ADVANCED GENERAL RELATIVITY ‘An intodocton to Regge Theory ad Wh Enepy Physice PD. Coli JOHN STEWART (uri, Ghnne and Later ME Create ulativity on Corse Space” ide Paice and CS, Chrke Deparment of Applied Mathematics and Theoretical Physics Sipermanfole BS. DeWitt, University of Cambridge Superstrng Theory, vol. 1 Introdurtion MB. Gree, 3.H. Schwars and E, Witten ‘The Interacting Bosor-Fernion Model P Tne and. Va cer ic Sltins of nis Pld xan D, Kramer, H, Soh MAI Liquid Meta: Coreg ond Theory, Mach Gigy Fel Thaori§.Potoraks Path ter Sethe in Quantum Field Tory RJ. River Gravitational Phi Sle and Galt Systeme WC, Slew Advanced Genera Relativity JM. Stewart i CAMBRIDGE URIVERSITY PRESSTeeter erate “Since eens” Resa Gene en ay af Congen Casbig i Paiasn Dui aie cal es fr hi ok ali fm be Bash Ly CONTENTS Preface 1 Differential geometry: 11 Dilferentiable manifolds 1.2 Tangent vectors and tangent spaces 1.3 Tensor algebra 14 Tensor fields and commutators 1.5. Maps of manifolds 1.6 Integral curves and Lie derivatives 1.7 Linear connections 18 Geodesics 1.9 Torsion and curvature 1.10 The pseudo-riemannian metric LIL Newtonian spacetimes 112. Special relativity 1.13 The general theory of relativity 2. Spinors 21 Introduction 22° Spinor algebra 28 Spinors and vectors 24 The Petrov classification 2.5. Spinor analyst 26 ‘The Newman-Penrose formlism 2.7 Focussing and sheating 28 Goldberg Sachs etal 2.9 Plane and plane-fronted waves u“ a 25 29 32 0 6 52 55 6 82 a 76 80 89 96 100 1053 Asymptopin 31 32 33. sa 35. 38 37 38 3.9 Introduction Asymtopia for Minkowski spacetime Asymptotic simplicity Conformal transformation formulae Seri and peeling ‘The choice of conformal gauge A spin basis adapted to. 7 “Asymptotic symmetey Spin-weighted spherical harmonics 3.10 Asymptotic solution of the feld equations 31 ‘The Bondi 4-momentum covector 3.12. The positivity of the Bondi mass 4. The characteristic initial value problem 41 42 43 44 45 46 ar ‘Quasilinear frst order hyperbolic systems of ‘The Cauchy problem ‘The characteristic initial value problem ‘The conformal vacuum field equations of general relativity Lagrangian and Legendrian maps Wavefronts and wavefront families Caustics in Minkowski spacetime Appendix A Dirac spinors Appendix B_ The Newman-Penrose formalism References Indes ns ns ne 12 124 wt at 133 130 us 149 155 160 16T 187 113 at 1s 193 108 201 a3 216 22 25 PREFACE General relativity is the flagship of applied mathematics. AL ‘though from its inception this has been regarded as an ex (raordinarly difficult theory, it isin fact the simplest theory to ‘consummate the union of special relativity and Newtonian grav- ity. Einstein's ‘popular articles" set a high standard which is now emulated by many in the range of introductory textbooks. Having mastered one of these the new reader is recommended to move next to one of the more specialized monographs, eg. ‘Chandrasekhar, 1985, Kramer et al., 1980, before considering re view anthologies such as Binstein (centenary), Hawking and Israel, 1979, Held, 1980 and Newton (tercentenary), Hawking and Israel, 1987. As plausible gravitational wave detectors come on line in the next decade (or two) interest will focus on gravitational ra dintion from isolated sources, e.g. a collapsing star or a binary system including one, and Thave therefore chosen to concentrate in this book on the theoretical background to this topic. ‘The material for the first three chapters is based on my lec ‘ture courses for graduate students. The frst chapter of this book presents an account of local differential geometry for the benefit ‘of the beginner and as a reminder of notation for more experi> fenced readers. Chapter 2 is devoted to two-component spinors which give a representation of the Lorentz group sppropriate for the description of gravitational radiation. (The relationship to the ‘more common Dirac four-component spinors is discussed in an ap~ pendix.) Far from an isolated gravitating object one might expect spacetime to become asymptotically Minkowskian, so that the de- scription of the gravitational feld would be especially simple. This concept, asymptopin (asymptotic Utopia) is discussed in chaptervii Proce 3, commencing with an account of the asymptotics of Minkowski spacetime, and going on to the definitions of asymptotic flatness and radiating spacetimes. (For a more detailed development of the material in chapters 2 and 3, the reader is referred to Penrose and Rindler, 1984, 1986.) The book concludes with a self-contained discussion of the characteristic initial value problem, eaustics and their relation to the singulasity theory of Arnold Exercises form an integral part of each chapter giving the reader ‘a chance to check hie understanding of intricate mater fering straightforward extensions of the mainstream discussions. Problems are even more important, for they are not only more challenging exercises, but can frequently be combined to produce significant results, encouraging the student to develop his under- standing by deriving much material which is not explicitly spelt ‘out, The brevity of this book is deceptive. Finally T acknowledge the considerable benefit of with many of my colleagues, especially Jiirgen Ehlers, Bernd Schmidt and Martin Walker. In particular I owe especial thanks to Helmut Friedrich for teaching me (almost) all I know about the characteristic initial value problem. John Stewart 1 DIFFERENTIAL GEOMETRY ‘The natural arena for physics is spacetime. As we shall se later spacetime is curved. It is necessary therefore to introduce a fair mount of mathematics in order to understand the physics, For- tunately we shall only need a local theory, so that problems from differential topology will not occur. 1.1 Differentiable manifolds ‘The simplest example of a curved space is the susface of a sphere 'S?, such as the surface of the eatth. One can set up local coor- dinates, eg., latitude and longitude, which map $? onto a plane piece of paper R*, known to silors as a chart, Collections of charts are called atlases. Perusal of any atlas will reveal that there is no 1-1 map of S* into RY; we need several charts to cover S*. Let us ate this more formally Ce A) peristrion Given « (topological) space M, a chart on M is a 1-1 map $ rom an open subset U CM to.an open subset @(U) CR, icy @ map $:M — RY. A chart is often called a coordinate system. Now suppose the domains U,,U, of two charts 44.45 over- lap in U,U,, Choose = point 2, in 6,(U, M'U;). It corresponds toa point p in U, Uy, where p'= 4, (2), Since p is in U, we can map it toa point 2, in $,(U;). We shall require the map 2, ++ 2, to be smooth, see fig ‘ure 1.11, s0 that in the next section the definitions of smooth2 1 Digeretiat geometry Fig. 1.11. When the domsine of oo share and ove they ave hit fo mah sur bes oy ogy") mat be snot curves and tangent vectors will be coordinate-independent. More precisely: (1.12) pernion Two charts 4,6, ere C™-related if both the map 206,77 + 8(U,OU,) + G(U NU) When problems of differentiablity arise we can similarly define (Chrelated charts and C*-manifolds. ‘The reader for whom these ideas are new is strongly recom- mended to peruse a geographical atlas and identify the features described above. Farther examples worth examining include the plane £2, the citcle 5", the sphere S® and a Mobius band. 12 Tangent vectors end tangent spaces 3 eee x IST Fig. 121 A curve A: — Mt MU) is smooth if is image under a ehart In amoath cur in B, 1.2 Tangent vectors and tangent spaces ‘Most concepts in physics involve the concept of differentiablity, and, as we shall sec, an estentil ingredient is the generalization of the idea of a vector. Simple naive definitions of vectors do not work in general manifolds. ‘The vector London — Paris may be parallel to the vector London —+ Dublin in one chart and perpen- ‘dicular tot in another. Some experiments will show that there. severe problems in establishing chart-independence for the usual properties of vectors in any non-local definition. The following ap- proach may not be an obvious one but it will capture the intuitive concepts. We start by defining a curve within a manifold (1.2.1) pernurton A G*-curve ina manifold M is a map of the open interval 1 = (a,8) € R + M such that for any chart 4, 60:1 — RY is 2 mop. There are a number of points to note about this definition Firstly C*-curves are defined inthe obvious way. Either or both of a,b can be infinite, By considering half-losed or closed intervals Tone of both enclpoints can be included. Finally the definition4 1 Digerentio yeometry ae) Ht) a 3 ge8 c Fig, 1.2.2 Two carves (0), ls) ane tangent at pF tei mages under & ‘hae @ re tangent a ep) ia implies that the curve is parametrized. As an abbreviat often speaks of the curve 2), with £ € (2,8) Let f: M — R be a smooth function on M. Consider the map Jod:1—» Ry tm f(A(e))- This has a welldefined derivative, the tate of change of f slong the curve. Now suppose that two curves A(t), (2) mect a a point p where 5.» Suppose that 4 (2a) 4 a fuey= Fen) ot for all functions f. ‘This is « precise way of stating that “A, pass through p with the same velocity”. To see this we cons (122) 0, Suppor that any chart wv domain of dependence ince p. Let @ map q.€ M to.2'(q) where 2'(g) are the coordinates of a Tncn (bi) ha ony for cr and ch 4 gi =[Leen] a2) ‘Proof: It is trivial to show that (1.2.1 implies (1.22). ‘To proceed in the other direction write fo. = (f09-!)0(G0). Now foo’ 1.2 Tangent sectora and tangent spaces 5 is a map R* ~+ Ry 21+ f(x!) = f(-"(z")). Also $0. is a map I+ RY, tvs 24(2(0). Using the chain rule for dfferentiat ger => [gue] oe (123) AA similar expression holds for f © 4 which then proves the result ‘Thus given a curve A(t) and a function f we can obtain a new number [a(f ©A)/dt],.,, the rate of change of f along the curve AC) at ‘We now use this result to remove the non-locality from the idea of a vector. (1.23) vermirion The tangent vector i, = (dA/d), to a curve X¢) at a point P om it isthe map from the set of veel functions f defined in « neighbourhood ofp to defined by Sere [Ziroal] £0), = 4,4). (1.24) Given a chart with coordinates =, the components of i, with respect to the chart are oa,= [Foo], The set of tangent vectors at p is the tangent space T,(M) at p ‘This accords with the usual algebraists! idea of vectors, as we sce from the following result (1.2.4) tmeones If the dimension of M isn then T,(M) is a vector apace of dimension n6 1 Different geometry Proof: We shove first that 7,(M) ise vector space, i. 7,(M) and a isa real number then X,+¥%, aX, eT,(M). In other words we have first to show that there is a curve v(8) through p, t= t, such that 8) = (Lou), =XyF + Yoh (125) At) a curve in RY (where Let A, w be curves through p with A(¢,) ji, = ¥,. Then o: tr GoA+50p~ 0) "= have their wsual meaning) and v: t+ 9” ‘M satisfying (1.28). ‘The second part of this proof is left as an exercise for the reader. Finally we have to show that a basis of 7,(f) contains m vec: tors. We first establish a useful result. Let g be a chart with ‘coordinates 2, Consider m curves Ay(t) defined as follows 9 (Ad) = (2(P)s 2) 2hP) + Ps --92"(P)) sel the bth corinat vss, We dnote he angel veto Spetoty §40)= (23), - 128) [Note the simple result (zo A,)'= 4,. Then using the chain rule (1.23) we have Ge), 4 Gord, =F soe o(ordl, Spelt oy Semon) +9 yan =[eayoe" : aS heuer yee [Faure 02) 1.2 Tangent sectors and tangent spaces 1 [Next we show that cach vector at p is linea combination ofthe (0/0s*),. To see tha let X, bea tangent vector to the curve A!) with 46) = p. Then = (for) = [fos oso.(0) = ES daree Veo) E (Ga) 1 es (128) where (1.27) has been usd in the last step. Thus x, D9 (ja), (129) ‘and so the (6/02), span 7,(1). Finally to show that they are linearly independent suppose that 5 A* (0/02"), = 0. Then yay = ‘Thus each A‘ = 0, ie, the (0/82), form a basis. From (1.2.9) we see that X,2" are the components of X, with respect to the given basis, ‘We shall subsequently use the Binetein summation conven- tion: in an expression where the same index occurs twice, once up ‘once down, it is to be summed. (In a fraction up in a denominator counts as down in the numerator and vice verse.) ‘Thus 54a), -4'(z), WARNING: Do not confuse the diferential operator acting on functions in R*, 8/8z* with the vector (9/3s*), in 7,(M) ‘The final result needed to capture the concept of tangent vector as a derivative islets an exercise8 1 Different yore (1.85) Beercise. Suppose that fg are two real functions in Mf and fg: M —+ Ris defined by fo(7) = fla)o(o). If X, € TCM) show that X,(fo) = (X fate) + 10V(%,9), (1.2.10) the Leibniz product rate 1.3 Tensor algebra Now that it as been demonstrated that 7, (3M) is a vector space it is possible to incorporate most of the machinery of linear algebra. (1.4) perisniox The dual space T3(M) of T,(M) i the veetor space of linear maps X: 7,(M) — B ‘Ty(M) has also dimension x. If acts on X, € 7,(M) the result is written either a A(X,) of < 2,X, >, scalar. Blements Of T3(M) are called 1-forms, forms or covectors. Lat f be any function on M. For each X, € T,(M) X,f is a scalar. Thus f defines a map df :7,(M) ~via WX) =F (sa) Since X, is linear so is df and so df € differential or gradient of f S(t}. df is the (1.3.2) Lema, Let § be a chart with coordinates 2'. Then the coordinate dif- ferentials dz* form a basis for Tj(M), the dual basis, with the orerty ((S),) 65. (132) 13 Tensor elgeba ° Proof: Using (1.3.1) and (1.2.7) we have (.(6e),)= (=), Fw OF Yoo ‘This proves (1.8.2). Next we show that we have a basis. Now. = o(e(&:))=0 = ‘Thus the de! are linearly independent, Finally consider the form w=a-(u(fa),Jar a “(5s),) ~((@z),) (#**(6),) for all m. ‘Thus w= 0, an so every element of T3(Bf) isa linear combination of the ds ‘We thus see that every I-form can be written, ve(v(le) ur an In particular if we set = df (+($5),)= (Gx) rt where we have used (1.2.7), and the shorthand f, = 8/02", This is the justifieation for the name gradient for df-10 1 Digerentat gometry Next we ned to consider the posit of coordinate transfor mations sa y'e"), Using the above revult gam so (ov ) date vie (22) “This given the rs for» change of basi in 73(M).1F(2/82"), isthe daa basis in Z(H), (182) implies that (Ra), =e Ga), De = (5pm), 8) ‘The positions of the indices should be noted Equations (1.3.5),(1.3.0) ate easily generalized. Let (c,) be any a 5) basis of 7,(M) and («*) the dual basis of Ty(M), i uty rahe (3.7) ‘Then the formula for a change of basis u* ++ O*, e, -€, is A) fey, (13.8) where A*, is # non-singular mat Suppose the Iform A has components A, with respect to the basis o*, Le, Az aet = A,o= Tat ‘Then since the uate linearly independent A (139) Fes ax! (1.3.10) Note carefully the similarities and differences in these rules Quantities which transform like (1.8.9), (or (1.3:10)) are called covariant (or contravariant) in old-fashioned textbooks. (1.8.8) perivttow (1,2) tensor $ on T,(M) is a map Ss T,(M) x T,(04) x T5(M) + Ry which is Linear in each argument. 19 Tensor algebra u A (1,2) tensor is determined by ite components S.y° with respect to a basis. Suppose (,),(«") are dual bases. Define Slegyesye*) ER. WX = Xte,, ¥ = Vey, A= Aw" ane two vectors and a form S(,Y,9) = 8(X%e,, Yep ho) XVASlesseyw") by the linearity VAS ‘These properties generalize emily to (rs) tensors Notice that two (r,s) tensors can be added and the product of fa scalar and a (7,4) ensor ie also a (7y4) tensor. ‘Thus the sat of (4) tensors forms a vector space. However the sum of « (0,1) tensor, (a form) and a (1,15) tensor is not defined. (1.84) Beercive. What is the tensor transformation law cor responding to equations (1.3.9),(1.3.10)? Suppose that the components of two tensors are equal in one basis. Show that this is then true for every other basis, (1.8.5) Brereise. Suppose that B,, has the property that when- ever C.," isa tensor then 20 is Ajay’ = ByC,*. Prove that By, is tensor and generalize the result to tensors of arbitrary order or valence. Suppose that 5 is (1,2) and 7 is (1,1). Consider the map UTM) x TM) x Ty(M) x TMM) x Ty(I) +R, U(X, YA, Zn) = SKY, AIT 0), ‘This is clearly linear in every argument and so is a (2,8) tensor with components Oye = SyoT ‘There is another tensor operation to be considered which re- ‘duces the order of a tensor. Suppose U is (2,3) with components Uy Let wey2 1 Digferentiat geometry (1.3.6) Bzercise, Show that this defines a (1,2) tensor V, the eon traction of U across the third and fourth indices. Since this definition was given in terms of components itis necessary to verify rot only linearity but also independence of the basis chosen ‘Two other tensor operations will be frequently used. Suppose Sis (1,2) with components $,°. Define Seas = HSS + Su° Sa? = HS" — Su! called symmetrization and antisymmetrization respectively. "There are obvious extensions ©. q, ins ~ Tae ~ Thad Ate Ta + Tau Ton [Note that these operations can be applied to subsets of the indices of higher rank tensors, A tensor which is symmetric on all ofits indices is symmetric, A tensor which is antisymmetric on all of its indices is antisymmetric or skew. (1.8.1) Beereise. Let M be a manifold and f: M— Ra smooth function auch that df = 0 at some point p € M. Let 2! be a coordinate chart defined in « neighbourhood of p. Define = Ot «> Beat By considering the transformation law for components show that F,, defines a (0,2) tensor, the Hessian of f at p. Construct ‘also a coordinate-ftee definition and demonstrate its tensorial properties. What happens if df #0 at p? F, 14 ‘Tensor flelds and commutators Until now we have considered only the algebra of tensors defined at one point, Next we consider the union of tangent spaces 7,(M) ‘at all points p € AL TUM) = Y 700), 14 Tensor elds and commstetors 18 (1.441) permarion A vector field on M ise map X + M > T(M) such that X(p) X, i @ vector in T,(M). ‘Thus a vector field isn specification of vector at each point of the manifold. Given a coordinate system 2" and associated basis (0/82, for exch T,(M), X has components X!, where wxi(Bs). mt eK) (Here X,(2!) means the vector X, acting om the function 2*.) Although it is not entirely unambiguous itis conventional to drop the sufix ‘p'. X is called smooth if the functions X* = Xe") are smooth for one, and hence all, charts in the atlas. Analogous definitions hold for eovector and more general tensor fields Now suppose that X, ¥ are vector fields. We have already seen that for any function f, ¥,f is a number foreach p, the tangential derivative of f slong Y at p. Thus there exists « function ¥ f MR, Yf(p) = Y,f. Since Y/ is a function we may compute (5x) 27=* (t2),0"a), x1 (oe),+* (Fe), (65), ‘This computation shows that there exists no vector Z such that Z,f = X,(Vf). For if there were then, XN) xorn=2r=2'(36) . But we have just seen that X,(¥'f) involves second derivatives of, J: However consider the possibility A XUN) ¥,0Cf) for all f“ 1 Different geometry ‘The right hand side is vm ($8) 7 (2), ‘Thus X, ¥ define a new vector field, the commutator [X,¥] such that [GYLF= X07) -%,001) forall f It has components [GY] = XY AYP ge ‘The commutator has an important geometrical interpretation deseribed in section 1.6. (1.4.8) Beercise. Show directly from the definition that boayl=-Mx, | KZ] = 221+ and prove the Jacobi identity WY), 21+ 0% 21.81 412.241 1.5. Maps of manifolds In this section we consider maps from one manifold M to another IN. In the most important applications M and IV are the same nif. (1.8.1) pernetion A mop h: MN i C™ if for every O™ function f 1 NR the function fh: MR is also O™. 15 Maps of manifets 8 Fig. 15.1, hs Af = W maps curves in M to curves in N The pushforward ‘Smaps tangent vectors of crv nM 40 tangent vets of eaves Now suppose that h: M + N is O*. Let y be a C™ curve in M. Then h(y) is a C™ curve in N and h can be weed to map tangent vectors over M to tangent vectors over NV. (1.5.2) verivrrion Given @ C™ map h: M+ N there is @ map h, : Ti(M) — Tyy(N), the induced linear map, forward which ‘mags the tangent vector of a curve 7 at p € M to the tengent vector of the curve hy) at A(p) € N. Of course this definition, while highly intuitive, is of little use for computational purposes. For those we need a further result (1.5.3) uesaea For each O™ function J: N+ R OX =X (0h, (151) and h, is a linear map.16 1 Disferentiayometry Proof: Let 7 be a curve through p with (0) From the defini (XM = [f (ho MN] (0) =F 2) 9-9}°(0) = XylF 2h) Purthermore the right hand side is linear in X, and so therefore rust the Teft hand side be. 410) ‘Once we have a map between vectors on the two manifolds we can construet a map between covectors on the two manifolds, ‘There is however a slight surprise; the map is in the opposite (15.4) perininion ‘The mape hi: M = N, h, + T{M) ~ Typ (N) induce a near map, the pull-back kr : Tyyy(N) > Ty(M) as follows: ifw € TrylN)y X,€T,(0) ten (h0)(X,) = (KX) (15.2) [Lot us now consider « concrete example in order to show how fons (15.1)(15.2) yield explicit formulae for the maps h, Let Mf be R? and NV be R? with coordinate systems (e,yse)s (20) respectively. Let A(2,y,2) = (2y)- Then for any ENoR Fo h(aus2) = Sey): ts (X,Y, Z), and g is function on M sa (8iEV 8), IX, has comp uren=(xbL a2) =a We see that A,X, € Tyy (lV) has components (X,Y). Finally if & € Typ (IN) has components (2,1), i, = Adz + dy, then (8,X,) = AN 4 AY = BU) u ‘Thus hy € T3(Af) has components (2, 1,0), Bu = Ma + pdy +0dz. (1.5.5) Beereise. Carty out the corresponding caleulation when M= BR and N= Ri (1.5.6) Beercise. Let h: MN be a smooth map between two manifolds M and IV. We may extend the definition of h* to map, fanctions f: N — R to functions h* f : M— R via Mf=foh. Show that A* commutes with the differential operator, ie wa) = ae"). 1.6 Integral curves and Lie derivatives Given a vector feld Xan integral curve of X in M isa curve 77 in M such that at each point p on 9, the tangent vector is X,. Existence and uniqueness of such curves is guaranteed by: (1.6.2) THEOREM, Beery smooth vector field X defines locally a wnigue integral curve 7 through each point p such thet 9(0) = Proof: Let 2* be a local coordinate chart and let} be the coordinates of p. ‘The equation of the integral curve is ue with the intial conditions 2'(0) = 2}. Provided X* is smooth the standard theory of ordinary differential equations guarantees the existence, at least locally, ie, for small of precisely one solution em"), [A very uefa result for what follows is1B 1 Differential geometry (0.8.2) Lema IX is @ smooth vector field such that X, # 0, then there exists © ‘coordinate system y' defined in « neighbourhood U of p such that in U, X= 9/0". Proof: (We give the proof in 4 dimensions; the generalization is obvious.) By continuity there exists some neighbourhood U of p, in which X #0. We choose a S-surface © in U nowhere tangent to X, with coordinates (y?,g°,y"). There is a unique integral curve of X through each point of B. Define the fanction y! along each such curve by y! = fdt, with y! =0 on Z, and the funetions Fuysy" to be constant slong the curves. ‘Then (y') is the required nate system. ‘Am integ ‘curve o(t) is complet of t. A set of complete integral curves is a congruence. A smooth vector feld X defines a transformation fi, : M —+ M for each reals as follows. For each p € M, h,(p) isthe point on the ‘unique integral curve of X through p at a parameter “distance” s from p, i. if (tis the curve and pis 4(,) then h,(p) is 1(t,+8)- If should be obvious that as + varies these transformations forma al rameter Abelian group, for identity, (2) woe now possible to give the geometric interpretation of the commutator mentioned in section 1 (1.63) rweonese Let X,Y be two smooth vector fields in M inducing groups (h(E). Eel p EM, and define g = kg(p)r = hale hele = kyle). Then if (m,) isthe group indeed by [X,Y] gyal) + O(a? als) in the sense that this equation holds true for any coordinate system in the atles. 1.6 Integral cares and Lie derivatives 19 Fig 1.61. The geometrical interpretation ofthe commutator. Moving do in the dection X for flowed by ty the divecion F tot in ot the sem Imoving along ¥ to lionel by te slong X tou, ea be each fom Gy moving dd log the commatn Proof: Introduce a local coordinate system 2! with p at 2! ‘Then considering integral curves of X waste (2) a0 (@ xpd (BE) at pone = Xpdet UMN), det where --- denotes third order quantities. Similarly Sat Yi dt+ 1Y,Y") dt ++ B24 [YF Ox) a] d+ OY), de = Khas Mp dt KN), dehy (7K) dade 107 Y) a ‘The corresponding formula for x, is obtained by interchanging X and ¥ and s and ¢. Thus VAX" =X" Y),dadeg > X,Vhp ded +20 1 Disferentat yometry ‘To this order we can evaluate the commutator at u instead which proves the result, Thus pywv forms a small quadrilateral if the commutator vanishes. ‘At the moment although we cen diferentiate functions we ean- not differentiate vectors of covectors. Differentiation implies a comparison between values at two spacetime points and we can- thot do this for vectors, Because the two vectors will ic in different ‘vector spaces, However given an additional smooth vector field X there is a comparison which we can make, leading to the definition of s derivative the Lie derivative (1.6.4) pernation Let p be a point of M and let X be « smooth vector field. Let 7 be the integral curve of X through p inducing ¢ I-parameter group of transformations (hy). Then if f MR is any real function on M, the Lie derivative of J with respect to X is ‘This definition, while exhibiting manifest coordinate invariance, is of little use for computations, We therefore establish: (1.8.5) THEOREM For any function f on M, (E52), =X (16.2) Proof: We use the coordinate system adapted to the integral curves set up in lemma 1.6.2. The coordinates of hy(p) ate v(k(P)=V@) + — ¥*Uhle)) = 9") 23, ‘Then from the definition (ExDy 16 Integral curses and Lie derivatives a where lemma 1.6.2 has been used. ‘Thus equetion (1.6.1) ie true his coordinate system, But being a scalar equation itis true Now let g = h_y(p) 50 that p = hy (a). Using the results of section 1.5 we have an induced linear map (h,,). mapping 7,(3) to 7,(M), and 40 we may compare vectore at ¢ with vectors at p. This leads to the notion of s derivative for vectors. (1.6.6) pernvrriox Using the notation of definition 1.6.4 the Lie derivative of a smooth vector field 2 with respect to X at p is 2, ~ (had (Ex2)p= | 4 well defined vector a p Once more we need to establish a suitable formula for compu- tations, (1.6.7) THroREM (£42), = 1X1, (18.2) Proof: Again we use the special coor congruence. Then, letting a = 2,3, rate system adapted to the (ean = [Pee =z fe = ate dt Li = (= Now the right hand side is uot a vector; see, ea the diseussion in section Ld. However suppose that we subiract Z,(*) from the2 1 ier gometry right hand side, (Since X" = 64, this is zero and we are free to do it) Then (x20 (X, 2}. Now both sides are vectors and so the equation holds in all coor- dinate systems since it holds in our special one. (Ci, exercise 1.3.4 again.) XZ) - 2(X) 1p (hy)" to define Lie Aervative of cows, I practice we can se shortcut, bed fon the requirement that a derivative mast obey the Leibniz rule. ‘Thus let w be a covector. Then for any vector Reld Y, o(¥) ie & function. Therefore in an arbitrary coordinate system, 2x (0) = X (0) (oD) .X" (or), wg VAX! + wy YEX We can now use the induced linear However we must requite the Leibniz rule to hold, i Lx (00) = (£xH)¥) + (ERY) (2x0) Vi 4 o(Ex¥ = (£xu)¥! YE X= KEY ‘Thus (Exe)! = 0.8K! $a XR ‘This has to hold for all ¥ and so we sce that the local components of the covector Lu are eX eX (163) (1.6.8) Bzercise, Write out formal definition along the lines ‘of definition 1.6.6 and develop the equivalent theorem to verify equation (1.6.3) (1.6.9) Beercise. Compute the components of £T where T is 8 (0y1) tensor. 16 Intel cares oud Lie derivatoes 23 (1.6.40) Beercise. IF X,¥, 2 are smooth vector fields show that fun byl ~ Ly ExZ. [Hint: at last a use for the Jacok identity!) (1.6.11) Ezereiae. X is a vector field on a manifold M and f : M + Ris a function. Show that Lie differentiation commutes with the differential operator ic. Exdf = d(£xf) Demonstrate also that Lie dilferentiation con symmetrization and contraction of indices. tes with (anti) (1.6.12) Beereise. Let 2',i = 1,2 be a Cartesian coordinate chart on R'. ‘The vector field’ X has coordinate-induced components X* = (2,21). Construct the integral curves of X and sketch them. Let be the (0,2) tensor field with components diag(1, 1). Show that fun ‘The finite transformations induced by translation along the inte- aral curves of X are seid to preserve 1, In the context of special relativity they are called boosts or Lorents transformations, (1.6.18) Bzercise. Let 2 be a Cartesian coor and let ny = diag(1,—1, -1, 1) be the usual Minkowski metric of special relativity. "For the purposes of this exercise may be assumed to be a (0,2) tensor. Let X be e vector field on M with, associated I-parameter family of transformations hy. Show that X preserves 9 if it satisfies Killing’s equation Je chast on R', Xn) = NeXt Assuming for the moment (or refersing to exercise 1.10.12) that ‘his implies that each component X* is a linear function of the a, derive the general solution, and hence obtain the Poincaré invariance group of special relativity. ‘The Lie derivative is one of the most important and least un- derstood concepts in nuatlrwtical physics. As final example in4 1 Disferntia geometry this section we consider Iinearized perturbation theory on mani- folds. In what follows Af = M, is & 4-dimensional manifold, but wged to n with only minor arithmetic changes. Let wal smooth manifold containing a I-parameter family of smooth non-intersecting 4-manifolds M,. The mas folds M, are to be thought of as perturbations of a specific fone, say M,, If one manifold is a replica of another we need ‘point-identification-map which relates points in the (¥o mani folds which are to be regarded as the “same”. We supply this by introducing a vector field V on NV which is everywhere transverse ive, non-parallel to the Af,. Points in the various M, which lie on the same integral curve of V are to be regarded as the same, The fof V is chosen 40 that the induced I-parameter group of transformations h, maps M, to Mf, In old-fashioned notation © choice of V is « choice of gauge. ‘Now let Q, be some geometric quantity on M and Q, the cor- responding quantity on Af,. This defines « field Q on N. Along teach integral curve of V we have a Taylor series OHH [Qe (L. Qe + UIs where « is small, The first order term is usually called the Fin- carized perturbation of Q. In general Q will satisfy some complicated nonlinear equation, but ifall quantities are expanded as Taylor series in ¢ and nonlinear terms are discarded, the restlt- ing linear equations for the perturbation may be both simple to solve and relevant. There is however a problem, because there is no preferred choice of V. (1.6.14) Problem. By considering the dilference between two choices for V show that the linearized perturbation of Q is gauge-invariant i £5 Q=0, for all 4-vectors X on Me. Verify that this is the ease ifone of the following holds: 4) Q, vanishes identically, fi) Q, ie a constant scalar field, LT Linear connections 2% 38) Q, bs tinea combination of prod , Son of prodscte of Kronecher dees with constant coefficients. 7 1.7 Linear connections ‘The concepts of parallelism and differen cial to classical mechanics ‘a curved ma ion of vectors are eru- _mcchnien How are we to introduce them in ; fold? ‘The Lie derivative is not sulficiently general since it depends on the choice of a vector field X. Tt turns out that the clearest way to introduce the new structures is axiomatically. (1.1.1) vernion A linear connection V on M is @ map sending every pair of smooth vector fields (X, Y) to a vector field Vy¥ auch that Vx(eY +2) =aVy¥ + VxZ, for any constant scalar a, but VU) IVY 400% when f is « function, and itis linear in X VasivZ = Vy Z + £Vye Farther, acting on functions J, Vx is defined by Vs. Xf. g¥ ts called the covariant derivative of Y with ; native of Y with renpect to X! Becste Va¥ ional linar in, V i ot enor, owner Uke the Lie derivate Ey, UV is ner ints deity @ (1s1) tensor VY mopping X to Vy¥, calle Ope 9 ing xY; called the covariant26 1 Digerenta yomery Let (¢,) be a basis for vector fields and write V,, as V,. Since Vey is 8 vector there exist scalars T%,, such that (Note carefully the (conventional) post Te, are called the components of the connection. Now ¥e, ete Then, using definition (1.7.1) suppose that X Va¥ = Vee, (M) = x*9,(%%) Xe AAV eY! = XO Y Ye, (una) ‘Thus Vx¥ is completely specified by giving the components of the connection, We may also write (17.2) as (Wa) S [eI 4 EYEE LX" say. (173) YS, are the components of the (1,1) tensor VY. Neither of the ‘90 terms in Y", transforms like the components of a tensor, but their sum does. in a coordinate induced basis (74) Before we give an example we need to introduce a new concept In Buclidean space if we move a vector without changing its mag. nitude or direction itis ssid to be parallel transported. This ean be generalized as follows: (1.7.2) periwmion I} Vp¥ =0 then ¥ is said to be parallelly transported with respect to X 127 Linear connections a We now consider « concrete example. Let M be R? with Carte- sian coordinates (z',2*), and e, = 8/0z". ‘The normal geometric concept of parallelism suggests that ¢, is parallelly transported ‘slong eq and so Pigg =O. Next consider polar coordinates (7,9), ¢, = 0/9r, ey = 7-8/0, ‘These basis vectors are parallelly transported along e, and s0 Ve, = Vie ‘They change direction when transported along e,. In fact «result from elementary vector caleulus shows that Yee, ‘Thus all of the connection coeficients vanish except Maat teed ‘his example illustrates the fact that the connection compo- nents cannot be the components of tensor. Por if they wee they would vanish in all frames i they vaished in one. ‘The general Ueansformaton law i given bythe flloving theorem (1.7.3) tuoReM Consider a change of basis, Sey. Then BEB! yy + (D> Botey(By). (1.75) By BJe, Vos Bi'e,) = WB e, + BeBe, F HOUT ey + Beg Beg,8 1 Disfeentia geometry which proves (1.75). The first term on the right corresponds to the usual tensor transformation law. cx the example above setting @, = €,5 1.04) Brie Cs ed '4¢ explicitly using formula (1.7.5). =e, and compute (1.75) Bzercise. Suppose that V,V* are two linear connections fon AM, Show feom definition 1.7.1 that DIXY: (Wx¥ -95¥)s fs a tensor. Compate also the components from (1.7.5) and show that they transform correctly. “We can now extend Vx to tensors via the Leibniz rule. Suppose that 7 is @ covector, Then 7(Y) is a function for each vector Y. Hence Vx (n(¥)) = X (9()}- But from the Leibniz rule Vx (af ¥)) = (VxM(H) + VAY and s0 V7 is defined by (Va) = X (WY) = ACV) (1.7.9) for all vectors ¥. We can evaluate this in component form by actting X = ¢,, Y= Ye, 9 = 7.0% where (u*) i the dual be ‘Then (1.7.6) becomes (Vent = elm) — ne (VAeN) (Syme = eal) ~ Psat (7.7) In particular veh Phe (7g) and for 8 coordinate basis (Fonda = May = Mea Pt (179) 18 Geodesice » (1.7.6) Brercise. Extend the defiition of V to tensors, Eig. show that if is a (1,1) tensor (I = (PY HP TY PAT (1740) 1.8 Geodesies In lat space a geodesic isthe shortest distance between two points, i.e a straight line, It has the property that its tangent vector is. parallelly transported along itself ifthe tangent vector is X then VxX = 0. We now generalize this concept to a manifold, (1.8.1) vermrion Let X be a vector field such that VxX = 0. Then the integral curves of X ore called geodesic: (1.8.2) TuoREM There ia precisely one geodesic through a given point p € M in @ given direction X,. Proof: Choose local coordinates 2'. Let the tangent veetor to some curve be X*. Then the curve is given by F0=KeO), 2O=s4 ‘The curve is geodesic iff 02 (VX) = XX = ED XEN But {,X4= Xf = df/dt and so a ‘am XA% 1 Dieta some Ee es) (a fetya-rcan (4) ( which oe ved abject he ial ate ey ys soe (4-6 ferential equ Dy the standard theory of ordinary diferent haton este ocaly for sal rales of nd i ign the Coretary. Locally we can always save (1.8.1) eabjest to the 2- point boundary consions 20) = 2} (1) = 24 where p and g Ere neighbouring points. ‘Thus locally there is a unique geodesic Joining two neighbouring points, In at space this statement is true globally but in a carved space it falls. For example the geodesics on a sphere are great cicles which intersect in pairs points. In a local coordinate system the equation of a geodesic is (1.8.1) B47 8" =0, vestigate the freedom in the (0); what choices ta! [ds ete, we have where 2! = da'/dt etc. We now i parameter &. Suppose that itis changed to fof # preserve the form (1.8.1)? Letting 2" (erate) (182) ‘Thus the standard form of the equation is preserved iff 3 = 0, ie s=at48, 0,0 constants |A parameter which produces the standard form (1.8.1) for the {geodesic equation is an affine parameter. An affine parameter {a defined up to a change of scale and origin. (1.8.3) pernation Tet» €M be ford. The exponential map at p sends each ne. for X, op to the point unit parameter distance along the unin 1S Gendesce a acodesic through p with tangent vector X,. In a small neighbour food of p the ezponential map has an inverse ie, there exists a neighbourhood U of p such that q € U implies q = e29(X,) for some X, €Ty(M). Now let (e,) bea basis for TM), and’ write 4 = eap(Xpe,). The (X}) re called normal coordinates of q and U is & normal neighbourhood Consider the geodesic through p,q. Since the tangent vector i X, the normal coordinates of any point r on the geodesie must be Proportional to X~. Thus the geodesic has the normal coordinate form XS, where s is some parameter. If tis an affine parameter we have (0) # 0. Now from (1.8.2) the equation of the geodesic is and since 5(0) #0 Since this has to hold for all directions we may deduce that Pig) = 0a p (1.8) ‘Thus we can always find coordinates, normal coordinates, for which the symmetric part of each connection coefficient vanishes at a point. This desirable property (1.8.8) would not be possible if the connection coeficents were components of & tensor. According to Newton’s laws a body moving under the action of no forces moves along a straight line. However if gravitation is taken into account there are no force-fice massive bodies. Ein: stein proposed treating gravitation as « geome jeal quantity as follows: a body moving solely under the attraction of gravity and tno other forces moves along a geodesic. Let x = 2” = (2!,2',2") the spatial coordinates and define »' = (t,2"). Then the 4-velocity ‘ofa particle is "= de*/ilt =(1,v), where v is the Svelocty, For2 1 Diferenta geometry {ational field with Newtonian potential ‘a body moving in a grav 6 0,-VA)=0,-8,)) eS LAS. Thus B+) ate from which we may read off the Newtonian connection as Pip =tar Paw $0 otherwise. (84) pence foal oor 1.94) Buereve. What the psi i er se eer. anal fe by snr ar = sect veto 1.9 Torsion and curvature [Although V is not a tensor there are two tensors closely related to the linear connection. (1.9.2) pertsrtion ‘The torsion tensor is « (1,2) ensor field T defined by TXY)= (VAY = VyX- GY), (18) Jor smooth weet fields XY. i in that it is not ob- his definition ie not entirely transparent the arguments, Since T(X, ¥) = ~T(Y,X) itis only necessary to check the firt argument, Teivilly TK + ZY) = MX) + TY): Now iff is «function (Byx¥ = VEX) - USI), UV KY — £9) X = (YDX = MGV F (VD), = fp) T (XY). TUX, 19) Tension ond eureatire 33 ‘To compute the components of T we introduce a basis (e,) of T,(UM). Since the commutator of two vectors is « veet dist scalars 7*,. = —7"4, such that levee) = a ceue (1.9.2) ‘The 7"), ate called the commutator coefficients. (1.9.2) Baercise. If the basis is coordi 8/82" show that 74, = 0, Verify also the RE discussed in section 1.7 ‘the example in aay, <1 =n ed Continuing with our computation of the components of T' we set X =e, ¥ =e, in (1.0.1) obtaining Tents) = Vatu Vie, ~ level BaP alter 0 that T has components T= Puy ~ Yu We have already seen that the geodesics fix only the symmetr part of the connection. ‘Thus the torsion represents that part tot determined by the geodesics. Since the ken here is that the connection is introduced to describe certain curves as geodesics, we shall assume that the torsion vanishes identically. Some non-wtendard theories, notably Einstein-Cartan theory attempt to attach physical significance to the torsion, ‘The second tensor related to V is considerably more important (1.9.3) pertution ‘The Riemann curvature tensor is a (1,3) tensor field defined by RXYIZ=VVZ—VyVyZ4Vxyi% (1.98) ‘al cach point. (It should be noted that some authors differ by an overall sign here.)a 1 Differential geometry RUXY)Z = Vy¥ 92 — Vpn VB + Vu HIV) WyZ—JOxVyS + NVA Yanso-wvnn® = FRUGYYZ In the second case we find R(X Y)EZ = V/V ($2) — VV y(F2) + Vox F2) = Vy (LV x24 (XZ) — Ve LPyZ HV N> Vix yF #UNYINZ = FRIXY)Z + (V(X) ~ XO + EGY) 2 RUXY)Z. ‘Thus R is indeed a tensor. Tn order to compute the components of R we choose a! and aetting X=.) ¥ = eq) 2 = in (1.8.3) we obt le.) Rlewegles = Vul¥.es) ~ VAP) + Vien VAT cep) — VAT aep) +1 ea 6 Ab es ell dee +o at PhP ye tals 6 We define the components of the curvature tensor by setting Re ea}ea = Rveatas ‘which implies that (0, WD" yy — Maal y+ eal ey Aa) EE) APP a alice Tal ae 19. Torsion end curvature 8 X» : Fig. 1.8.1 On parallel transport aroun the cose cusve par the vector Xp Changes ite ale In many cases the basis will be coordinate induced, so that the commutator coeflicients vanish, and we then obtain the standard formula = Fal as (295) where the vertical lines around the suffix F mean that it should be exempted from the antisynunetrization operation (1.9.4) Exercise, Let X bee vector field with components X* with respect to a coordinate induced basis 0/02". The components of WX have already been defined as X",. Let the components of VOX be Xa = (Xi )igs Show that XM = Rat . (1.98) the Ricei identity, ‘The Ricci identity has an important geometrical interpretation Let A,(t) be @ I-parameter family of curves passing through the point p(t = 0) in M. As » varies the points of constant ¢ define ‘curve in M denoted by A). Por # = 0 3,(+) degenerates to the point p. Let 7,5 be the tangent vector fields to the curves A.(0),A(a) respectively. Frou the geometsical interpretation of ‘he commutator given in section 1,05 follows that [7, 3] . Now3 1 Dieta geometry let X, be any vector in 7,(Mf), We may define a vector feld X at points on the curves as follows: for any point (2,4) propagate X paralelly along A,(¢) ftom p, ue, VpX = 0,X(p) = X,- Now let (da, dt) be the closed curve which starts at p, moves out dé along D,l8) tog = (dt) and then de along Ay(s) to r = (8-+ da,dé)y ‘aad finally back to p along X,,.,(0)- Let X,(daydt) be the vector in 7,(2f) obtained from X, by parallelly trensporting it around the Closed curve O(de,dé).” The difference X,(ds,dt) ~ X, is a wwelldefined vector at p which can be computed for infinitesimally Small ds,dé as follows. The paths pq and rp do not contribute to this difference because X was parallelly propagated along these curves. Thus the diference is (to first order) Xo ldadt) — X, =X, =X, = 4X (Phe best way to understand this equation is to write it out ex: pliitly in terms of coordinates as in the geometrical interpetation fof the commutator.) Purther since $= 0 at p, VsX = 0 there. For infinitesimal de it follows that along gry UjX = d¢V,VoX, Now since [2,5] =0 and VpX =0 V4¥ 5X = STIX. Therefore van [Xstdesdt) = %] ws fear] RSTO, ‘Thus R(S,T)X meanures the change in X after parallel trans port around a closed curve. This is intimately related to the fact that if 2 # 0 covariant derivatives do not commute, See also the following theorem, (1.9.5) THEOREM In a simply connected region U, the curvature R and torsion T ‘vanish if and only if there exists « coordinate system with respeet to which the connection coefficients venish throughout U. 19 Tosi ed curves a7 Proof: Suppose that there exists a chart such that T',,, = 0 throughout U. Clearly the components of R,T° vanish in this chart, Since they are tensor fields they therefore vanish in U. Conversely suppose 7; R vanish in a simply connected neighbour- hood U of a point p. Let (¢,) be a basis for 7,(M). Since R=0 ‘any clement of 7,(M) remains invariant under patallel transport around a closed path in U. Thus if q is any point in U, paral- lel transport from p to ¢ is path-independent. ‘Therefore the (e,) generate a basis at each point of U by parallel transport. Clearly = 0. If (o) is the dual basis of covectors it follows that 0. Now let y' be an arbitrary coordinate system, and set ehdy!. Then onVe se. =0 But since the torsion tensor vanishes P is Indices. Therefore ° ie = Ch Theta We may deduce that ina simply connected 7 v ‘connected region ¢!; is the gra- dient of a function, which we may call 2, so that a) = dz). Te follows that ¢, = 0/0z* and so the original bases were coordinate- induced. But by construction the connection coeficients vanished in this basis. TF the torsion vanishes there i another geometrical interpreta tion of which will later prove to be of grat sgniicence. Prt consider two coplsnae geese in Ry ey st affine parameter tet Z, be the lin jing points af the same {Ast varie the direction and length of 2, wil change nearly Even i the geodesics ace not coplanar Z, has constant. "elo ity” and aero “aceleraton™ This iden carries ove ret to 8 ‘manifold with a connection (1.9.6) perirnion Let X be the tangent vector too ongratnce of caves. Any ve tor Z such that |X, 7) = 0 is called © connecting vector of38 1 Digferentiat gometry Fig. 192) and p(t) ato ac Tahgent vector Bld Xp ann ‘ine parameters The “ronnecting sector ring geodenes of congruence with tsa euve joining points a the sane the congruence. (The reason for the name comes from the geometrical interpretation of the commutator.) (19.7) tHr0RE Suppose the torsion T vanishes, Let X be the tangent vector to congruence of geodesics, and let Z be a connecting vector for the congruence, Then the “acceleration” of Z, VyVyZ satisfies the geodesic deviation equation VV =R(X,2)X. (1.97) Remark, The significance of the name is as follows. Suppose 2y 4 fare two neighbouring curves of the congruence. Then there is ‘a small number ¢ such that exp(cZ) maps points of A to points of jt the same affine parameter. (1.9.7) then determines the “deviation” of the two geodesics. 19° Torsion and curvature 30 Proof: Since [XZ] and vanishes, UZ = V,X. Therefore VxVyZ= VV X VV yX + R(ZX)X = RUXZ)X, since X is geodesic, "The component form of (1.9.7) is ez ae saa X XZ, (198) where { is an affine parameter for the geodesics, Notice that only Reqag can be determined from experiments based on measurement of geodesic deviation Now itis obvious from the definition that Ry, ‘The Riemann tensor has further symmetries (1.9.8) Twrores Suppose the torsion vanishes. Then Rpg = (199) and Reece = 4 (Reon ~ Rane) (1.9.10) Further the Bianchi identities mnt = 9s (1.9.1) hol, Proof: Since each of these is a tensor equation it suffices Lo prove iin any one coordinate system, ef shall prove (1.9.11), leaving the other two as an exerc reader. In wormal coordinates, We for the 0 smal coordinates we may strange that T,0 1 Digferentatyometry at a point p. Since the torsion vanishes Mjyqj = 0. Thus at 7, Pi = 0. It follows that at p Penny =P ntg~ and antisymanetriation now gives the result. ‘The result (1.9.10) shows that when the torsion vanishes the [Riemann tensor ean be determined completely from measurements of geodesic deviation, Another tensor that will be required in the sequel is: (199) perinion ‘The Ricei curvature tensor is ¢ (0,2) tensor field obtained by contracting the Riemann tensor Ry = Pe (1.9.12) (1.9.10) Bzereise. Show that the Riemann and Ricci tensors for the Newtonian connection (1.8.4) are Ry, otherwise, (1.9.18) otherwise, (1.9.14) 1.10 The pseudo-riemannian metric ‘The final structure to be imposed on the manifold is a metric. Tt ‘ill be shown that metric structure induces a preferred connection. (1.10.1) peeiNrrion A metric tensor g at a point p in M is « symmetric (0,2) tensor. It assigns ¢ magnitude V{g(X, XJ] to each vector X in 110 The preuderiemannian metric a 7,(M), denoted by d(X), and defines the angle between any two vectors X,Y of non-zero magnitude in T,(M) via (XY) a(x)a07), 05,9) = ase 1 X¥) = 0 then X end ¥ ore orthogonal, The length of a cure with tangent ctr X Between fy and y i L(G) = Se a(X)dt. If (c,) 18 « basis of T,(M), the components of g with Tapes tothe Basis oe gu, aleve) ‘The simplest example is R* with Cartesian coordinates (2,42), where 94 = diag(1, 1,1), hhas all the properties. In the Minkowski space of special relativity with (42,942) coordinates then a, = ingly “1,—1,-1), hhas the right properties If 9(X,¥) = 0 for all ¥ in T,(.M) implies that X = 0, then g is said to be non-degenerate ‘Henceforth this will be implicitly assumed to be the case. It follows that there is a unique (2,0) tensor g with components g" satisfying ca ives the matrix (gt) is the inverse of the matrix (gy). Since Doth are non-degenerate we have s (non-natural) isomorphism, T,(Af) + T3(Af), 1X, given by Xa, m= 9X! (0a) the covector and vector, calling both X. A. similar isomorphism holds for tensors, eg. aT,2 1 Digerentiat geometry where the indices are raised and lowered with g*, 94, respectively. (NB: the leftright ordering is preserved!) In general such as- sociated tensors will be regarded as representations of the same object. (1.10.2) pernurios The signature of « metric is the difference between the number of positive ond number of negative eigenvalues, Ifin e n-dimensional ‘manifold the signature is n then g i positive-defimite or Rie- mannian, while ifthe signature is +(n~2) then g is Lorentzian or paeudo-riemannian. A standard resull from linear alge> bra, “Sylvester's law of inertia", guarantees thet the signature is basis-independent ‘Accepting for the moment that special relativity has local va lidity, we deduce that spacetime has dimension 4 and Lorentaian signature, Using standard results ffom linear algebra we see that ‘at any point p coordinates can be chosen so that the metric takes the form ding(1,-1,-1,-1), there, Tt is now no longer true that 9(X,X) = 0 implies that (0, and just as in special relativity we classify vectors at point as timelike, null or spacelike according as o(X,X) > 0, o(X,X)=0, (XX) <0. (1.10.3) Beereise. Show that in Minkowski spacetime the null vee- tors form a double cone, the light cone separating the timelike from the spacelike vectors. Deduce that in a curved manifold of dimension 4 with lorentrian signature, a Lorentzian manifold, the same structure exists in the tangent space of each point. (1.10.4) Beercise. X,Y ave orthogonal vectors in a Lorentzian manifold. Show that if X is timelike then ¥ is spacelike, whereas if X is null, Y is spacelike or null, but if X is spacelike then ¥ ‘ean be timelike, null or spacelike 1.10 Phe pene remannian metric 8 ‘The existence of a mettic means that there is a preferred con nection with zero torsion. This is constructed explicitly in the following theorem. (1.105) tHe 0REM Ifa manifold possesses a metric g then there is a unigue symmetric connection, the Levi-Civita connection or metric connection V ach that ve= (aoa) Proof: We shall demonstrate existence and uniqueness via an ex plicit construction. Suppose V is metric. Let X,¥,2 be vector fields. Then since g(¥,2) is «function X(¥,2)) = V5 (9(%2)) ¥x9)(% 2) + oT s¥,2) 4 91,952) (YYZ) + (V9 92) Similarly 2,X)+4(Z,VyX)y 2 (X,Y) AV 2X,¥] + (XV 2¥). Adding the first two equations and subtracting the Uhied gives HV) = HZ XY) + ¥(Q(Z.X)) + X(O(¥, 2) Z1XYN) + 900% X1) ~ 9X 1¥, 20) ‘where the symmetry of the connection has been used to set VY -WX=IKY] ete Now if (c,) is » vector basis we may set finding Tae ale: Ve) = G5el = Hela.) + €,(840) ~ eld Vadast Vets Beale (1108)“ 1 Digerati geometry I the basis is coordinate induced the lat three terms vanish and wwe have the famous formula defining Christoffel symbols 19" (ana + Bin ~ Sma) (1.10.4) Whenever If possesses a metric we shall usually use the metric connection without explicitly saying so, However even in a metric ‘manifold not all connections are metric, a8 we shall soon see! If Af is metric the Riemann tensor atsociated with the metric connection has further symmetries, We define FE (1.10.8) THeoRet ‘The Riemann tensor associated with a metric connection has the additional symmetries Reyer =O Ry Proof: Since these are tensor equations it suffices to verify them for a normal coordinate system for which = Raa (1105) R, (1:10) itm = 2Fimte + Bin ‘Equations (1.10.5) then follow immediately ‘There are two more curvature quantities that can be defined in ‘a manifold with a metric: (1.10.2) periwrnion The Ricci scalar is Ray (10.7) The Einstein curvature tensor is (120.8) 1.10. Phe posde-riemannian metric ro (1.10.8) Beercise. Show that the contraction of the Bianchi identities (1.9.11), the contracted Bianchi identities are VG) = VARs ~ fe\(R). (2.10.9) (1.10.9) Beereise. Propose « definition of a metric for Newtonian spacetimes. Is the Newtonian connection defined by (1.84) a mettle connection? An isometry isa symmetry of the metric tensor i.e. is invari= ‘ont under displacements along the integral curves of some sinooth vveetor field K or, in the language of section 1.6, K preserves g Extn = 0. (1.10.10) (1.10.10) Exercise. Show that (1.10.10) implies Vyokky =O. (1.0.1) ‘The property (1.10.11) is Killing’s equation and its solutions are called Killing (co-Jvectors. Isometries are intimately related to conserved quantities. This fs a deep result usually referred to as Nocther's theorem. Various statements of the theorem of greater or less generality can be found in the literature, e.g., (Amol'd, 1978). We now verify this for two particular eases. Consider first frely falling particle whose sgeodesic worldline has tangent vector P. Define E = P*K,, where isa Killing covector. Then VnB = PV,(PK,) P'P'VK, since P is geodesic, POPU AK, by symmetry, © since A is Killing ‘Thus B is conserved along the worldlines of P. As a second ‘example suppose that T"™ represents the (symmetric) ener- ‘gy-momentum tensor of m continuous distribution of matter, isfying V.T"" = 0. If K is a Killing covector define the as- sociated current as J* = T*'K,. ‘Then by an almost identical6 1 Differs yomery calculation one may verify that VJ" = 0, ise the curcent is conserved. 'A very common isometry is when g is invariant in time, ie, there exists a time coordinate t, and a vector T = 0/8t such that £9 = 0. Such spacetimes are called static. Another com- ‘mon example is axisymmetry, If is an azimuthal angle about some axis, and £9 = 0, where A= 0/0), then the spacetime is azisymmetric. 4.10.11 Beercse. Wentify the conserved quantities E and J* for je and for axisymmetric spacetimes. Killing covectors KC satisfy a second order ea VI, = RA Ke (11012) ‘This can be verified by using the identity R*,q = 0 and normal coordinates 1.10.12 Brereise, Consider Minkowski spacetime with Cartesian coordinates 24. Use (1.10.12) to show that the components of any Killing covector are linear functions of the 2". (This result was used in exercise 1.613 to derive the Poincaré invariance group.) 1.11 Newtonian spacetimes From a classical point of view, spacetime, the arena in which ma- terial processes are imagined to take place, is a 4-dimensional smooth real manifold M. Absolute time is smooth map t: M — R with non-vanishing gradient, dé # 0. ¢ is unique up to linear transformations (’ = at +6 with a > 0. Two points of events p,q ate simultaneous if and only if t(p) = f(g). The set of points simultaneous with p forms a space section 5, di- viding M into two regions, the past end future of p. In exch S, a Buclidean metric h with signature (+++) is defined. There is a lot of experimental evidence that the space sections afe flat, eG Euclidean geometry. We shall usually choose a coordinate system 2° such that A = diag(1, 1,1), ie, indices can be raised LAL Heaton spuctinee a Fig: 111.1 The coontnate system ina Galilean spacetime ‘and lowered at will in the space sections, The structure (My!) is @ Galilean manifold Let A be a vector field. In general A(t) #0 and we say that A is future pointing, spacelike or past pointing according as A(t) > 0,= 0,< 0. Spacelike vectors are then the “vectors” of classical physics. "Timelike paths (integral curves of everywhere non-spacelike vectors) represent the history or motion of parti- eles. Galilean coordinates are defined as fellows. First choose ‘8 smooth timelike path P representing the history of the origin and a set of 8 spacelike vectors (c,) at each point of P, and particular point o on P. Let « be an arbitrary event. The surface 5, intersects P in a unique point f, Ja is a spacelike vector and so there exist $ numbers 2* such that fa = 2%e,. We adjust proper time so that t(0) = 0. Let t= tf). Then the coordinates of « are 2*(e) = (ty2") = (t,2',2%,2"). (Poje,) forms « Galilean reference frame. We mus investigate what happens to the co- ordinates when a different frame is chosen. Two frames F, F" have their coordinate systems related by ¢ teonst., ose" +a, (UL) where O is time-dependent orthogonal matrix representing rota tions of (,) anid d"() represents the change in fie ations,8 1 Differential gometry ‘The transformations (I.11.1) form group, the kinematical group "in order to do mechanics we need a time derivative for vectors, "The covariant derivative associated with A will not do, (why?) and the remaining allernative is £, where T'= 0/64. Thus a particle has velocity 2° = £72" = (1,¥), where v is the usual “velocity”. Unfortunately however this definition is frame-dependent (1.11.1) Problem, Let d/dt = Ly = 0/0t in a Galilean frame F ‘and let a(t) bea spacelike vector. IFF" isa second Galilean frame da show that 7 2M ee, (ana) where w is a certain spacelike vector, and interpret this equation. (Bint: first use (1.11.1) to express (dt, da”) in terms of (a dz” ‘Then compute the components of T’ in the frame F. Finally cevaluate the lef hand side of (1.11.2) in this frame.) (1.11.2) Beerise, Let x(t) describe the path of a particle. Show that 3 i frame dependent, but iftvo pasticle paths 1,2 intersect at f point p, then Ax = x, ~3, evaluated at p is frame-independent. Show further that the acceleration X(t) of a particle is frame- Aependent, but for two particle paths touching at p, (i. with the same velocty,) then A3 is frame-independent. ‘The main result in classical mechanics is Newton's law of motion: there exist special frames inertial frames such that the acceleration ofa particle of mass m is related to a force vector thy mk (as) (1.11.3) Beercise. Under what subset of the group of transforma- tions (1.11.1) is equation (1.113) invariant? In order to make (1.11.3) frame-independent we have, following exercise 1-112, to introduce s universal standard acceleration field fa, depending on position and velocity, and write equation (1.11.3) (2,%)) (11.4) 111 Neston apaclines 9 Which is now fram: dependent. Inettial feames are those in ‘a vanishes. ‘How can one measure a? Consider laboratory mechanics. Since we cannot shield off gravity we have, in any inertial frame a)=f+u8, ‘where g is the gravitational acceleration and is the passive ‘gravitational mass of the particle. We rewrite this as mm teed In general the fame one uses a Taboratary isnot an inert one and nsing (111.2) tice we have gaat tes? seat ige detent -exk-wx(oxx). (LIN) By doing enough experiments with varying x, % we ean certainly measure f/m, «, a+ (u/m)g. Now by considering particles made feom diferent materils ne might hope to measure, and hence fe wrt Howe nat mere Pe Bi ‘xperiment shows tht to one part in 10" y/ is independent of the materia, aod so by stable choice of wit, yin = 1 Thar only a+ g can be mesure, and not g separately as no separate physical easing; we can and must incorporate t into gravity. We now reermalate Newton's lew follows There xis free-fall mations which ae charnceroed with respect to «8 special class of galilean frames, the Newtonian frames by the equation of motion R= -V4, (ais) in which the gravitational potential is a frame-dependent smooth function. If a force represented by a frame-independent spacelike vector fis present then (1.11.8) becomes f. (nt) (1.11.4) Exercise, What subset of the group of transformations (1.11.1) eelates Newtonian frames by preserving (1.11.1)? Show further that under a uniform acceleration 1, 4 transforms as Ob ox. (tna) m(e+¥¢)50 1 Differential yomery ina Nowtonian spacetime Fig. LIL? Two goose pal We can rephrase these results in geometrical form. We rewrite the free-fall Iaw (1.11.6) as BET ge o, io otherwise, (tts) The= ter ‘and by convention greek indices range over 1,2,3. This New- tonian connection is symmetric and so the torsion vanishes in Newtonian gravity, A simple calelation, exercise 1.9.10, shows that Rigg =O otherwise. (1-11.10) Te Fallows from theorem 1,106 that there is no metric of which R us the Rieman tensor. A physical interpretation of has bocen given by Ehlers, 1973. Consider two particles at the north pole, Suppose there is a smooth straight cube leading 10 the outh pole, and particle 1 is dropped from rest down it, It just reaches the south pole and then retuens to the north pole, Par- ticle 2 is thrown vertically upwards with sufficient velocity so fs to return to the north pole simultaneously with particle 1 Consider the spacetime diagram for this experiment, figure TAL2. Transporting the initial tangent vector T, 0 path 1 1.11 Neston specter 5 paallelly along path 1 slays produces «vector tangent to path 1, since path 1 is a geodesic. The same holds for path 2 It however T, is transported along path 1 we obtain the vector 7} at the end of the experiment. To se that 7; # Tay note first that « parallelly propagated basis remains a basis. Initially 1 > 0 and finally 75 < 0. Thus if 7; = 7, then there exists a point ¢ on path 1 at which the parallelly transported 7; satisfies 77 = 0. Therefore at q 7), do not form a basis for vectors in the tr subspace, whereas they do initially and Snaly. Thas parallel transport of timelike vectors is path-dependent. However using (1.11.9) and the geometrical interpretation ofthe Riemann tensor fiven in section 1.9 itis easy to see that parallel transport of spacelike vectors is path.independent By moving to a suitably accelerated frame we can always at- range, using (1.11.8), that d= 0 s0 that the connection vanishes for this frame, ‘Thus normal coordinates correspond to freely falling fames. Next consider two points at 2°, 2" + Aa". The relative gravitational acclerstion between them is AS = 9 ,A2% = Ryde, age ‘aot qa Since this isa tensor equation the relative gravitational accelers- tion or tidal force is @ tensor which cannot be transformed away. Indeed (1-11.11) is the equation of geodesic deviation. Finally consider the relation of the gravitational field to the sources producing ‘There isa lot of experimental evidence at a terrestal and solar system level for Poisson's law Vig = AnGp. (11.12) Using (1.11.9) this becomes w= tp, (1.11.13) aan equation to whic we shall retus3 1 Different yore 1.12 Special relativity ‘The postulates of special relativity are that spacetime is a smooth “manifold M with « Lorentzian flat metric m, such that in certain coordinate systems, inertial coordinates (0s -1). ‘The null cones of 7 describe light propagation in vacuum, timelike geodesics describe fre fall motions and [are [ Vradeie = [ima Anton ang tinete path representing pate gies the tripe time ms messed hy undo lack ode with Tere neil connor we dened opt Ueat ta sheataltd, (1123) where d*, [4 are constants and [TnL =. These transformations form a group, the Poincaré group. Simultanelty and causality fae determined by the null cones. Special relativity is well verified by local experiments in high energy physics. Under conditions where gravitational effects ean be ignored there is also extremely good evidence that Maxwell's equations fre valid. Let (2!) be an inertial coordinate system. Let the com- ponents of the electromagnetic field be B°, By the electric charge density be py and the current density be j°. These quantities can becombined into a Maxwell feld tensor Fy = Fug anda (current density J* according to i By= Foy Be=— Hon P?, ‘end Maxwell's equations take the form Faw Phy ard (122) ‘The energy density of the electromagnetic field is W 1(B° + BY), the Poynting vector is S = (1/4z)E x B and the Mazwell stress tensor is tos= (E) ("+ 88 BaBs~ Bas) 118 Special relativity 83 In a source-ree region these quantities satisfy WHVS=0, S4v.t=0, sponding to conservation of energy and momentum, The energy-momentum tensor is defined by LY agit, (ie) Ga Ft Jn. inal coordinate ten 2" =, To" 0 stem 2 =, To 8 and T= 7 The conseraio awe Become — FP iny,) (za) re, (a4) Next we shall consider fluid 7 ° jer fluids in special relativity. ‘The energy ‘density isthe sum of the rest mass density and the thermal energy density. Under normal laboratory conditions the second contribu tion is negligible in comparison with the first, but it cannot be ignored in astophysice an conmology, We peer fluid velocity, repl . th of 1g 4 by U", a unit timelike vector, U'U, = 1 We start by considering the projection tensor : uu, (1125) (1.19.1) Bzercise, Show feom the definition that i) al <0, "o understand the mame let p © Mand consi ime Ie pM and conser a wm 1, +, Let he the space of all vectors proportional to Und MLW Boake apace fac orthogonal. For any cr vi WVU RV, where the first term isin JY, the second in W. Thus h is @ map T,—W. The energy-momentum tensor of « perf fea is taken to be eee T= pU'UE ~ ph (1.12.6)st 1 Differential geometry ‘The conservation equations are presumably TM, = 0. To verify this consider o=T*, : (pU*) Ui + pU* US — p' + Up U* + (UO + U uv) Multiplying by U, gives O= (PU) — PU, + PAU" + PT ae p08 + (P+ 7 p+ (e+ r=0, (127) z ivative along U, and 0 = where p denotes p,t', the convective deriva Te An alterativ frm for dln Y /dt where BV i 9 small comoving volume element. ‘Thus equation (112.7) can be written BBV + (p+ 9)(6V) (By + 6V) = 0, where 5 = pSV is the energy contained in the volume element ‘This is the fist law of thermodynamics. Next consider Wer. =O (e+ 96 ~ hp, wau wt, the Buler equation, provided that p-+p is interpreted as the inertial mass density Balm ctrl or Bid o ny he Kind of fa (nee TAVIV" > 0 forall timelike V* phere, 1 eV 113 The genera theory of reality 38 Fig. 113.1, The Pound:tebiaexpetnent in ease b), The wordings ofthe bake Band op T of the tower nc iyperitan Imo patons siesta a ‘ards om the baie Sty apt they areeocived withing sme separation ‘Srpcovtesponding to red sit 1.13. The general theory of relativity 1 has been concluded that Newtonian theory is inadequate to describe high velocity phenomena which require special relativity while special relativity disregards gravitation. Further while we can define a (relativistic) inertial frame at each point, the rel ative orientation of these frames is not well-determined, It is clear that these two weaknesses are related to etch other. Grav- ‘tation modifies Newtonian theory by preventing exact inertial frames from existing and permits only approximate local inertial frames attached to freely falling particles, and these local inet tial frames are accelerated relative to each other, ‘This suggests that the incorporation of gravity into relativity theory requires ® change of geometry such that the system of timelike straight Hines is distorted into a system of curved lines representing rela- tively accelerated freely falling particles. There are two important experiments which support this viewpoint tis observed that light rays passing close to the sun are de- flected and bent towards the sua. ‘Thus the actual light cones i56 1 Digerentat geometry ‘a gravitational field are not the null cones of the Minkowski met- fie. ‘The Pound-Rebka experiment shows that the preferred frames of special relativity coincide with the local gravitational inertial frames. In this experiment 7-ray photons were emitted ‘vertically upwards by nuclei situated at the base of a tower, and bsorbed by similar nuclei at the top. A change in frequency was ‘observed. Let B,T be the bottom and top of the tower of height I let P be the photon, Fa freely falling body and let U,, U, be the A-velocities of the emitter and absorber respectively, see figure 1.18.1. ‘The analysis of the experiment, assuming special relativ- ity is a sufficiently good approximation, depends on whether the incetal frames of spesil relativity are a) attached to the earth, or by) attached to frely falling particles, or c) some other possibility. In case a) the worldlines of Bare parallel straight lines. In ease 1) if we ignore the slight variation of g with height the worldlines of B,T are hyperbolae with common null asymptotes. In both fees the path of P is null straight line. (1.13.1) Brercise. Show, using special relativity, that the fre- quency shifts (red shifts) in the two cases are a) 0, b) 8X/ = Gif? + smaller terms. Experiments have decided in favour of b) ‘with surprisingly good accuracy. Thus there are no global in- tetial frames in relativity theory, but local gravitational inertial frames ate identical with those in which Maxwell's equations are valid locally. (Notice that there is am assumption underlying ‘exercise 1.15.1, namely that the nuclei represent ideal clocks in tpite of theit acceleration g. This, the clock postulate, seems easonable because the gravitational force on the nuclei is many ‘orders of magnitude smaller than the nuclear forces dominating the emission and absorption.) ‘These considerations suggest the following hypothesis known as the strong principle of equivalence, (1.13.2) axtom In the neighbourhood of each event e in spacetime there exist pre- ferred coordinate systems called inertial at e, For each class 143 The general theory of relativity oT of physical phenomena except gravity, (mechanics, electrodynam- tes, hydrdgmamin, atomic phgece ele) st of loa laws con be selected which, if ezpressed in terms of inertial coordinates and ‘aluated ate fake on some standard form with standerd constonts independent of the spacetime location and the gravitational tidal field. (As 0 corollary, the local inertial coordinates are realizable approximately by frely falling non-rotating smell laboratories.) At his princi is acened it flows that spec clativity ‘local aproxiaton sald only when geval a oe teenegle, further implies hl nw revatio ol ys ical ows fhnowa in the spi lavityeprexination, can be certied overt the general casei prope formulate "he flloing“heunie dracon son i deo Er, 1, Tes clear hat gravitational hor of spacetimes shou he flwing feat nn nt emectines shoal have instein’s field equa- 48) it agrees locally with special relativity, ') it singles out a class of preferred relatively accelerated world: lines representing freefall, ‘Admits the definition of tidal field tensor related to the source density of gravity by a law similar to and consistent with the experimentally verified Poisson's law, 4) it is capable of describing light deflection. ‘The step taken by Einstein in 1915 to satisfy requirement a) was to assume the existence of a pseude-iemannian metre 9 Newtonian theory suggested that b), c) could be satisfied by intro. ducing 8 non-integrable symmetric linear connection representing the gravitational field. (A connection is integrable if parallel transport is path-independent.) However this forced Einstein to introduce a non-minkowskian metric. For the strong principle of equivalence, ( precise form of a) ) requires that there exist coordinates 2" in some neighbourhood of each event e such that MaMa Gm =O Mg Date, (1181) Now (1.13.1) implies 94.,, = 0 at e. This is a tensor equa- tion true in all coordinate systems. Then theorem 1.10.8 impli38 1 Digeentil yeometry that P,,, is the unique Levi-Civita connection associated with dur and if 94 = tu globally then condition b) is violated. In ‘lew of the success of special relativity many physicists are re- luctant (o introduce a non-minkowskian metric and even when confronted with the above argument, reluctantly admit 4. but also carry along the old 9, in their theories. ‘The role of ma, in ic. Re null cone structure cannot represent the propagation of light, and it fol- lows from the Pound-Rebka experiment that 7 does not determine nuclear proper times. Indeed it turns out that in most of these ap- proaches not only is non-measurable, but itis not even uniquely elated to measurable quantities, and so itis best discarded, ‘Thus the first main assumption of general relativity theory is the existence of a siioath manifold Af with a pseudo-riemannian metric 9 of signature (+ ~~~), The null geodesics of g represent light rays, timelike geodesics represent the world lines of freely falling particles, and the arc-length along timelike curves, rep= resenting the (arbitrary) motion of particles, measures the time this Poinearé-invariant approach is problen shown by a standard clock. his assumption satisfies the condi- tions a), b). It incorporates the experimentally verified features ‘of Newtonian theory, (non-integrable connection, curvature ) and of special relativity, (light cones, Minkowskian inner product of ‘vectors) and discards the ilLfounded special structures, (abso- lute time, integrable conection) of the earlier theories. Moreover the strong principle of equivalence connects the theory with local familiar physics In special relativity the energy-momentum tensor Ty satisfies TP, = 0. According to the strong principle of equivalence the ccortesponding conservation law in the presence of a gravitational field is Ph .=o, (1.43.2) for this is the covariant form that reduces to the special relativity form in inertial coordinates. ‘Condition c) still has to be satisied. It was shown in section 1.11 that Poision’s law could be written in the form (1.11.13) Ra = ~An@p. 1.19 The general theory of relatnty 80 Nov for a perfect uid with dovelocity U* the covariant form for the left hand side is R4U'U*. ‘The energy density is T,U°U* and using the equivalence of energy and mass this is a suitable candidate for p. A second quantity which in a weak fed i almost equal to the density is T= 1", = p— 3p. (In the solar system P/p $10") Hence for any value ofthe arbitrary constant OT, +(1-AguTIW'U', corresponds very nearly to the Newtonian density, Thus we set RyU'U" = ~ArG(AT, + (1— Agg POU. The csc way fo iy hs equation forall nee Uo Ray = Ar (NT, + (1 — Aga) (1.33) ‘The argument for (1.13.3 i not eatiely compeling, and this has led toa numberof alimatve bt considera ative but considerably mre compliated theories of gravity. = 7 Fortunately the parameter \ can be determined, We first take the trace of (1.18.3) R= -teG(4 — an)r. (asa) We next take the divergence of (1.18.8) and use the contracted Bianchi identities R*,, = 110" aud the conservation law (1.13.2) to obtain, using (1.18-4) 4rG(l ~ YT, = 2eG(4 - 99)T,, In general 7, # 0 and so \=2, Using the Einstein tensor Gy, = Fa} oay (1183) canbe witten asthe Binatein (1915) ltd equations G, = 867, (ass) Notice that in this derivation condition d) has not been used. Tt can be shown that Binstein’s theory dacs indeed satisfy condition sand inde tis soion ean esd innate sno of the alternative theories. =60 1 Digests geometry ‘A complete description of solutions of (1.18.6) is not yet pos- sible. The remainder of this monograph is concerned with (often Guite intricate) techniques far extracting information about grav- fly. Readers with litle experience of Binstein’s theory will find a ‘succinct description of some exact solutions in chapter 5 of Hawk- ing and Bllis, 1978, although they may find it helpful to work through the frst two sections of chapter 3 here first. Once the first two chapters of this book have been mastered the compre- hensive survey of solutions given in Kramer et al., 1980 should prove helpful Given such con which some simplifying feature obtains, ‘The major problem ronlinearity and so the major simplication occurs when the prob= Jexity it makes sense to look for problems in lem is elose to a known exact solution. Here we present linearized theory close to flat spacetime, i¢., where fields are everywhere weak. We regerd this as a theory on Minkowski spacetime, and fssume a global Cartesian coordinate chart 2! has been selected. We write the metric tensor as d= Mat hier (2.13.6) where ¢ is “small”, and for the rest of this chapter higher or- dee terins will be discarded implicitly. For example we write the verse metric as gt aaah, (aas.7) where indices on “small” quantities are raised and lowered with the metric. Much of the calculation is routine, and so is left as (4.18.3) Bzereise. Show that Paws Define ean) (0088) 3 (bain t Pa Baha tan” (uas9) Neat show thatthe Binsin eld equations (1.185) become , 9 naBaa — Baga” ~ Fis = 16xGe"'Tiy- (1-18.10) 14S Phe gnarl theory of ratty a Using problem 1.6.14 we deduce that under & change of gauge hg hat Leta = hs + May (1.13.1) trary vector. explicitly that the curvature tensors are ‘gauge-invariant. Is this surprising? Under» gage change (LIS, Fg! Hs hy +p whee DE, = &u"- In flat space Lg = p can always be solved and so, loosolaroiein te alana cei (4.18.5) Exercise. Show that in de Donder geuge ‘equations (1.18.10) become wgeuge the field Ry = -168Ge" (1.13.12) By the above remark this equation can always be solved. (4.18.6) Brereise. Consider the external gravitational feld of a static spherically symmetric body of mass eM situated at the eo- gudinte origin == 9 = «= Die it develo 0" = (1,0,0,0) fake Tig = eM8(x)U,U,. Assuming that the field is both weak ‘and a function of R = \/z"+yFF 37 only derive the linearized (1.13.7) Bzercise. The problem in the previous exercise has an exact solution, the Schwarzschild metric appropriate to a mast eM, Lineatize the exact metric and compate it with the solution of (1.18.6). This completes out survey of the foundations of Bias theory of gravitation. =2 SPINORS 2. Introduetion Although the treatment of spinors given here is totaly indepen- dent of quantum mechanics it may help those with a elementary tunderstanding of that subject if we recall some of the properties o ors, ‘The wave function ¥ of an electron has pe Paul sn ope {ho complex components corresponding in some particular refer Siete tz mata opter nthe Hibs apcspaned mas Can bevpreeted as ost of hes component long the seesknat seen str wrdo f= aye) a wat vector thespian dcton & tid = uo, +09, + wey In the Hilbert space of ¥ the components of # can be represented by the Pauli spin matrices (a) aCe) ae so that the ttl spin operator ithe matrix tee(he “AP) ean ‘The theory is required to be invariant under transformations of the form et i va F 21 Intrusion 8 where Q isin $U(2), and | denotes complex conjugate plus trans- pose. To obtain the relativistic generalization of this one simply adds a ‘timelike’ component to ¢ viz. o, = I, where J, is the 22 tunit matrix. Now (2.1.1) becomes ww ul aia (Same SE) ea Henceforth we restrict Q to be a member of $U(2,C), so as to preserve the Hermitian asture of (2.1.3) Now the determinant ofthis mata is precisely the (Loventrian) norm of the 4-vector w'. "This deteriinant is obviously invari ant under the transfonmations (21.2) and so every element Q af SL(2C) defines a Lorente transformation. Since $1(2,C) is sm ply connected, Qis continuous with the entity; so therefore ithe Lorents transformation, and ao the latter must be proper, ie. i preserves both time and space orientation, Further this map from $4(2,C) to the Lorente group L(A) is (21) for —Q defines the same Lorents transformation Unlike Dirac we choose not to form s-component objects fom Instead we regard (2.13) a8 a method for encoding real vectors into 2% 2 Hermitian matrices (Se) a We posit the existence of more fundamental objects, complex 2. vectors which will be called spinors, Tt will turn out that all real spacetime vectors and tensors can be described in terms of the tensor algebra over these 2-vectors, their complex conjugates and their duals. The Dirac 4-spinor formalism can also be formulated in terms of these 2-spinors, as will be described in appendix A. On the other hand spinors need not be describable in terms of conventional tensors, although they will be in orientable causal spacetimes of signature (+ ~~~). ‘Therefore we shall adopt the viewpoint that (2-component) spinors are more basic than tensors or Diae spinorsot 2 Spore 2.2 Spinor algebra Let § bea 2edimensional symplectic vetor space over C, as defined below. (224) perminion isn] = -Ing). (A 2-form|, | on S is non-degenerate if [a6.en) ‘The set of symplectic tronsformations forms the symplectic group of appropriate dimension. wh OY & [As we shall se $p(2), the symplectic group of dimension 2 is isomorphic to SL(2,C), which provides the link to the previous section, A symplectic structure may be unfamiliar and 30 we Shall outline some of the basic features. With the standard Bu- Wdean scalar product no non-zero vector is seléorthogonal; with fs Lorentzian scalar product every null vector is selforthogonal, Dut with a skew-scalar product every vector is selF-orthogor Indeed since $ is 2-dimensional the space of vectors orthogonal to a non-zero vector £ consists precisely of vectors proportional to €, For if 7 were orthogonal to & and not proportional to it the two ‘vectors would constitute a basis for $ and so every vector would be orthogonal to 6, which would contradict the non-degeneracy condition [Let S* be the dual space of $. We build the tensor alge over $, S* in the usual way. Given bases of $ and S* we may define & non-natural isomorphisin between the two spaces: {Wo ‘eetors are regarded as being the ‘samme if their components 22 Spinur algtra 65 respect to the two bases are identical. However the skew-scalar Product defines a natural isomorphism: to the element € of $ we associate [f, ]in S* which is a linear map $C: 9 [én). (A demonstration that itis an isomorphism will be given later.) Unfortunately the geometrical properties of symplectic spaces ‘are not widely known, although a lucid introduction can be found in seetion 41 of Arnol'd, 1978. Therefore it is advan- tageous initially to do calculations with respect to bases. A standard basis may be constructed as follows. Let o be an ar- rary vector in S. Let s be any non-parallel vector. From the paragraph above it follows that [o,«] # 0 and so we may normalize ¢ to set foe] = 1. Clearly «is not unique, for we funy ed an srbiearymaltiple of © toi without upsetting the normalization (2.2.2) verisiriox Ifo is a non-zero vector and sis chosen to set [oy is @ spin basis for 1 then (0,1) Given an airy aia ast we may define component ve tors wi ape tothe bass HE Bin he componente are (4, A=0,1, where S S204 (224) Clearly o*=(1,0), = (0,1) Pure mathematicians often abhor the use of components, be- fause of their ambiguity; does w* + o* represent (a component of) a vector suum, or simply the addition of two components, i, ‘numbers? However they have been unable to provide an alterna, tive notation capable of the same expressive powers, (See e.g, the discussion in Schouten, 1954.) We shall use the ‘suffix notation’ whenever it is the most convenient approach, An obvious exam- ple is the natural symplectic isomorphisin between S and S*; we shall identify isomorphic partners by the the same (greek) Kernel6 2 Spore character, while the position of the sulltx identifies to which space the clement belongs. Thus ¢* (€ S ) is isomorphic to £4 (€ 5°) Wercan get round the mathematicians dificlty by introducing Penrose's ‘abstract index’ notation. Briefly this involves the fol- owing. We build an infinite set of copies of $ labelled $4,,... ter We also need a “sameness” map (implied by the concept of Copies) which ilentifies two vectors in different spaces. We also Ge the same for the dual spaces. Then the notation is as follows: A, vector in 54, 4%, another vector in $4 4+ 94, their sum, also in S", €, the “same” vector as £4, but in 5®, fin the dual, Le., the isomorphic partner induced by the Symplectic structure, in Sy, aagE*s the saume as [ny €] iC. 'A complete rigorous account is given in Penrose and Rindler, 1084. Indeed mathematical physicists have been using this no- tation for tensors without knowing it. Now to satisfy the pure mathematicians we should rewrite equation (2.2.1) as follows. We ficst introduce aspin basis ¢" where the lower hatted index labels the 2 members of the basis, and the upper index indicates that the basis members are in S.A suitable choice is eg, ¢5* = 0% ect =, We then write e While rigorous, this notation becomes increasingly more difficult to read as equations become mote complicated, #0 we shall adopt f policy of partial vacllation, Henceforth spinorial and tensorial indices with a caret always refer toa basis, Indices without a caret fan be given either interpretation unless an explicit statement fo the contrary is given. Although this sounds ambiguous we shall develop the notation in such a way as to prevent actual Peg +E! = eu ambiguities. ‘Clearly the symplectic structure can be identified with an el ement of S* x S* which we write as fy = —Eyay 80 thot 24 Spineralglra or [Gol = cang'n!. ‘The condition that {als cata ivion that (2,¢) is @ spin basis for Tn components with respect ta this basis we have co (4 9) This isa non-singular matrix, and we identify its inverse, up to & conventional factor of ~1, with an element of $ x $ viz, wa( 9d m= (4 a): Although the minus sign may sem at this stage a tile unnatural its value will soon become apparent. (2.2.8) Beereise. Show that ‘The convention on kernel letter of di nel letter of dual elements means that ay an be regarded as an index-lowering-operator. For [esml = canta” = (eau). ‘The bracketed quantity i cleanly the dual of ¢#, and so we have & uk = eye (222) Since ¢,y is non-singular the map €~ [€, ] is an isomorphisn a ingul pe [6 Lian as claimed easter. ‘This equ i concomits * Thi ion has an important concomitant. gular we have t YEE = Cele PE = 18,8, whee 46 ete spinor rancher dls o anit matin, Note cacefully the positon ofthe indices. Evaluating the in the right hand side we obtain eee 663 2 Spore ay (22.3) Just as the metric tensor ean be used to raise and lower indices in conventional tensor analysis so the symplectic structure can be tused om spinors. Again note carefully the position of the indices in (2.2.2),(2.23). Ifthe contraction is across adjacent ‘northwest” and ‘southeast ws occur. (This is why €*? twas chosen to be minus the inverse of € ,y:) (2.2.4) Beereise. Use (2.2.2), (2.2.8) to show that indices no minus For this reason ¢ is usually used as the kernel letter for the Kronecker delta gy and that (2.2.5) Exercise. Show that for any 64, &,4 and only if, AC gg = e4* = 2 Show further that 74€, = 0 4 are proportional [Next we consider a linear transformation applied to a spin bs aot tat, ayes (2.24) Since the saine equations hold with indices lowered it is easy to ace that (6,2) forms a spin basis if and only if ab— fry = 1yi-., the Caserta mata (" 8 nn 10), Sethe ‘that @ linear transformation preserve spin bases is precisely the condition that it be a symplectomorphism we deduce immediately that Sp(2) is isomorphic to SL(2,C) as claimed earlier. (2.2.6) Beeveise. Demonstrate this directly by computing @*. It should be obvious that the ual symmetrization () and an- tisymmetrization [] operations con be applied to spinor suffices, re the dimension of $ is two, 7 yao} = 0 for any multivalent spinor r, for at least two of the bracketed indices must be equal. In particular we have the Jacobi identity eqotor) =0 = Canton + Eackn + Eavtoe (228) 22 Spiner algebra 9 ‘This is usvally used in the form given by the next equation. (2.2.7) uemnsa Let r cp,, be 0 multivalent spinor. Then Fat = Fan. + Hawt c (228) Proof: It is clearly sufficient to consider the case where + has ralenc wo, Then multiplying (228) with the CD indices raised into top gives eaare® ~ tan + ta Tao = Haute” Since r4 Tas) +7 a0; the result follows immediately. This special cate of & more general rent for which an elegant proof has been given by Penrose and Rindler, 1984. (2.2.8) rHeonens Any spinor, pit the sum ofthe totally symmetric spinor rp) ‘and (outer) products of with to nd (outer) products oft with ttllysmmatre spinors of lower Proof: The result is obviously true for spinors of valence 2, and. we proceed by induction on n, the valence of r. Ifthe diference of two spinors a, of the same valence isa sum of terms, each of which is an outer product of «with a spinor of lower valence, we write a 648. Clearly ais an equivalence relation. Now Fase. + Fac. 1 “ pte TR He a) Consider the difference between the first and tween the first and any other term on the right, eg -70 2 Spine ‘This is clearly skew-syrametsic in AB and so by lemma 2.2.7 ean be written as cy times a spinor of valence n — 2. Applying this to each term on the right after the First we find that all n terms on the right are equivalent to the frst so that rue..») ® 7awe..8) ‘Thus ifthe result ie true for valence x1 then itis true for valence 1, which proves the result ‘Thus “only symmetric spinors matter”, or mote precisely every educible representation of $L(2,C) can be realized by symplec- tomorphisms acting on symmetric spinors; see e.g, Naimatk, 1964, for details. Because $ is a vector space over C one might naively assume that the complex conjugate of «spinor in S could be defined as an clement of S. However there are (at least) two reasons why this is nat s0, Let us denote the operation of complex conjugation by bar. Ifa, are in S and cis in C the conjugate of a + <8 would bbe & + af rather than & 403 and so complex conjugation is an anti-isomorphiem and not an isomorphisi. Secondly if com: plex conjugation were an isomorphism then ‘real’ and ‘ima valence one spinors could be distinguished. But iti clear from the discussion of transformations between spin bases that any valence ‘one spinor can be mapped to any other by a symplectomorphistn, s0 that a ‘real spinor would be iomorphie to an imaginary’ one, is clear therefore that complex conjugation must be regarded as ‘an anti-isomorphiem from $ to 2 new vector space 3. To empha- sie the fact that 5 is different from S, not only does the kernel letter of a spinor acquis oe bac, but the suffix also gains a prime with similar results for lowered indices. ‘There is clearly some ational redundancy here, but apart from an exception given below it has become standard. As usual we regard complex con- jugation applied twice as the identity map and identify 5 with S, &¥ = a4. The symplectic structure of $ maps to a symplectic structure for S. It is then apparent that the complex conjugate of S* is the dual space of 5 , which will be denoted 3°. Note fur 25 Spmors and veto n ther that in component terms Zz is always written a ey and not as Zyy. - (This and the raised index version isthe exception mentioned above.) We may now build grand tensor algebra out of $,5°,5, 8°. Because Sand 5° are to be regorded asthe ‘sane’ vector space the order of unprimed indices matters, eg Ment rhe in general. Similarly ey EM oH py But since $ and $ are different veetor spaces we do not need to distinguish between say $5 and Sx 8. ‘Thus all primed indices can be shulfled through unprimed ones and vice versa, et Note however that there is » price to be paid for this convenience. We can no longer regard eg., ry, a8 a 2 x2 matrix if we are allowed to set ry, = ry... The only spinors that we interpret in this way are the Infeld-van der Weerden symbols defined in the next section. ‘Therefore the possiblity of index reshullling must bbe rescinded for that patticular example. 2.8 Spinors and vectors A Hermitian spinor r is one for which # = +. For this to make sense r must have as many primed as unprimed indices, and the relative positions of the primed indices must be the same as the relative positions of the unprinied ones, ‘The simplest example is an element of S x 3, e.8 7°". Let (0,0) (3,2) be spin bases for S and 5. Then there exist scalars &,1)¢,2 such that E080" ee + Gat 4 to" ‘The assertion that + iy ermitian is equivalent to the statement that € and 9 real aul that ¢ and o are complex conjugates.n £ Spore ‘Thus the set of Hermitian spinors +" forms a real vector space of dimension 4. In section 2.5 this will be identified with 7,(M), the tangent space at a point in a 4-manifold MF. Similarly the set of Hermitian spinors rq» forms a real vector space of dimension 4, dual to the one above. In section 2.5 this will become the cotangent space T7(M). We denote the 4 real components of 44° by rt. This ie best understood as ¢ relabelling AA’ M+ a. (The explicit details are described below.) Every spin basis defines « tetrad of vectors (I,n,m,m) via ote, nt ate act me =o", (23:1) aden Me aban Ty = Dae Clearly 1 and n are real while m and 7 are complex con- jugates. We may compute the varfous vector contractions ax evidently IJ is shorthand for 4,1" which expands to 0454040" =0. ‘while all other contractions vanish. (2.8.2) veruition A.sct of 4 vectors which possesses the properties of exercise 2.9.1 is celled a Newman-Penrose null tetrad. This is usually abbreviated fo NP null tetrad or even null tetrad. Another useful Hermitian spinor is Sanne =Cantaps — GAR = MAM, ‘with tensor equivalents 9,,,9° which are obviously symmetric, (2.8.8) Bsercise. Use xy sat ~ 4p to show that Dgny)— Im My, g" =n” = Ame), (2.8.2) 23. Spinors and vectors 8 Heence, or otherwise, show that aot gals ata ete. ‘Thus g has all the properties of a metric tens ‘even more explicit deine a tetrad of vectors ‘To make this 2M tn), p= 24m), Pi(m— ma, = BM). (2288) (2.3.4) Exercise. Show that (e,) form an orthonormal tetrad, sus = ding(,—1, -0). ‘Thus the existence of » spinor structure fixes the signature of spacetime to be (+ - ~~). We can spell out the correspondence 4” ++ r# explicitly by writing Kanon = 9549 440 (23.4) where the a’s are the Infeld-van der Waerden symbols. (Note ‘that the primed index is always last. See the remark at the end of the last section.) A suitable choice is rg, 4 1,23, (235) where the o, are the Pauli matrices of section 21. (It isa tedious exercise to verify that they satisfy (2.3.4).) (2.9.5) Exercise. Raise and lower indices to show that eae AI Mo,, where 1 otherwise. (2.38) (The summation convention does not apply to equation (2:3.6),) If wis any spacetime vector we may write 20 thatum © Sport ‘This is the identifica “Appendix A the Infeld-van der Waerden symbols will not appear again! wn discussed in section 2.1. Apart from We now proceed to the spacetime interpretation of spinors. Clearly as we have already seen, every univalent spinor x* dee fines a real null vector &* vin kY = «784, This i's special case of a more general result, (2.3.8) TueonEw Every non-vanishing veal null vector k* can be written in one or other of the forms Be a bee (23.27) Proof: eis clear that k* defised vig (2.8.7) is real and nll Suppose conversely that &* = wis real and null. The nullity condition is cautamst* all = 0, which says that the 2x 2 matrix u*” has vanishing determinant, s0 that the rows/columnns are linesrly dependent. This means that there exist univalent spinors x, such that Ma og AKA ‘The reality condition is ve Multiplying by x implies, assuming « 7 0, that «,47-= 0. In other words 2 is proportional to x. Renormalizing now gives (2.8.4) where the sign ix Uhat of the (real) proportionality consta Every univelent spinor «defines « null vector A+, However fe defines the same null vector for real @ and so there is some additional ‘phase’ information in x. To extract this information wwe complete x to @ spin basis (x,2) where jis unique up to an additive multiple of x. Now neat + AR), 23 Spiners and vectors 6 is a unit spacelike vector orthogonal to A, and is unique up to an ‘additive multiple of K*. However the 2-plane in the tangent space spanned by k*,2 is unique. Another real unit spacelike vector orthogonal to &* is Bar (etg” — Wa'y ‘Together s,¢ span a spacelike 2-surface (in the tangent space) orthogonal to £. Now suppose we make a phase change to mie sr ete. Clearly & r+ eM and since (1,1) form a spin basis, ise gu = 1, it follows that y+ e-"p. Tt follows directly from the definitions that cos 26s + sin 20, tH sin 208" + cos 200", Penrose has suggested the geometrical interpretation of a univa- lent spinor as a flag. The “lagpole’divection of «* is defined to be parallel to the direction of #*, while the ‘lag? itself ies in the two: plane spanned by &* and s*. Thus the flagpole lies in the plane of the flag. Changing the phase of e by @ leaves the flagpole invariant but rotates the flag plane around the pole by an angle 20 a measured in the spacelike 2-plane orthogonal to the pole spanned by s° and ¢*. Note that a phase change by 7 leaves the flag plane invariant, but reverses the sign of x, This is another ‘manifestation of the fact that the map S£(2,C) —» L(4) is (2,1) So far our considerations have been local; we have considered spinors and tensors defined st a point. Suppose now that itis possible to define spinor fields in some region of spacetime. We ‘must then ask whether this places any restriction on the spece- time? Every non-zero null vector must be either future pointing ‘or past pointing. We see immediately from theorem 2.3.6 that we have a consistent method for distinguishing the two possibil ities; a null vector will be said to be future pointing if the + sign occurs in the decomposition (2.8.7) and past pointing oth- crwise. Thus our spacetime must be time-orientable. Further the sffect of a phase change on a flag can be used to define a space- ‘orientation inthe spacetime: the flag plane rotation corresponding to 0 > Os defined to uchanded, ‘Then the orthonormalPy 2 Spins tetrad (2.5.8) has the standard orientation, ie. ¢y is future point- ing and (e,,¢;,¢)) form a right-handed triad of vectors. (Such a tetrad will be called a Minkowski tetrad. ) Thus a necessary condition forthe existence of spinor fields is that spacetime be oti- entable, In a non-compact spacctime the existence of 4 continuous vector fields constituting a Minkowski tetrad at each point is not xy but also auficient for the existence of spinor fields iiven in Geroch, 1968,) Most physicists would accept fons as being very reasonable and we shall henceforth, assume them without further comment. 24 The Petrov classification ‘The Petrov classification of zero rest mass felds rests on two pices of algebra. It will be recalled that in tensor alternating tensor is defined by tet = Hag Cast = = 2A where the last equation refers to the components with respect to 8 Minkowski orthonormal tetrad such as (2.3.8). Cleatly its spinor ‘equivalent will have four primed and four unprimed indices, will change sign under the appropriate index interchanges and will be constructed out of the ¢ y's. Some experimentation reveals that ‘essentially the only candidate is Gnet = Cancow morn = Meautootaetmn’ ~ tackaotaaton)s where \ is a constant. ‘The second equation (24,1) now fixes JF = =o that only the sign of A needs to be determined. Thi i a straightforward calculation using the last equation (24.1) and (2.3.3) and gives = (2.4.1) Baercise. Carry out the calculations outlined above. ‘As we saw in section 2.2 itis possible to decompose any spinor into its totally symmetric part plus sums of products of ¢4p’s with lower valence symmetric spinors. The following theorem shows how totally symmetric spinors can themselves be factorized. 24 The Petron elasifention n (2.4.2) tHeonene Suppose rap. c is totally symmetric. Then there exist univalent spinors a 4sBgy---s%¢ such that Fane = 8% The B,..-19 are allel the principal spinors of 7. The corresponding veal null vectors ore allel the principal dictions of © sumabant Proof: Let + have valence n and let € (ey). Define TG) an OE This is clearly a homogeneous polynomial of degree n in the (complex) 2,y and so can be factorized completely + (a2 — 2,918.2 — 8,9)... ~ Tavs which proves the result sin dy ate abit we and y are arbitrary, Notice that neither the order nor possible factors in the a8y--.y77 are specifable ‘Asa first example we consider the totally skew Maxwell tensor F.y for the electromagnetic field. ‘This has a spinor equivalent F yaw = Fama: » We define ea = 3, we = aay wher the at equality flows fom the smmetty properties af Fr Using lemma 2.2.7 we have neatly Pe Fansw = Frau + Onto and a second application gives Panam = Fossovviny + anton + ean® 2 Spinone “The first term on the right vatishes because of the symmetry properties of F so that p, autor + €apdam If we define the dual of Fas Py = dea°(Py then it is straightforward to show that Fy tile = Wanton (2.4.2) s0 that dag and Fy, are fully equivalent, (2.4.3) Exercise. Derive equation (24.2) We now use theorem 2.4.2 to decompose yy as Fa0 = % 8p) (243) ‘There are two possibilities here. Ifa and fare proportional then a is called a repeated principal spinor of $, and ¢ is said to be algebraically special or mull, or of type N. The corresponding real null vector a, = c,d, is called 8 repeated principal mull direction (pnd). If @ and are not proportional ¢ is ssid to be algebraically general or of type I. (2.4.4) Beereise. Show that ¢ is null if and only if 4 RyP = [BP — [BP = A Nese APY = BB = 0. Suggest an explicit example of «null electromagnetic field ‘As a second example we consider gravity. We use a result, to be derived later, that the Weyl tensor Cys.g be defined later, can be written as Yweotwatoo + ¥aweotasteo, (244) where Vagco is totally symmetric. The dual of the Weyl tensor is JC 24 The Paros clasifeaion 0 Fig. 24. A Peotose diagram of the Petty dasieaion, TI Indicate increasing spoiacton oft eaten — and Canoe +i et = Wavetamtorns $0 that Cog and Wyycp are fully equivalent, Theorem 2.2 now implies that Yauco = aaBateSay (2.43) ‘There are six eases to be distinguished here. ‘Type I oF {1,1,1,1}. None of the four principal null directions coincide. This is the algebraically general case. ‘Type I or {2,1,1). Two directions coincide. ‘This and all subsequent cases are algebraically special. ‘Type D or {2,2}. ‘There ate two (different) pairs of repeated principal null directions Type IIT or {3,1}. Thee principal null ditections coincide, ‘Type N or {4}. All four principal aull directions coincide. "his is known as the Petrov classification, and the increasing specialization is succintly represented by the arrows in a Penrose diagram, figure 21.1-(I spacetime iy emypty, O is lhat space!) The escription of this elassicition by any other appioach iy vx80 2 Spinors tremely messsy, W has two sealar invariants T= Oypco¥?, T= yg? We Day. 62? and for type HIT that ‘The condition for type IT is thet I T= J =0, Type Dis fixed by Voqna¥ac”*¥" ner) while type N is fixed by Youn" Pooye ‘The tensor equivalents are even more messy. See eg. Pirani, 1905, for the details. (2.4.5) Problem. A skew aymmetzic (possibly complex) valence 2 tensor or bivector Fy is anid to be self-dual if Fr, = iF, ‘and anti-self-dual if Investigate the spinorial ‘equivalents and show that! s) every bivector is the sum of a self Iedual bivector, b) the product of a self-dual fand an anti-selédual bivector vanishes. Can this be generalized to d-index tensors with Weyl symmetries? (2.4.6) Problem, Show that the space of valence 1 spinors at a point can be mapped onto a (double) null cone and hence onto & unit 2-sphere if +k" and —E* are regarded as the same direction. Show that the Weyl spinor Wayop defines a finite set of points on the sphere, and examine the vatious Petrov types. A Friedmann- Robertson Walker spacetime is isotropic at every point. What can be said about the Weyl tensor (and gravitational radiation) in such spacetimes? Determine the Petrov type of a Schwarzschild spacetime from the properties: ‘8) vacuum apart from a static point source, ») axially symmetric about the spatial direction of the source, «) time-symmetric, 2.8 Spinor analysis So far attention has been focussed on the algebra of spinors, i.e, the spinor structure at » point. Given a spinor structure at each 25 Spinor analysis at Point of spacetime one wants them to mesh together smoothly. Consider first the possibility of a Lie derivative of a spinor field. Wauch a derivative existed we could compute eg, Lyeqq. This object would have 2 spinor indices and be skew, and hence would be proportional to ¢,y. It follows that £. tional to 9. Le., X would have to be a conformal Killing vector field. Thus Lie differentiation of spinors isnot possible in general One way to define a spinor covariant derivative would be to note ‘that spinors can be interpreted in terms of tensors, and to ex: tend consistently the tensor covariant derivative. ‘The viewpoint adopted here though is that spinors may well be more fundamen- tal than tensors and s0 an axiomatic approach will be adopted. In the definition given below not all ofthe axioms are entitely in. dependent, but this slightly redundant description is clearer than minimal definition would be propor (2.5.1) permrnion Let 9 and be spinor fields such that @ ond @ have the same vax lence. The spinor covariant derivative is defined aziomatically asa map V,=Vyx.t 0-40 yy. auch that 1V0+8)= 9,040.6, 4) V,(09) = (V,0¥-+0,9), iti) ¥=V,0 implies B= 9,0, iv) Vien = Vee =O, +2) V, commutes with any index substitution no involving X,X?, vi) V.V,f=V,V,f for all scalars f, sti) for any derivation D acting on spinor fil there existe @ spinor €X*" such that Dy = E**'V yy.¥ for all ¥. (Here a derivation isa linear map acting on spinor/tensor fields which obeys the Leibnis rule ii) and annihilates constant scalar fields) All of these conditions are conventional except for vi ‘needed to ensure that €** isa (possibly complex) tangent vector in the spacetime. ‘This identifies the 4-dimes ial vector space of,8 © Spire section 2.3 with the tangent space T,(M) at a point p € Af, and its dual with the cotangent space T7(AM). (Up to now we have constructed merely a 4-dimensional vector space at each point.) [Note that condition vi) is equivalent to the absence of torsion. (2.5.2) tutoREM 1, exiete and is unique Proof: The somewhat complicated proof is given in section 44 of Penrose and Kindler, 1084. V, has two desirable properties. Firstly if V,x* = 0 then the corresponding flagpole and fagplane are paralelly transported. Secondly condition iv) implies that V,gqy = 0 s0 that, acting on tensors, V, is the metric covariant derivative, Tn tensor analysis the most ellicient way to compute the curva ture is usually via « tetead formalism. One introduces a tetrad of vectors ¢,* and a dual basis of covectors et, ie, €,"€4 = 6". Here hatted indices label the vectors, while wabatted vectors la- bbel the components with respect to some arbitrarily chosen basis. ‘The notation is the same as that for Minkowski ttrads introduced earlier. "The Christoffel symbols are replaced by the Ricei rotation coefficients defined by eet = ec Veer = ~8i° Fs (252) For a torsionfree metric connection the Riemann curvature tensor of chapter 1 can be computed from the Ricci identity sa = 2V;.Vegtan for any vector u. Now u can be written as a linear combination of the e's vit. 1 = wje!,. Here the 1's are scalars. (Note the absence of any tensorial index in u;.) Further it is easy to show that Vj,Vq is a derivation with the additional property that it 25 Spiner analyse 83 annihilates all scalar filds, not just the constant ones. Thus Race = V7 alte BWV V gels Hue IW Vae ) If follows that R cia V Va (25.2) Although (2.5.2) may look unfamiliar it leads dsectly to its spinor ssqivaent ‘We introduce a spinor dyad ¢, ‘The defining formula is dual 4, (25.3) suitable choice is ¢ Mtg Oy = 04 ent with (2.2.2), (2.2.8) and exercise 2.2.4. We could now introduce the spinor Ricci rotation coefficients Pasco = tanto tor Veota's (254) although a more useful notation will be introduced in the next section, The spinor equivalent of (2.5.2) is Ravcoamero: = Mnatuer VV gle oe” wy) natu” wVVgera tee: =Mastemd.Vyiy tee (2.8.5) Here c«. denotes the complex conjugate and the fact that Vi,y, is a derivation and equation (2.5.3) have been used. Net westar te eatunte th ight hand wide of (2.8.8). We ist Dev =e? Vico Y poy 2°? (VeoVow ~ VooVeo) eno" + Voy Vo") VoueVin’a 4 Spinors Similarly we define Dew = Voo¥ ny (2.5.8) Beercise, Let D be one of the operators Vj,Vyy apy yep Suppose that the tensors (spinors) Uy,U, have the same valence, while U,V have arbitrary valence. Show that (pu +uDv. D(U, +U,) = DU, + DU,, Show further that Dy Now by successively applying lemma 2.2.7 to the index pairs CD and C'D' in the fallowing equation we obtain 29.4 = VooV nn: ~ VopVoor YoveVownt Hoo Vow + HepVac¥ny® + teoton Vee V Hoo VewVa" YowVercr vee Hoc¥ ew Ver — soctoo Ver ‘Thus 2.4 = ovDen een (258) Now considera typical term on the right hand side of (2.5.8) “8 aco a This is obviously symmetric in CD and if one applies Dep to cant y = 0, i can be seen to be symmetric in AB. We now use lemma 2.2.7 to decompose it into irreducible symmetric spinors rmullipied by es, Because ofthe symmetry properties there wil bbe mo terms consisting of» aymmetiie valence 2 spinor multiplied by an ¢ and we need only apply emma 22.7 to the index pairs javco~*eactoyay (2-5-7) “sD icot m = Yancoy (2.5.8) 25 Spiner anaiais 85 ‘and multiplying (2.5.7) by €®°e"4 gives (1/6)e5 04% (259) Also Devt» = Fanon (2.5.10) is symmetric on AB and on C'D'. Combining all of this into (2.5.5) gives Rancox von = tx otonl¥ancn ~ Meuayctoym)t Cwotco® snorn: tee. et (2.5.4) Bvercise. Show thatthe identity Ry, OR, Thus ® yp. and A have tensor analogues #,,,A, Contracting (2.5.11) gives or equivalently implies that A is real and @,,,y, i Hermitian, Raw =~ Mana + OMagtarps 28,5 +689, (2.5.12) R since @ is clearly trace-frce, ‘Thus A=R/M, &, Bas j Raye) (2.5.13) ‘The remaining terms in (2.5.11) form a Hermitian spinor which we write ar Cuca = ¥avcrtanton + Pewontsatco, (25:14) which is known as the Weyl curvature tensor. By shifting found the terms in (2.5.1) one eventually obtains ©, $4 geFa9) ~ ANG. 990 ~ Psi Gay) + FR 94 (25.15)* 2 Spi Carca has the sane symmettis as Ry and is in addition trace: frees Te represents the patt of Ry. not determined by local matter. It is often referred to ar gravitational radiation although it may be time-independent. (2.5.5) Exercise. Show thet for an arbitrary spinor &, Danke = Fanen’? ~ 2G Emer — rane) = Fascn’?> Oyf? =H3Ay Onake (These are the spinor forms for the Ricet identities.) where (2.5.6) Exercise. The Bianchi Identities are V'R ‘ace R ‘Obtain the spinor equivalents att = Hed" Rates won xntero® ancot ER ssa = Cantco¥ a eumtooPanow ~ tanto ¥eoxwt 2N(rokerp€oratme ~ Eautcv outa)» Vamos 80 guAe VP oP anco vel a ‘One may well ask why sich clomsy formulae could be use- fal, Consider first a spin 1 zero rest mass fleld in vacuum, e- clectromagnetism. Maxwell's equations are equivalent to VPs t iP) or from (2.4.2) 7 (2.5.16) iy. The WM ban Similerly consider a spin 2 zero rest mass field, Le., grav Bianchi identities are Vi Regu 25 Spinor analyse 7 It can be shown that in vacuum this implies VPP ace (25.7) (2.5.7) Problem. Show from exercise 2.5.6 that in a vacuum spacetime Vee aco By differentiating a second time, (use V,,4 ;) show that D8 vc = OV eran¥on where 0 = V4.0 is the usual wave operator. What Petrov classes admit exact plane gravitational waves D¥ascp = 08 ‘What can be said about lincarived gravitational waves on a given background? In fact zero rest mass fields of spin + = tn where m is an in teger or zero have irreducible representations in terms of totally symmetric spinors g4q,.c of valence n satisfying Ve bane 0, (25.18) Htm > 8 there is © consistency condition, the Buchdahl constraint ‘that must be satisfied, (2.58) Problem. Let yop, be 8 valence 2s totally symmetric spinor, representing a tero est mass spin 2s feld. Show that for #23/2 the Buchdahl constraint, Buchdah, 1958 WH ober. anc =O, ‘ust hold. Investigate what heppens for = 2. (2.5.9) Problem, Let o* bea constant spinor in Minkowski space time, Let x be a (complex) scalar field satisfying Ox = 0. Show that Fan = VV Vo Ay ee-Beu)y88 2 Spiners is totally symmetric and WM bso. = ‘Thus every x satisfying Ox = 0 generates a spin s zero rest ‘mass fleld, In fact the converse is true in any simply con- nected region. See Penrose, 1965, especially p.l05. (x is a Heerte-Bromwich-Whittaker_Debye-Peneose potential.) Besides the examples given above there are some theories w! fon. Let , be an is obvious fare actually simpler in a spinor formul arbitrary spinor field. The following iden Vee da = AV woo + 2% aV cei) pata) + eaabeV ends (25.19) where the indices between the verlical lines are to be excluded from the symmetry operations, and lemma 2.2.7 has been used in the last line. Next define the vectors M, AVeed*s Riv = bana, (25.20) and the (anti-self'dual null) bivector Py = 8abatoe (25.21) Ik is atesightforward lo establish the identities (2522) (25.23) (2.5.10) Beerve. A Dirac-Weyl spinor satisies Vand (25) Show that R, = 0 and hence thatthe tensor equivalent of (2.5.24) PVRS R, VS 0, ‘considerably more complicated equation. 2.6. The Neamen- Penrose formalin 9 (2.5.11) Problem. Dirac's equation couples two spinor fields Vata mires Vad nw Hb (25.25) where jis a real constant proportional to the mass of the fed. Show that $4¥ on8 = —Hb axes xe then equations (2.5.25-4) imply VES + M, =~, VaG +N, -20C,, where Gy N, are the obvious analogues of Fy, M,. Hence establish the tensorial equations governing Dirac’s theory FuV PEt RYE = ~IBE Ce, GaV GE +O VGS = ~ MGC CG, = PCy, 2.0 ‘The Newman-Penrose formalism ‘As was mentioned in section 2.5 many explicit calculations are most efficiently performed using « tetrad of vectors. While o thonormal tetrads had been used for some time, the idea of adapting tetrads of null vectors to spinors seems to have been developed frst by Newman and Penrose, 1962. The idea is most easily understood as the translation of a spin basis (0,«) to the eal spacetime, “As was shown in section 23, (0,1) induce four nll vectors, a NP null tetrad via Meo, ntact, : 1S eden = balan ay. OY) Obviously land n are real while m and are complex conjugates. ‘The directional derivatives along these directions are denoted by the conventional symbols D=ry,, a-w90 2 Spiners Clearly V, is a linear combination of these operators. Indeed exercise 2.3.3 v, ing + nd = mat — mE, D410 —F5— mb. (263) ‘The ides now is to replace V, by (2.6.3) and then to convert all remaining tensor equations to sels of sealar ones by contraction with the NP null tetrad, Although this may lead to a vast set ‘of equations, they usually posssss discrete symmetries, and be- ‘cause only scalars are iavolved they are usually easier to handle in specific calculations. As presented above the NP formalism is independent of spinors, and indeed there are a number of practi tioners who claim to understand nothing of spinors! However the apparently fortuitous notation was not chosen at random, but has lors are usually the most ficient tools for earrying out calculations. We start by considering the connection. In a tetrad formalism this is described by the Ricci rotation coefficients (2.5.1). Each sealar coefficient is of the form oral origin, and with practice 5 (basis vector). (directional derivative)(basis veetor)), (2.0.4) and the obvious spinor analogue is given by (2.5.4). Each is of the form oVByy where @ and 9 are (0,4) and V is one of D,A,6,8. Note that since oatt = 1, o8Vi, = «400, and so there are three independent choices for a,f. Since there are four choices for V there will be 12 complex rotation coefficients cortesponding to the 24 real rotation coefficients, These are given in the box in appendix B. We could ‘also derive them vectorinlly by noting that since II, = 0 it follows that "V1, = 0,0 that there are only 6 choices of vectors in (2.6.4) ‘and there are again 24 real coofcients. To see the relation between, 2.6 The Newman-Penrote formalin a the two representations cousider e.g, Deby) 04041" Dog. + 0°84" Dog =ADox larly complicated We now derive the NP description of electromagnetism, and for simplicity we assume there are no soutces. As we have seen the ‘Maxwell feld tensor can be described by a symmetric 2-spinor @ap- This generates 3 complex scalars via = F000", b= hyo”, b= dante. (26.5) (@,, contains n e's and 2m o's.) Recalling that 49 = oA? 40, and that $49 = €,°¢n Gop one immediately obtains Pap = Frau ~2rocatnt Ootaty» (2.6.0) It is not difficult to relate the @, to the components of B, Bas defined by the orthonormal tetrad (2.8.3) induced by the NP null tetrad. For example Fy -iBy= Fey biRy = (FER Ys HP LEP) (4 0)C— using the symmetries of F, = Wapeaye'to! sing (2.4.2), ~ Bb gat'ol™ 2. (2.6.1) Brercise. Show that bn BiB, ‘The Maxwell equations are +6, ByiBy = ~24, 0 bay “Va uGew = AoW peen: (26-7)2 2 Spore ‘There are four complex equations here given by A',B = 0,1 cor responding to the eight real Maxwell equations. For example we sony multiply (2.5.7) by 046! obtaining (Ao* ole — 43" 0 Wandoc (07 ED ~ of B)Idstatc — ds2nte ~ $yt0%e + $y0n0cl = 0s = H9t© Dig + Dd, + $41 Doc ~ #0" Dip + $30"Doy $0°Bic+ $40°Btg ~ 4,0%boy — $40°B00 =, Di, ~ Bb = (x ~ 2a)by + 268 — by ‘A complete set of source-tice Maxwell equations is given in appendix B. ‘We may decompose the traceece Rice! tensor a Foe = Banu ooo 3", anwna8AT™, (26.8) Pay = any POO", a, ore, oy = Baan Ore Since #,, =, there are 9 independent real quantities, as might have been expected. “The ten independent components of the Weyl tensor may be combined into 5 complex scalars W, defined by Vawcpo"o%o%o”, Waacnoo%0%e?, Ypene"aiee, (289) Vaveot eee”, 2.6. The Newman-Penrose formation 93 ‘The equations defining the curvature tensor components in terms of derivatives and products of Ricci rotation coefficients, the so-called field equations together with the Bianchi identities are given in appendix B. Although they look fearfully complicated. things are not as bad as they seem, Since we are dealing only with scalars then for any NP quantity 6 W.Va (2.0.10) Before we translate this into NP notation we note the followi ‘equations which follow from (2.6.3) Dig = 204 ~ chy ey = 004 ten cane 0. Bty = Hog — Beay ¢ » For practical reference these equations are repeated in appendix B, together with the equivalent equations for the derivatives of the NP basis vectors [Now reconsider (2.0.10) multiplied by say [*m* m'DY,§— 69,9 = 0, DBs ~(Dm'Vyb—SD9+(8N)V,8=0. (28.12) Now Dm! = Dio!) co — m2 + otal — zal = Ht — nn + (eam. Similarly BI = (4+) ~ pm" — ome Thus (2.6.12) becomes (FG ~6)D$—nAb+(pHe-2)64-+0B9. (2.6.13) (Ds-sD)6 ‘The four commutator eqsations like this are given in appendix B.oo 2 Spiners Although the system of NP equations may seem, at first sight, to be horrendous, there are two diserete symmetries which can be used to simplify calculations, It is eagy to see that the following ‘two oper ions preserve spin bases oti, Av iod, io er ott, het, OM no, (2.6.2) Protlem. Investigate what happens to the NP null tetrad, the connection coefficients and the Maxwell scalars 4. Hence show thet the 4 Maxwell equations are just ‘, * and. trans: forms of one equation. (Similar simplifications oceur for the “eld” ‘equations and the Bianchi identities.) [A common practical problem is the computation of the NP scalars. Suppose a coordinate chart (2*) and NP tetrad have been given, 40 that the components, eg., (2), are known funct ‘Clearly itis an elementary exercise in algebra to compute both the covector components, ¢g. (2) and the line element g,,(2) using exercises 2.31, 2.3.3. However since, eg., =m" DI, = m'PV yl, involves a covariant derivative it looks as though itis necessary to compute the Christoffel symbols from 9,y. Fortunately this is not the case ‘Suppose first that all of the NP scalars are required, One uses ‘the commutator equations, e.., (2.6.18) listed in appendix B. For each fixed a one may regard 2" as a scalar Beld on the manifold fand apply the commutator to it. To make this clear we attach caret to the index, Then 2* = m* is also a scalar field on the manifold. Thus (2.6.18) reads Dm! — 61° = (#4 Blt nt + (pee) m* bam, (2.6.14) Since we are dealing only with scalar functions the remaining derivatives in (2.6.14) are partial ones, i.e, the connection is not needed. Applying the commutators to each of the fields 2* one obtains 24 real linear algebraic equations for the 12 complex eon- nection coefficients, and these can be solved to obtain the NP scalars explicitly. The “eld” equations can now be used to ob- tain the curvature components, using only partial differentiation ‘and simple algeb £6 The Newmen-Penrose frmaliom 9% only small subset of connection coefficients is needed there is an alternative procedure which has been implicit in the literature, oC has ony eecetly eon devrbed egy by Cae Note that pee eee eee “DI, = m* Di, ~ = 2 mV my, (2.0.15) where the second equality flows since ia al vet Ay allows snc is nll eto. Easation (2.6.15) involves & “curl” of a vector and so only ies of vector and vo only patil derraies are needed. if (2.6.3) Brereise. Obtain the expressions = Fe Vil — HEV my + fm : (rs 0g —at Vimy + mV, amy) — mT Vm The remaining expressions can be found using the / and * transforms. Bee _In actual ealclations ws it ten made of Larents transforma Sion to simplify calculations. We need therclore to deteribe hv such transformations ales the NP coeflient, Let (02) be pn basin Then (2) = (30,4) is aloo a spin buss for tay comer 2A, Setting A= aexp(i0) we ee that the NP (teed leana , (2.8.16) Such transformations with O ae called boosts and i 8 #0 a= 1 we have a spin. The general case is a combinatior . two. The behs ase 1 std abl of ba anstema “t xmas is gen (8) = (ous), where ie amity oe ne NP ten ttn acing (26.17) (2.6.17) i x of the NP scalars under such rotations is given explicitly in appendix B.96 2 Spinsrs By interchanging o and «the ‘transform, one may also construct rnall rotations about n. ‘There transformations depend on two real and two complex parameters, in total six real ones, and 20 constitute a representation of the Lorentz group. 2.7 Focussing and shearing Perhaps the simplest application of the NP formalism is to the propagation of light rays We therefore consider a beam of light rays. ‘The beam can be represented by a congruence of geodesics with tangent vector [* and connecting vector 2*. The concept of “connecting vector” was fixed by definition 1.9.6, and it may prove helpful to review the discussion of geodesic deviation given in section 1.9. By definition their commutator vanishes a= (27a) At some point pon each geodesic we choose a spinor o such that ‘OAS is tangent to 7. We then complete o to a spin basis (0,1) at pp. We next propagate 0 and «along 7 by the conditions Do! = De =. (2.72) ‘This means that o and + ate paallely propagated along + and so remain a spin basis at each point of 7. Further identifying !* swith 42" itis lear that Dl* = Do*2*’ = 0, so that 7 is affinely parametrized. ‘Suppose that at p the connecting vector is orthogonal to y it 0. Then Vet, = 3V(M) = 0 where the third equality fellows from (2.7). Thus 2* is always orthogonal to 7. We now construct the NP tetrad ([,x,1,7) induced by the spin basis. Since 2* is real and orthogonal to [* there exist a real w and complex z such that PVE, (273) 27 Roewing ond shearing 7 ‘We next compute Det =z" UDI +361" + 231", oo" Du bof" Ds +40" Ds = toto" + 20460" + soto" + 20"FoA Multiplying by 0,24. gives ~Ds = ~f0,b0" ~ 20,504, Ds=~p:— et. (214) The interpretation of «i follows. Consider th ; ; ollows. Consider the projection of 2 onto the spacelike 2-plane spanned by m‘,7, or equivalently esse" from (23.8). Waite the projection VA(2e) — yey) = 2(m 4 8) + y(m — mY where = ztiy, which is consistent with (2.78). Thus = describes the projection in an Argand 2-plane spanned by m* We now analyse equation (2.7.4). Suppose first that ¢ = O while pis real, ie, De = ~p2, oF De=~po, Dy =~py. “hiss an isotropic magnification at aa fon at arate p, Next suppose that 7 =0 while p = iw, so that Dz = iwz, or oo De=-wy, Dy =ue, which corresponds to rotation with ang rotton with angle velocity w. ‘Next consider the case where p = 0 and o is real Then De=-0z, Dy=ey,principal axes along the x and y axes (2.11) Beis, Show that making © by e® slates the (Gicplsear ey Bn ce (274) nar equation, the general ee» ape peat Si th ee The congrrnen or mere preset veseton ofthe conning eto ente a otogeal pace BGI cpanel sosed and sened. Tha csi be ea ted eperineulyo lls Supe s oague screen with 8 tottus pled etapa oo staonny bea Sut psand eseund sen ws ured fo cope the ige Stet by eving te scm cen lie othe ew the eee ortbe mage meld change neue hv ect Thi wr ere! atria serine comolgy One ewan Siem lace inngeothoat fst Guetnging thee "Since Do* = 0 it follows immediately that « = ¢ = 0. Let se neha rte! toa fly of pert, he Seve font ey there oon slr ele ow ch hal yu (2.75) 4 ‘Then itis eany to show thet 4uly= (278) in fact the converse holds: (2.7.6) implies (2.7.8) as a consequence eevee nen Thera sector Sela, eeting 28) is said to be hypersurface-orthogonal. Clearly hypersurface: ‘orthogonality implies mira lily = which implies that p is real, In this ease the NP field equations (a), (b) of appendix B reduce to Dp=P 400+ 8m, Do=2pothy (272) 2.7 Racssing and shearing 99 2B jy — 2, (278) ‘Now suppose that we have an intially parallel circular beam, ie, p= = 0, in a region in which &,, = ¥, = 0. The beam enters region where Buy #0,y = 0. It is clear that ¢ remai zero,but, what happens to p? Now for all reasonable matter the ‘nergy density is postive, which we express by the weak energy condition: 20 forall timelike *, (279) where 7°" is the energy-momentum tensor, (See evg., Hawking and Ellis, 1973, for a detailed discussion of energy conditions By continuity this must hold for null vectors and so T.,I*D > 0, Now the Einstein field equations (8r)(Ry ~ rs) <0, and so #,, > 0 for all physically reasonable matter. &,, represents the matter crossing the geodesics transver- sally. Now Dp = p*+9,. > Oand so provided the strict inequality holds, p becomes positive. It is easy to see that p — co in a finite parameter time. For suppose w is an affine parameter and p at v=0. Consider first 4 tee ‘The soltion ofthis equation via. 1/(o5~p) certainly becomes ine finite when v= 1/p,, But forthe exact equation Dp must be even larger and so infinite growth cannot be avoided. When p —» so we have a caustic which may alo be « curvature singularity, (Geometical singularities are discussed in some deta in Haws ing and Elis, 1973) In optics this behaviour with =O sealed anastigmatic focussing. Now suppose Yy # 0; which ples that 7 #0, Wy represents the gravitational redation crossing the ‘ays transversally. If sheat is prevent then caustic sil aeeure, even if %5 = 0, Dp = p00 and the tern generates fucusting. This v1 estigmatic Tens. Dp=p'100 2 Spinors (2.7.2) Bvereise. Let D = dfidv and solve (2.78) for ‘A6(%),® yy = BBlv), where A,B are real constants, B > 0, and p= o =O intially. (‘The general ease is a “superposition” of such sources.) We still have the problem of solving the NP equations. Ale ‘though this i formidable task, some sub-prablems are relatively ‘easy. Consider the “D-equations” which are the restriction of the INP equations to a null hypersurface with [* tangent to the gen- crators, We demonstrate that this reduces the NP equations to a linear system of ordinary differential equations. For simplicity we shall assume a vacuum and assume the tetrad has been chosen to set n= ¢= 7 =0, We also assume that Ws is given as free data, Define rac tf) Ga) eG ‘then equation), b) of appendix Biply Dxaxter We next define ¥ via X = —(DY)¥~!. Like X and Fy ¥ is Hermitian, ‘This equation can also be written as DY = ~XY. Now ¥ has to satisfy > -FY. ‘Once this second order linear equation has been solved for ¥ we ‘may make a non-linear transformation to recover X. Now equa- tions e), 4), e), Ba) of appendix B form a coupled linear system for the unknowns r,c1,8,¥,- Once these have been solved the ‘equations f), g), h), Be) can be treated as a linear system for ‘qs%of Wa. Finally i), Be) form a linear system for ,,, and then Bg) ie a Tinear equation for W,. 2.8 Goldberg Sachs et al. Although the title sounds like the name of a stockbrokers? firm, it actually refers to a group of theorems. It is clear from the diseus- sion in section 2.7 that the anslysis of the field equations would 28 Goldberg Sasha 101 be considerably simplified if « vanished fu vanished for a congruence of null ‘geodesics. These theorems establish necessary and sulficient con- ditions for this to obtain. While these theorems appear to have no immediately uilicable consequences they serve a useful pedagog- ical service an intoducton to manipalating the NP equations lowever as a precursor we consider the problem of integrability conditions. Suppose the scala field ¢(,2) satisfies the equations % jan, a Be = 9lhah (281) where f,g are known functions. Can a solution f c ions. Can a solution for ¢ be found? Locally this is a well-understood problem of the type 4 = B, and has a local solution if and only if @ x B= 6, of Of _ 09 ae (28.2) [Next consider the generalization BY d=o (2.8) hess AB" ate loc reese vaste fds which eure forming, i, there exist scalar fields a and Bs fece fields a and 8 such (4,8) = 044 68. (2.84) ‘The reason for this name should be obvious from the geometrical interpretation of the commutator given in section 1.6. (2.8.1) THeoRExe A nezenary end sii coiton for slain of theo (2.8.3) to exist is sfnlivnnl V9 BV,S (285)102 2 Spinone Proofs As in chapter 1 we adopt the nota suppose first that (2.8.8) has # solution ¢. Then (2.84) implies A(B8) ~ BAe) = aA$-+ BBY Setting Ag = f, Bd = g we obtain (2.8.5) immediately. Suppose conversely that (2.8.5) holds. We adopt the special coordinate system of lemma 1.8.2 to set A= 0/By'. ‘Then we may solve ABH f vi doy)= f" Has )ds where = yess. We denote this solution by ¢ = AY. Let Bd g~h, Wemay choose j= jy") to set h= 0 on an inital surface y! = const. Now Ah= ABA ~ Ag BAAS 4(A, BIA“ — Ag Bi+(aA+AB)A*~Ag by (284), 1Bb— Ba by (285), Since h ially we see that h = 0 and so BY = 9, ie ¢ solves (2.83) We start with two results concerning solutions of Maxwell's equations and shearfree congruences. (2.8.2) THe MARIOT-noBINSON THEOREM Suppose thet Fy, is a null electromagnetic field, ‘Then the re- peated principal null direction generates a geodesic shearfree null congruence. Proof: Let the repeated principal spinor be 0, and set ¢4p = {40 ,oy- Complete o to a spin basis (0,2). Then from (2.6.5) it follows that 4, 0, $2 = 4. Now the Maxwell equations 28 Golderg Sache eto 103 (Ma, Me) of appendix B imply that & (2.6.11) wehave Do, = ¢o,. I follows that 0. Further from DI = Diora" = (et pote ear, so that the congruence with tangent vector I" is both shearfree and geodesic, (although it need not be affinely parametrized.) ‘There isa converse result. (2.8.3) noninsow's tHEoREM Suppose spacetime contains « geodesic shearfree null congruence ‘Then there exists @ null electromagnetic field, i.e, «solution of the Mazvell equations with repeated principal null direction tangent to the congruence. Proof: Just as in the construction described in section 2.7 we may choose a spin basis (0,1) with 045" tangent to the congruence, and o*,14 parallelly propagated slong the congruence, i., Do* = DiA = 0, Then equations (26.11) imply that x ‘The Maxwell equations for a null feld 44 = $o40y reduce to (Mb, Md) vie. Dé= ob, 59 =(r- 2896. (288) In general the equations (2.8.6) have no solution. However [D,§] = —(& + B)D + 76, and so 1° and m* ate surface-forming. If we introduce (28.6) become Ing as a new dependent variable then equations Do=, b= 7-28. ‘The integrability condition becomes Dir =26\ =p = (a+ Bp + ple ~ 8). (8) Using the field equations ¢), e), k) of appendix B this condition is easily seen to be satisfied identically. ‘Thus the integrability condition for the Maxwell equations (2.8.6) is always satisfied at « point if the Maxwell equations hold at the point and the feld104 2 Spinone ions (cyek) hold there, It is therefore possible to integrate the equations obtaining a solution satisfying the requirements of the theorem, From exercise 2.4.4 and theorem 2.8.2 it is clear that plane electromagnetic waves can only propagate in spacetimes which admit a geodesic shearfree null congruence. We must therefore ask what restriction the latter condition imposes on a spacetime. "The necessary and sufficient conditions are given by the following theorem. (2.8.4) tue couopEne-sacus THEOREM In a vacuum spacetime any two of the following conditions imply the third: A) The Weyl spinor ¥,yco & algebraically special with n-fold repeated principal spinor 0, (n= 2,3,4), B) either spacetime is flat or o generates « geodesic shearfree congruence, ©) V4*W jgor contracted with (5—n) 0's vanishes. Proof: Assume A with n= 2, Banca = %a2u%eBny- Clearly 040% avon and s0 08 V OH aucn + SaucneQoVP" =O. (2.8.8) Now suppose that in addition B holds, If spacetime is lat then condition C obviously holds. Ifinstead o is geodesic and shearfree then 060g = Deg = 0 and 0¢0,72"'0° = 0. It follows that the last term in (2.8.8) vanishes and so C is true. Suppose instead that C holds. Then pg nn gs Panev0*e 2.9. Plane and plane fronted waves 105 and 80 06007 a6 which proves B. The proof that B and C imply A is slightly messy. See e.g., section 7.3 of Penrose and Rindler, 1986. (2.8.5) Bzercise. Show that A/and B imply C,and A and C imply B torn = 3,4. 2.9 Plane and plane-fronted waves In elementary courses on electromagnetism “plane waves” are defined to be solutions of Maxwell's equations in Minkowski space- time with e+" dependence, where h, is a constant null covector. Because ofthe linearity the Fourier decomposition is not essential, and so we consider a (sight) generalization. (2.9.1) perwvition A plane-fronted electromagnetic wave in Minkowski spacetime is a solution of Meswell’s equations possessing « I-parameter family of null plane wavefronts, u = const. where e(u)2",—h, null end nonzero (2.9) In the special case where h, is constant we have @ plane wave. Note that on a surface u = const. h, is constant and so the wave fronts are still plane. A sta nary torch should emit plane 2, but itis waved around the waves should be plane-fronted, clear that definition 2.9.1 is highly non-covasiant. If we want to study wave propagation in a general spacetime then we would expect ‘plane-fronted? waves to be the simplest, and so we want a covariant definition of ‘plane-fronted’. Such a definition is suggested by the following lenin,106 2 Spnare (2.9.2) urna Consider « plane fronted wave in Minkowski spacetime, and define 1, = Vyu, where w is defined asin definition 2.9.1. Then, is mull, seodeste, and free of rotation, divergence and shear. Proof: It follows from (2.9.1) that ~h(u)=2Ki(u)VQu, (2.02) = (1 Rw) (293) Since k, is null, 4 is /,. Since [, is @ gradient Vjsly = 0, and so Dl, Maly Mah = VOGAE) and I, is geodesic. Now differentiating (2.0.8) gives E(u), = AVL HLTA. (2.9.4) Let us complete [* to a NP tetrad. ‘Then transvecting (2.9.4) with m! shows that MI, + 1,6 = 0 and Am*él, = 0 or c = 0. Similarly p = m*5l, = 0. Note further that the special case k, = const, implies X= 1, and so from (2.9.4) we may conclude that Vuh =0. ‘Thue we are led to the following coordinate-free definition (2.9.8) permrion Consider « wave with wavefronts w= const. Let |, = Vu. Ifl is rll, geodesic, and free of rotation, shear and divergence then the wave is plane-fronted. If in addition Vl, = 0 the wave is said (0 be plane parallel or pp. (In this case, as we shall see in theorem 2.9.4 the nullity condition ean be dropped.) 2.9. Plane and planefroted woses 107 In the electromagnetic case Robinson's theorem guarantees the ‘existence of plane-fronted and pp waves, and they are all of type 1N, We turn therefore to the gravitational case. It follows from the Goldberg-Sachs theorem that plane-fronted waves can only occur in an algebraically special spacetime. Suppose that we have a plane-fronted wave. We may without loss of generality assume { is affinely parametrized with = 0. ‘This eate is still too general for a complete analysis and so we investigate pp waves, We now state and prove an exact result on pp gravitational radiation, Although the proof is long it is worthwhile working trough it. This isa classic example of the mixed use of tensorial, spinorial and NP techniques, and the complete discussion is not otherwise available (2.0.4) tweoREss Suppose « vacuum spacetime contains @ non-zero covariantly constant or parallel vector field | v=o. (295) Then: 1) Lis null, and the spacetime is algebraically special of type N swith repeated principal null direction t, 2) coordinates (u,v,:2,y) can be found 40 that the line element is ae 2H(2")du’ + 2dudu~dz*— dy, (2.9.6) 9) H(a*) = R(J(u,2)) where z= 2-4 iy and fis analytic in z but otherwise arbitrary. Proof: We start with part 1, Equation (2.9.5) implies Vi.Valoc: = 0, and the vacuum condition then implies Vanco!” (2.9.7) Now if PY # 0 there must exist a 6, such that 0° = (?2'B,, # 0. Then equation (2.9.7) implies V suep9? = 0. (298)108 2 Spars "This means thet the Weyl tensor is of type N with repeated principal spinor 3 i. Fancy = ¥ B49? (This can be seen by completing @ to a spin basis and considering ‘the Weyl equivalent of equation (2.66).) Now from (2.9.7) i fol lows that 3p[29 = 0 from which we may deduce that there exists 2 7 ich that [ogy = op7p The reality of ! implies that [py toust be Hermitian, i, ly = Rapyy where is eal. Thus is aul. By incorporating a factor VB into 3 we obtain lop =98p (2.99) ‘To prove part 2 we first note that equations (2.0.5, 2.0.9) can be rewritten as VaalBoop (2.9.10). which implies 3° ,yy =0 oF Paade haggis (ean {or some a,qy, and equation (2.0.10) implies that a, is Hermi- tian, Further, in vacuum, equation (2.9.8) and the Rice identity (exercise (2.5.5)) imply = aye02” = Dawe = VaraVoy" 9 or, using (2.9.11) Vea" gia + Fates" On examination the last term on the right hand side vanishes and the vanishing of the first implies that V;,a1y = 0, s0 that locally Via for some scalar field a. If we set 0, = e-"3y then (2:90.11) implies Vo. 20. (2.912) Now (2.95) implies Vj,ly = 0 and so locally {, = Vu for some sealar field w. We may choose a coordinate system (z*) with 2° = uso that 8. (2.9.13) 2.9. Plane ond plane rowed wocer 109 Next complete o to a spin basis (0,4) and define (Yampa (2.9.14) This implies HL, (2.9.15) or, using (29.13) 0m, =0, where greek indices range over 1 to 3. ‘Thus there exists a scalar field = = 2(27) such that locally m= 92+ SMe (20.16) (Note that (2.9.15) says nothing about m,.) However since we Ihave not fixed precisely the spin basis we have the freedom to add a multiple of o to ¢, and it is clear that this can be used to set f= 0 in (2.9.16). Now (2.0.14) and the fact that Vm, is symmetric imply vam = ahh for some scalar function g. Thus mV m, = FV my, = (2.917) We now set x = 2-!(2 + iy) and X, = V2, ¥, = Vyy. Clearly X,X* = Y,Y! = -1, XY* = 0 so that u,2,y can be taken as independent coordinates. Further (2.0.17) implies VyX = VV = VX = VY ° so that 2,y span a Euclidean 2-space. (The connection com: ponents vanish identically) This 2-space will be called a wave-eurface, We now choose a wave surface in each u = const. hypersurface such that the family forms a S:surface (with coordinates 4,29)10 2 Spins ‘We define a scalar field v so as to vanish on this hypersurface, and to be propagated by De=rVyw=1. (2.9.18) Setting V, = Vv we have on the initial hypersurface mV, (2.9.19) Now Dim*v,) = (Dm V* + (WV, )m"P 04 (W,Vj)m"l using (2.0.14) and VV =9,(Kih)m* = 9,(1)m" ‘Thus (2.9.19) holds everywhere, We shall use uu,2y as coordinates. It is clear that (1,0,0,0), 2(0,0,1,). It we set Hl, + Jv, +Om, + Om, ‘while (2.9.19) implies that then (2.0.18) implies that J ‘Thus 2, = (H1,0,0), where H is tobe determined. The line element (2.9.6) now follows from 9,5 = ah) ~ 2 Fy ‘We end by proving part 3. It is easy both to obtain the inverse of the metric tensor from (2.9.6) and derive = (0,1,0,0), n° =(1,-H,0,0), m* 2°17(0,0, 1,4). (2'9.20) ‘Thus D = 0/6v etc. We may now apply the commutators given in appendix B to the scalar fields u,2,2,y in turn. For example (2.9.20) implies 2.9. Plane and planefromted woes mu and the commutator equations applied to u give one | r= a+B, pap. After constructing all four sets of four equations like this, and recalling that Vl, = 0, we find that all of the connection scalars except » vanish, and DH=0, 6H =-0, (2.9.21) ‘Thus His v-independent. All of the field equations of appendix B are trivially satisfied, except for (j) and (n). (n) asserts &v = 0 in a vacuum, ic., 65H = 0, Since the wave surface is flat this is equivalent to Het Hy, (2.9.22) which is suficient to prove assertion 3) ‘The final equation (}) gives on | on Oy +" Be0y" (2.9.8) Baercise. Desve equations (2.8.21-23) v= Bu =~ We may write = Ae, where A(u,z) and 6(u,2) are real. If the phase 9 of the wave is constant the wave is said to be linearly polarized, while if Ais constant on each wavefront u = const the wave is called plane. If A vanishes for w u, the wave is said to be sandwich. Equation (2.9.6) represents a remarkable line element, We shall ye only one ofits features. A geodesic will be said to be type s in the 2-surface w = const. v = const., and type 2 if the 2-surface : = const. Type 1 geodesics reside in R ‘and s0 are complete, ic, infinitely extendable in both direction Type 2 geodesics reside in a 2-surface with line element de! = 21(a,2)du? + dauua 3 Asympopia ee Fig, 3.41. The hypersurfaces on which the ADM and Bondi masses ate reused, fiat asymptotically as one moved further away from it. By “going to infinity” one might acquire the Killing vectors necessary for integtal conservation laws, Another viewpoint lends credence to this ides, A “gravitational energy density” would presumably be 1 quadratic function ofthe connection coefficients. Since these can be made to vanish locally such a density could not be loealizable, but its integral over « large region might be well defined. ‘The investigation of ssymptopia depends on how we approach it, Proceeding along spacelike hypersurfaces does not_ produce very interesting results. Imagine an isolated source intially at rest suddenly emits « burst of gravitational radiation. Consider first a spacelike hypersurface which intersects the source during its ‘quiescent state, and which leads to a “mass measured at infinity”, the ADM mass, of M. Suppose a later spacelike hypersurface intersects the source during or afte its active phase. One might expect the mass of the source to have decreased because of the ‘energy emission. All of the radiation will however intersect this hypersusface, and allowing for its contribution one might expect the later ADM mass to be the same as the earlier one, which is in- deed the ease. Ifwe want to measure the mass loss of the source it is clear that we should bend up the later hypersurface so that the 3 Intaduction us already emitted radiation remains in its past, never intersecting If we bend the surface up until it becomes timelike then radiation emitted to the future of the intersection of the hypersurface with the source will intersect the surface a second time, and so will con- tribute twice to the mass! It seems clear therefore that we should ‘bend the hypersurface such thst it becomes null at far distances from the source, i., we should approach asymptepia along null directions. The corresponding mass is the Bondi mass and one might reasonably expect it to decrease for radiating sources. Another major problem is that we have no exact radiating so- lutions which become flat asymptotically. Thus we can only guess what structure to expect far from an isolated source, ‘This re- quires great care. If we set up axioms which are too strong we ‘may find that only flat spacetime satisfies them! On the other hhand if they are too weak we may be unable to draw useful con- clusions. Pethaps the best way to get a feel for what is involved is by constructing asymptopia for Minkowski spacetime, While it is certainly possible to discuss asymptopia by taking appropriate limits as “r ~+ 00” the anslysis hax to be done rather carefully, as wwe shall see in section 3.10. Thete is however a useful technical (tick due to Penrose for avoiding such limits. One performs an isotropic scaling of (proper) distances with the sealing decreas- ing to zer0 as distance from the source increases; the total sealed distance to asymptopia then remains finite. ‘Technically this is done by a conformal transformation. By convention we denote the original metric, the physical metric by j, and consider the transformation to the unphysical metric g given by (14) where 9 > 0 is a smooth function which tends to zero fat from the source. The adjective “conformal” is appropriate here since angles are preserved under (3.1.4). Unfortunately there is another convention in common usage in which the physical metric is g and the unphysical mets is 9, Le, (8.1.4) becomes (3.15)mm 2 Spinors By making a coordinate transformation U =u Vos et J H(ay2) du, with constant = we see immediately that this is 2-dimensionsl Minkowski epace, and so type 2 geodesics are complete. [Note further that there is some coordinate gauge freedom. 1 is defined up to a constant factor, and so w is defined up to constant linear transformation 1 @ = cu +d. m* ie fixed up to ‘rotation me" and the addition of a multiple of, and this gives Zee =e" + p(u). ‘The consequent change in to preserve the Tine element is [ov ~ Rlse"P'(w)] + 90] + where q(u) isreal. It is now straightforward to show that by means of a gauge transformation every geodesic can be converted to type 1 or 2. Thus the spacetime is geodesically complete. (2.9.6) Problem. The discussion given here is based in part on Pi- rani, 1965, and Bhlers and Kundt, 1962. An altemative discussion of plane linearly polarized waves is given in section 35.9 of Misner, ‘Thorne and Wheeler, 1973, Examine and compare these various approaches. (2.9.1) Problem. Consider the pp wave vacuum spacetime with Tine element ae Hy 2,9)da# + 2dude — da? — of thie section. Is the choice If = 1y(u)(z* ~ 3?) consistent with the vacuum condition? Make the coordinate transformation (4s942,9) — (UV.XGY) with SUj2say)X, y= bX v with —a'fa = ¥/b = f. Suppose the metric represents a weak sndwich wave, ic. f vanishes and spacetime is fat for u < uy and for u > u,. Determine the linearized solution of the field ‘equations, What happens ¢o test particles as the wave passes through them? What condition has to be imposed on f to ensure that the wave is sandwich? V4 Yea! + YH) 3 ASYMPTOPIA 3.1 Introduction title of this chapter deserves some explanation, Far from an isolated gravitational source one might expect an idealized sim- pile description of spacetime — asymptotic Utopia. However the remark of the British Prime Minister Chamberlain in 1989 on Caechslovakia is relevant here: “This is a far far away country bout which we know very little.” Just as in special relativity, integral conservation laws can be Aerived in the standard way from veetor divergence equations e.g VJ" =0. (44) However energy conservation is expressed locally by the vanishing of the divergence of the energy-momentum tensor T"* vir" =0, (312) ‘and such an equation cannot lead to an integral conservation law. (The integrand would be vector-valued and one cannot add vectors ‘a different points.) The situation is ameliorated ifthe spacetime possesses symmetries. For suppose K° is a Killing vector. ‘Then the current J defined by raTK,, (3.13) satisfies (3.1.1). For example the 4 translational Killing vectors in Minkowski spacetime lead to conservation of energy-momentum, In general however spacetime postestes no Killing vectors, For an isolated gravitating source one might expect spacetime to become6 5 Amymptopia We shall use this convention in sections 8.10 and 3,11 to minimise the number of characters in the equations. 3.2 Asymtopin for Minkowski spacetime We introduce standard Cartesian coordinates 2* and the standard Minkowski metric jy, = ding(1,—1,—1,—1). We shall examine some exainples of conformally related metries ay = %y (24) where My is smooth and non-zero ‘As a fit example we investigate y= 1/X%, where X?=Hyats', (8:22) con the complement of the light cone of the origin. The fi tion is whether we cas identify the “unphysical” spacetime with metric g given by (3.21). This turns out to be particularly easy if we introduce new coordinates y* via the equations jas'y, (323) aat/X8, ota ytly, Yt y 0 that dy! = =X"? [6 — 2Xta 2!) des. [Now a simple calculation gives fady'dy! = Xi gdetdet, 0 that sdy'dy! = Dighy detde’ (324) ‘Thus as seen in the y-coordinates the “unphysical” spacetime is flat. tie clear too that in the complement of the light cone of the e-origin, curves tending to z-infinity, ie., X* > skoo (where the sign is positive for curves within the light cone and negative ‘otherwise) approach the y-origin. I Aaymteia for Minkoush spacetime ut Consider fist timelike geodesics through the 2-origin aan, (28) where #xT*T* = 1, T° > 0, and A> 0, Tis clear that X* We may introduce @ new parameter p = ~1/2, where the sign is ‘chosen to ensure dy/d > 0. In terms of y" the curve becomes Note that, setting dy fd ee (28) ‘Thus a futuresdicected timelike geodesic through the a-origin ap- prosching z-infinity maps to a future directed timelike geodesic approaching the y-origin (from below), Consider next a timelike geodesic not passing through the 2- origin arse, where 4,7" are as before, and C* is spacelike and orthogonal to {T*. By means of a Lorentz transformation we can arrange that r (1,0,0,0), (sere) 1= O21 CR” ‘This curve also approaches the y-origin from below. However itis not a geodesic, for FO=T, HO) =-20" ‘Thus all future-pointing timelike geodesics approach the y-origin from below. All geodesics with the same z-velocity are tangent to cach other atthe y-origin, ie, they have the same y-velocity there However two parallel geodesics will have a relative y-acceleration ‘athe ycorigin equal to the spacelike z-separationns 3. Anyrptepia (3.2.1) Beercise. Perform the same analysis for spacelike geo- desies. By considering null geodesics as limiting cases of both familios of timelike geodesics and families of spacelike geodesics, show that a faturedirected generator of the fature mull cone of the z-origin becomes a futuredirected generator of the past null cone of the y-origin. How do two parallel null geodesics differ saymptotically? In practice (3.2.2) is a rather inconvenient choice of conformal factor. Consider physical Minkowski spacetime in spherical polar coordinates = dt — art art, (32.7) where ay? = do? + sin’ ay? ‘We now introduce the standard null coordinates vetty udu (828) so that A —ujPae’, (32.9) Now consider a, (e210) with iseyeey” (2.41) 4 (u- oF 2 a= ae ayy 2) ‘To interpret this metric it is convenient to introduce new coordinates Sarctany, q=arctany,

, where P = —m, R= 0. Spucsie geodesics traveling trey fom the origin have tp — oo, beygee dey pve oe R-1,T — 0, i.c., they approach spacelike infinity, I°, given by R= 2,70" Again the asymptotic bchviour of parle spaclle geodesics Isa lated eater To stuly the ssympltic behavieur ef mull geodesics note fist that conformal tonsormations map mall Yectors to ml vectors. Next consider radially outgoing null geodesics C= Pte eye oy wm ey where eis faite. Thor to inp pe Inie Pek my DH Ro 2p, ‘This is polat on the mill ypersuriace T + R= + which wil be called future wall infinity", (The sym bot is pronounced “seri™) "Similarly radially ingoing nal fodesics originale on past null infinity 7” given by T ft Note that paral) geodeis have diferent end points. ‘We can of course pec the representation of the eytinder abe taining a Penrose diagram for Minkowski spacetime, fig. 8.2.2 While this is perfectly satisfactory for spherically symmetric spacetimes, in more general sitwal ons the diagram is sometimes drawn 00 fig. 3.21 HMuwrerr while thin ta sesfect) alishatorsm2 3 Anmtepia near I+, itis « misleading picture of the structure near 1°, for I° is really a point (3.2.9) Beercise, Locate PI+,-./*,I~ for the compactified spacetime with 3.3. Asymptotic simplicity ‘An isolated system is one for which spacetime becomes more like Minkowski spacetime ax one moves away {rom the source. Using ‘our experience gained in section 3.2 we try to capture this idea in ‘definition. Unfortunately “asymptotically at” has been defined in a number of mutually inconsistent ways in the literature, and ‘4 new name is needed to avoid confusion. (The two definitions sziven in this section are now universally accepted.) (83.1) vermox A spacetime (if) is asymptotically simple if there ezists nother manifold (M,9) such that 1) Bf is an open submanifold of M with smooth toundary Off, 4) there exists areal function Mon M such thet guy = WPiyy 0m MM, 2 =0, VF 0 on dit, 7 8) every null geadesic has two end points on OW, 4) yy =0 near Bit. (4,3) is called the physical spacetime, and (Myg) i the ‘unphysical spacetime ‘This definition is adapted from Hawking and Ellis, 1973. If con- dition 4 holds it is not necessary to assume that VS # 0 on f 7 ‘The idea of conditions 1), 2) is to construct BUF as an “infinity” for spacetime (lig), which is to have some of the properties of “infinity” for Minkowski spacetime. However it is by no means clear that we have really captured the concept of “infinity”. To 3S Asymptotic soupy 3 demonstrate this we need to consider geodesics, and so we require the connection. (9.8.2) Bzercise. If 944 = Wigs, g** Christoffel symbols are related by 1-27", show that the Mea Bey 20 TA OVNI. (3.3.1) Let 7 be a null geodesic in (M,g) with affine parameter » such that v= 0 on Of. The equation of 7 is Bot yy dah dat Pe pe Ge tae pe 0 (332) Suppose jis the corresponding geodesic in J, Using a diferent parameter (0) and equation (3.3.1), this becomes et pa dab das wae =, (sas) where a‘ denotes a v-derivative. ‘This is clearly the equation of F, the equivalent geodesic in Jf. If we demand that be an affine parameter the right hand side must vanish. ‘This implies = const/9?, and absorbing the constant into # we obtain aoa er (334) Now on dif, @ = 0, V,0 #0, and we may choose » s0 that near BM, v ~~. Thus 5 ~ —1/v becomes unbounded; 4 never reaches dif, which apparently seally is “at infinity” and can only bbe reached by conformal rescaling. ‘This would appear to justify condition 3). Condition 4) is designed to capture the idea that spacetime shall be empty sufficiently fae from the source. As we shall se in section 3.5 it can be weakened considerably Clearly Minkowski spacetime is asymptotically simple. Are ‘here any other candidates? It seems reasonable that the conditions ought to include the exterior Schwarzschild space- ai = (1- Benfede! av". 2mjr) —r'eB* (8.38)1m 3 Anymptopia Setting w= t~ (r+ 2mln(r/2m—1)), 1= 1/r gives aa? = -9(P(1 — amide? ~ 2dudl ~ a3?) (3.36) ‘The analogous Minkowski metric (u = ¢— 1, 1 = 1/r) ai? =174(Pa? ~ 2dudl ~ a2") Now Af is given by 0 <1 < 1/2m. For M in the definition we ray take 0< 1 < 1/2m, 20 that Bit is 1 = 0. We abo set =. ‘Then on Mf , =0 , V2 = (0,1,0,0) 4 0. Thus conditions 1), 2), 4) ate satisied, However condition 3) cannot be satisfied, for there exist closed null geodesics at r = Sm which never “escape to infinity”. However as 1 0 the Schwarzschild spacetime does become flat asymptotically. We therefore weaken the definition requiring it to hold essentially only in a neighbourhood of = °. (6.8) veriwirion (8,3) i weakly asymptotically simple if there exists ‘asymptotically simple spacetime (31',5’) and a neighbourhood of OM in Mt exch that 0" (\AK' is isometric fo an open subspace O of Although it may take a few seconds to understand this definition it apparently includes all known examples of spacetimes one would wish to regard as flat asymptotically. Before exploring asymptopin further it is clear that we need to ‘establish the relationship between both the connection and curva- ‘ture tensors in the physical and unphysical spacetimes. For later ‘convenience we collect together all the relevant formulae in this section, ‘As suggested in section 3.1 we define Way 9 = 79". (34) 3.4 Conform traeformation farvlae 125 Consistent with this we choose AB a rtzAt, (342) plus their complex conjugates. Indeed unless we require ¢ to transform differently to & this is the only possible choice which js continuous with the identity. (A complex 0 would lead to tor sion, Penrose and Rindler, 1984.) It i important to note that we have not introduced a spin basis. Therefore equation (3.4.2), ‘and indeed all subsequent equations, are to be understood in the abstract index formalism. When we introduce spin bases we shall take care to introduce a diferent notation for spinor components. We also require a covariant derivative V..y on the unphysical spacetime. ‘This must satisfy the conditions of definition 25.1, plus equation (3.3.1). Some experimentation shows that the only plausible candidate is defined via the rales: 1) Waax = Vax forall scalars x, 2) aba = Vals + Toe Fanny = Vanna + Taner where Tay = Vay ln, 3) Fak! = Vank? ea? Tonk Faun Vag ~ eu Tons ‘with the obvious extension to higher valence spinors, 4) Vos = Vas + Ty ~ sag "ETe The formal existence and uniqueness proofs given in Penrose and, Rindler, 1984, (3.4.1) Beereise. Show that Paakec= Fy =O, > Vantage = Vane? =O. (3.4.8) Beercise. Check that rule 4) is consistent both with rule 2) and with equation (3.3.1),126 5 Anympopia We need to establish # rule for index raising and lowering. If the kernel letter of spinor caries a tilde then 22, Z,y are to be used; however if there is no such tide then the standard ¢4?, are used. ‘As one might expect some key concepts turn out to be confor sally invariant. Asan example let $45. be totally symmetric spinor representing a zero rest mass feld. Let $40.0 = Banc (343) ‘Then the sero rest mass field equations are conformally invariant, “Oyc= 0 > Vb c= (SAA) (3.4.3) Bzercise. Verify equation (8.4.4) the transformation formulae for the various parts of the Rie ‘mania tensor are quite complicated, although considerably simpler in spinor form than their tensor counterparts. We merely state them here, relegating the derivation to a tedious exercise, ‘They Vvco = Vanco (3435) $V AVA HPoaM(VED), (84.8) ape FOV Vayet ——(B.4N) In vector form the last two equations are Rama-lonn+ Uo.nyv'm, (8.48) 28, +.97'V,9,0- 1AM on (349) where 0 = ,V. (04.4) Bees. Oblain the ration between Cay 48d Ogg met 1g on a spinor €, oF 24, using rule 2. Hence, using the Ricci identities (exorcise 25.5) obtain equations (8.4.5-7) and (8.4.8-9) 35 Feud pecking a7 3.5 7 and peeling ‘Using the formulae of the last section we can now derive some of the important properties of nll infinity. For convenience we set 18, = =¥.0, noting that Vigny = 0. The boundary Oi of it is 0, and 50 n, is normal to it. For the results that follow condition 4) of definition 8.8.1 for asymptotic simplicity can be related in a number of ways (3.5.1) uemsca Suppose the energy momentum tensor of “physical” matter falls off suficiently fast near Bit Ty =O) as 20, (352) Then the boundary Dif is « null hypersurface. Proof: Using Binstein’s fel equation and assumption (3.5.1), we deduce that near Oi A 01), y= 010), We see that A vanishes on iif, and then it follows from equa- tion (8:48) that myn" sist vanish there. Thus OA ia w aul hypersueface Locally Jif must be either to the past or the future of Of. (8.5.2) pernunion 17 Ht les to the past then the future end points of null geodesics form a null ypersurfece J* or future mull infinity , while fit lies to the future the hypersurface is past mull infinity, 7 Itcan be shown, see eg. section 6.9 of Hawking and Ellis, 1973, that each of these two hypersurfaces is topologically S? x R. For the next result veneed s slightly stronger assumption about the matter.128 3 Asymptopia (8.5.3) tHeonEe Suppose that T* vanishes in neighbourhood O of Then ¥jnco = O(9) in B, and in perticler Vann vanishes on J. Proof: The vacuum condition implies that 94 yyoy Noting that Wye = Pawon fom (3.4.5), and using (3.44) we have U4 (0"W .gcp) = 0.08 a0 W pcp = a, (352) By continuity (3.5.2) holds also at % But from lemma 3.5.1 n* is qull on Zand so there exists an «such * there. It follows (cf. theorem 2.94) that there a such that Vasco = Vatatcty ont (353) Next we differentiate (5.5.2) again and restrict the equation to. obtaining OV ep Pann + ¥anco¥ ent (354) Note that (81) implies Vaxetayta on Performing this sym- retriration, and using the Jacobi identity, equation (2.2.5), we find ftp VE v0 FigeVayeHancn =O on, which implies that Ve» (Wlatateto ont. (355) Te will be shown later (lems 3.94) that this implies thet vanishes on Zand #0 Yancv =O) ath (35) This is a special case of general result on peeling. Let 1 be ‘8 null geodesic in M reaching J at the point p. Let ¥ be the corresponding null geodesic in Af. We choose a spin basis (5,7) 35 F and peeing 129 at one point of 7 with I tangent to 7 there, and we propagate the basis parallelly along 7 vie De = Di* =0. (35.7) aa = ditt, wher 4. Now we are fece to choose whatever conformal transformation behaviour we wish for 0 and « provided that e*? = o4\? — «Ao? transforms according to (3.4.2). The most convenient choice in an affine parameter along thie context is which implies Now using (8.5.7-8) and rule 3) from section 8.4 it follows that Dot De4 = (2-5) (3.5.10) But since n,, 1" are normal to.J,n*, m* and sh are tangent to.F, and since © vanishes on Fit follows that "80 is regular there, Thus o*, 4 are regular on.¥ Further since Do = 0, 7 is affinely parametrized. We choose the origin and scaling of the parameter Pio set an a ate. (3s.n) Pquation (85.9) implios that, acting on scalars, D = 9-*D ot r= OF, which shouldbe compared with equation (3.24). Thas from (38.11) we may concade FeO on z (3.5.12) Phe next resll is often refered to as the peeling theorem.130 8 Anenptopia (854) te0Rem Suppose thal gap. Feprezens spince ero reat mass fed which is regular at J nd let Hy = ba .c0*0? 8, where there are is’s and (22~1) 0's. Then the phyrical componente satisfy 1-0) (35.13) Proof: Clearly the dy are regular at Z But eN(O-)(O789).. Ee, ‘Thos 9-29-09, is regular at in other words the physical components satisfy (3.5.13). (8.5.5) Beample, Consider electromagnetism, for which ¢ = 1, Sav..c = Bau: ASin section 2.6 set Bane”, $e) = 840081", dey 40) ‘Then equation (8.5.13 ane since the lending 1/¥ term represents a null electromagnetic ficld. [But as we have seen such fields correspond to plane waves. Thus 86 The choice of conformal gouge 131 far from an isolated source the leading term in the asyw ficld is « plane wave, while the next order term is algebraically general. (9.5.6) Protlem. Show that the gravitational feld far from an isolated source can be written schematically as Sonep = M14 WH, EL, +0" where [iV] represents a Weyl tensor of type WV ete 3.6 The choice of conformal gauge The conformal factor used to compactify spacetime has never been fixed precisely, except in examples. It ean be multiplied by any real strictly positive scalar field without changing any ofits prop- crties. This “gauge freedom ean be put to good use in specific calculations. In this section we set up the key equations which can be used to determine a specific choice. ‘We shall make a slightly stronger assumption about the physical ‘matter than in section 3.5, namely that in @ neighbourhood of 7 By = 010), (s6.) and further that this asymptotic relation can be differentiated ‘once. It follows immediately that 8,, = ofS). Then, setting n= —¥,0, equation (3.4.9) implies 98,,-V.n,-4 (362) We now differentiate (8.6.2) obtaining Bay WVHA VV. +b Now setting 2 = 0 gives AHF VAVint iV) =0 om (3.63) Next consider identity (3.4.8) in the form AA HIT = tan’ + of) (3.04)ase 3 Anmpropia Differentiating this equation gives }OV, (GM) ~ $9,009 - n°V.n, + 0(8) 1V,(0M) ~ fn, n'N4, + 2 +0(0), 79,4 —20An, where equation (8.6.2) has been used. We now divide this equation by and subsequently set £2 = 0 obtaining on Zé (8.65) In, = -}9,(00) +08, We can now explore the freedom left in 2. As stated above we should consider the changes a f=, (30) where 0 £0 on.F. (This condition arises because we have postu- lated V0 # 0 on. If@ vanished there then 1 would vanish to second order.) (6.6.1) uenca Under a conformel gauge change (3.6.6) we have on F y= 0m, 9.8, = 00m, +5n'V 8 (3.6) Proof: Set gy, = MG, and Guy = VGqy 20 that 94 = 09, Then setting , = V0, we find where rule 1 of section $.4 has been used. Rule 4 implies ¥, VF — May /0+ a9 ty) OV amy tty Hngty — OV by — Many 2(02/8)tats + 9a9"ln.ty ~ (9/6)t.t ‘These equations simplify considerably on giving (3.6.7). 3.7 A spin bass adapted to. 133 Lemma 3.6.1 is the key to the choice of conformal gauge. Per- hhaps the most useful choice for 6 is that which sets VA, = 0, To see how this is possible aote that equation (3.6.2) implies 9400 = Vn, on.F Hence on F 29480 = 0, = 09m — gyn" 0 = la(0030 ~ 440). Now A is the directional derivative slong the generators of Suppose that on some intial 2-surface Sin F transverse to the jenerators we choose 0 arbitrarily. We ean then propagate 8 via a0 jon) on% This sets 7, = 0 on% We shall assume that this has bbeen done, and henceforth drop the earets. ‘The freedom in the choice of on S, can be expressed as residual gauge freedom Mah=0n, a6 0 om % (368) ‘and will be used later. 3.7 A spin basis adapted to.” Jet us recall that we have defined mz = —V,Q where n® is on f, and have used the conformsl gauge freedom to set Vin=0 ont Clearly there exists & spinor « defined in a neighbourhood of 7 such that We can complete « to a spin basis (0,4) in this neighbourhood, Now on 7)134 3 Asynptepia and so there exists » 8, such that = Btw (32.1) © implies 1(9, ~ B,) = 0 s0 that , must ier experience in the proof of theorem 2.9.4 it seems worthwhile to explore the possibility of making & vi, [As we found eater, it may take some time to prove an apparently ‘subsidiary result but this is ultimately time saved! (8.7.1) TH 0REM A phase change can always be found to ensure that isco constant on. Proof: Te is clear that (3.7.2) will hold on it Vai = 118, - VP Niaw there. ‘Thus in order to obtain (3.7.2) on itis necessary and snfficient to be able to solve ve (3.73) there, On must be a gradient, but we cannot conclude that this has to hold in a neighbourhood of 7% since © vanishes at J. ‘The integrability condition for (3.7.3) has to be weakened to B= VA+20,, (37.4) in a neighbourhood of for some vector field v, and some sealar field A. An equivalent formulation of the condition (8.7.4) is that oat VB = Me (318) 37 A spin bait dap te. 135, ‘There is an equivalent spinor formulation. Equation (8.7.1) ean hhold on-7if and only if there exists a spinor field 74¢ defined in a neighbourhood of such that Vote = Hite + Mer there, This equation implies VaVete = VA + BB) = thes on. Thus an equivalent formslation ofthe integrability condition (87.5) for equation (3.7.3) is that on 5, ne (3.76) for some spinor field Loy, We now recall that ot 7¥ yoy vanishes, and so from equation (2.5.6) and excerise 2.5.5 IL Vate = Cea Tante + Caan Ae atynctan + oPeoa at” ‘hus the integrability condition (3.7.0) on-F can be reformulated ~Meutmctow teuPerwwt? = Mclyer (87-2) there, Although it may seem somewhat inelgant itis advantageous to extract the 32 scalar equations in (3.77). Tt i clear that on snultiplcation by cc the equation becomes trivial, while mal Aiplying it by 04 is equivalent to mtiplying i by eo", The 16 remaining equations either define consistently components of Liye ot produce the equations B=, = 0, FAS Uhe vanishing of @,,, by, is already guaranteed by equation 116.5) since 00 1.) red on ¥. Thus the integrability condition ste the sin ASO. (378)138 3 Anymptopia Can we satisfy it? "We are assuming. is topologically ? x R. Let S be a cut of J, ise 4 spacelike 2-surface in.Forthogonal to the generators of J ‘There exists another null hypersurface 3. (at least locally) which contains $ and whose generators are orthogonal to S. We use Ds to define a spinor feld o on $ as follows: o%6* is tangent to the generators of Bs, I follows immediately that x = 0 on S. We may propagate 0 from $ onto.¥ via the condition Ao* = 0, which implies that "7 =O on. We have yet to use the restriction Vin, = 0 on.F which produces depe at paesin (379) on % Field equation (q) of appendix B implies Ap= 2. (3.7.10) Now the assertion that p is rel is equivalent to the assertion that is hypersurface orthogonal, cf equation (2.7.6). By construc- tion 1 is hypersurface orthogonal on $, to that p is real there. ‘But then esuation (3.7.10) implies that p is real on.F, and so ! is hypersurface orthogonal at each point of Z Thus our initial cut hypersurface, $/.5, define a family of cuts/hypersurfaces on % Now from the D, 4 commutator we deduce immediately that [ell = (74 DIF (e42)n— (Ft lm (7 +) which vanishes on J ‘Thu this family of eats/hypermurfaces can be obtained by Liedragging S/2.- along.7. On S, the 2-meti fu of! = =m, and the 2-connection is essentially m*Bm, = {0 . The 2carvture tensor can have only one ral independent ompenent, and the defining equation for R® rust be ofthe frm] fan 94 deems quadratic in the connection. Inpection of eld ation (0) ofappendix B shows that to within a factor R® mut be, +-A- Thus the integrability condition (31.8) is that the 37 A apn basis adopted to F 13 cot Sand those obtained by ie lished? Now the rennining conformal freedom is given by equation (3.6.8) viz. : ian ing it, be flat. Can this be a6, a8 0 ony with @ arbitrary on S. Under such a transformation R? 0° 42009 2v7ev,8. (his is essentially equation (3.4.8) but with differen, dimension: dependent coeficients.) We can always choote @ on S to ensure that R°) vanishes there. Now a 9a ~ By ‘oo Weamy = 0 implies £ ay = 0 implies £5 = 0, and by construction £,t 0. "Thus 98) =O on Sand so f= 0 there. Thus our dhe of on ha ened that ech fhe st of 7 Bt is the integrability condition (27.10) for equation (31.3) ean be satis fo ‘Thus we may always chose one (ar) ‘This ensures that in addition to (3.7.9) above we have on. (3.712) However there is still some conformal gauge freedom left! We may set Q-f=ro, with ont. (a7.13) Under (3.7.13) n* and its gradient change, but as examination ‘of equation (3.6.7) and the argument leading to it shows, the condition F = 1 on. implies aA Vin ont Thus nothing that we have done already on. T's affected by is acon. smal gauge change (3.7.13). Next we sct |, consistent with,138 8 Anyptpia Dp =0 = Dp. ‘Thus all of the NP connection scalars except ¢ vanish on J. In fact o contains all of the dynamical information regarding the ‘outgoing radiation field. (The gravitational peeling theorem shows that on-7the dominant I/r term in Bycp is @, and we regard this ae determining the outgoing fleld. If the roles of I,n were interchanged the dominant term would be @, and so we regard ¥, as describing the incoming field.) To see the réle of note that since Vagep vanishes on J the field equations (a,k,p,q) of appendix B imply Bq =-08, %=-be, &_=—As, Fy, and in addition equation (b) implies that Do = 0. In older accounts $j. = ~A¢ is called the news function, N. ‘We can extract some further information intrinsic to./from the Bianchi identities. We frst set ¥aven = %ancor and note that since from equation (3.5.11) we may take DQ = =1 on. then DW, = x, there. The Bianchi identity (Bg) of sppendix B implies immediately usa. (e714) ro that these that Equations (Bj,k) imply that D&,, = D&y, ‘quantities vanish to second order. Now equation (Be) impli w= 68, (a.115) Finally although DA need not vanish on it must be real, and 0 (Be) implies xa toe +89 = 5, 4200 +8. (3.7.18) 58 Asgmpticsynmetry 139 We are therefore justified in regarding ¢ as measuring the outgoing asymptotic gravitational field. The other NP quantities are “pure gauge” 3.8 Asymptotic symmetry In the previous section the cuts of were locally fat 2-surfaces, i.e. locally coordinates 2,y could be found so that the line element of the cut was ast = da? — dy’, We must emphasise however tht this is deceptive globally, for cach cut is topologically a $*. We may however include the global topology at minimal cost by regarding the cut as a Riemann sphere with independent coordinates (C,@), where Ca etin nin, and ae = ~atat. (3.8.1) Although this may appear merely a cosmetic change, it is rather ‘more than this. For we may now make an (inverse) stereographic projection =e cot 10, finding after some algebra EP a1 4 CaO, ast a aos soap. (982) "hus by a conformal transformation with factor Pt where P L(+ 60) = Seosec $9 cach cut becomes the unit sphere, We shall vse this alternative representation frequently. ‘The construction of section 3.7 established that given one cut there was a preferred spin frame and system of cuts adapted to ‘hat cut, We shall call this a Bondé system based on that) he only thing that we have aot fixed is the coordinate chart ies that all aris are exqually good. However we140 3 Astonia may prefer those adapted to the above structure, and this ean be used to define an asymptotic symmetry group. Pot ‘Consider first a relabelling of the generators ¢ + ¢ = 2(¢,2) I we require the transformation to be both conformal and locally proper, then we must require B= 0 OMe. Writing @ = 24+ i9 we find immediately that o_o BOF, Be By’ ye" so that @= J(¢) where fis holomorphic, This fixes the form of df, and since the result say not be known we present it formally fas a lem (3.8.1) ven Let f be a holomorphic bijection from the Riemann sphere to itself Then fis « fractional linear transformation (FLT) ach (38.3) wae (383) 10) where, without loss of generality, ed —be = 1, (In fact provided f is not constantly infinite we may drop the bijectivity requirement.) Proof: Let T be s FLT which maps {(0) to 0 and f(co) to oo Then 9 = Tf isa bijection which leaves 0 and co invariant. If (0) ~0 then g could not be a bijection in a neighbourhood of © and to we may conclude that o(6) = CR) with A(0) non-zero. However HO =a = ET bijection mapping 0 to 0. 7 HO 1 oT is also 8 holomos 28 Asymptotic eymmetry at hhas a non-zero limit as (+0, Thus h is a bounded function and by Liowville’s theorem A is constant, It follows that T/(C) = KC for some constant k and so f must be a PLY. 1 conventional he normaliaation 1 follows that if we wish a cut to remain the unit sphere under (8.8.3) then we must make another conforn I transformation a8 = Keay’, where K(C,G) ean be deter 1d from the equations above as +g CFIC + (CEH ‘Thus lengths within the eut scale by a factor K under the trens- formation (3.8.3). For our theory to remain invariant under this rescaling “lengths” along the generators of f must scale by the same amount, ie., we need to impose (3.84) which integrates to give K [ut ald]. (3.85) The transformations (3.8.3.5) together with formula (8.8.4) clearly form a group, the Bondi-Metsner-Sachs group (BMS) ot asymptotic symmetry group. The largest proper normal subgroup is the eupertranslation group (S) walt, G=6 (3.88) ("The reason for the name will appear below.) ‘The factor group GMS/S is the group of conformal transformations of the sphere (5.8.3), whichis isomorphic to $£(2,€), and represents the proper ‘orthochronous Lorentz group. ‘Thetis also a 4-parameter normal subgroup, the translation group (T) given by (3.8.6) with A+ BC+ BE Tae . (387)Me 3 Annona where A,C are rel. (8.8.2) Problem, Let t,2,9,2 be Cartesian coordinates in Min- art sat ty? +2, Introducing owski space-time and set u =e cot 10 show that (et iy) sr), 2=r(C +02, y=nir((-0)4, 2 = (Ce -1)2, where Z = 1/(1+ ¢@). Now make a translation nr feta the, erat aath yrey Show that wera ut (A+ BC+ BE + C6) + O(1/r), Cas C+ O(n), ~d, Thus a real translation in Minkowski space-time generates a member of 7, thereby justifying the name. The “supertranslation group” contains all of these, and of course very many more. (3.8.3) Problem. Working in frame in which the Bondi cuts are flat, show that under a supertranslation (3.8.6) G2) = of a(6,8), 648) + FaL6,8). A good cut is one on which o = 0, ¢ that produced by the null cone of a point in Minkowski spacetime. (A Bondi system of good ceuts can be found in any stationary asymptotically simple space- time.) Show using problem 3.8.2 plus @ conformal transformation, that in Minkowski spacetime the subgroup of supertranslations preserving good cuts is precisely T. Next consider the following riodel, An isolated system is initially stationary, it starts to rar diate and then stops, so that the spacetime eventually becomes stationary again. Are the initial and final Bondi systems of good cute related by a supertranslation, and if so is ita translation? A symmetry group of considerable significance in sp tivity is the Poincaré group. A Poincaré transformation may be 8.9. Spinaighed spherical hornenics ua regarded as the composition of a Lorents transformation and a ‘translation. Similarly a BMS transformation may be considered 8s the composition of a Lorents transformstion and 4 superteans- lation, Thus there are many Poincaré groups at % indeed one for ‘each supertranslation which is not a translation, and no one of them is preferred. This causes considerable difficulties, e.g. there is no generally accepted definition of angular momentutn 3.9 Spin-weighted spherical harmonies In order to simplify the subsequent discussion we shall introduce the machinery of spin-weighted spherical harmonies. ‘These have plenty of other uses as well, particularly in the representation theory of O(4). The standard reference is Goldberg et al, 1967, bbut in order to conform to the conventions of Pentose and Rindler, 1984, some changes in the notation have been made. ‘These will bbe deseribed later Let be a sphere of radius with line clement as = REO" + sin 846 tm myde"da", (3.9.1) where m,7i span 7S), are null and m,7i* = —1. Note that this ‘Hoes not fix im for there isthe “gauge” freedom. (39.2) where ¥ is teal but otherwise arbitrary, This is called a apin through y. We may project space-time vectors and tensors into S ‘using the projection tensor P: the projection of wis P,eu® where M42 = ~min, ~ mitmy. (Projection simply discards components transverse (0 S.) Let n, 4,.., be a tensor in S, ie, it as been rojected into S on every index, where there are p indices in the Vist index set and g in the second. Let Demmi T, Ue is clear that under spins (2.9.2) the scalar 9 transforms as wet, (3.9.3)ua 3. Arymgtopia where s = p— ais the spin weight of 7. By projecting the vector equivalents of equations (2.6.11) (which ave given explicitly in appendix B) into S we obtain fm’ = (4— Pym", 6m" = (4— Pym, iS We can now define new derivations in S, viz 3, 3, pronounced edth, edth bar” by: By = me ge 6 MEV My ty. 28 3.94) meng Oe) where V, isthe covariant derivative in S, or equivalently the pro- Jection of the space-time covariant derivative, It is easy to ee that if y has spin weight s then 3n has spin weight «+1, and Dy has spin weight # — I. Thus acting on quantities with definite spin ‘weight 3,3 can be allocated spin weights of 1 and -I respectively. (8.9.1) Beereise. Show that acting on a quantity 7 of spin weight 1 3 and 6 ase related by a Bn in — sa - Bn. (39) ‘Thus neither § nor a,B have spin weight, ie they donot transform homogeneously under & spin. '} and 3 can be expressed explicitly in terms of polar coordi- nates (0,4) or complex coordinates (,¢. It is easy to obtain the Christoffel symbols for the line element (3.9.1) as nt Thy, = —sinOooe8, TY, = —cot®. [An approptiate choice for mm is RA, ~isin®), (3.98) which is consistent with (3.9.1). Then a trivial ealeulation gives <= R(1,iconee®), m= att aR 3.9 Spine phere! ren us Now we can write (3.05) explicitly as _ (ing (0k zl on= CF (Ge snag) (08) jetting i 8 ; By = OO (55+ seas) linea ‘The operator 38 clearly has spin weight nro, but it is not the same as 3. It is easy to derive the commutator elation (92) (3.9.8) Exercise. Show that if 7 has spin weight » then (~ 93) = En (39.8) With this representation it ip possible to define an orthogonal set of functions which are complete on the sphere, and are indeed tigenfunctions of 35. The conventional spherical harmonies will bedenoted ¥;,,(0,)- The spin-weighted spherical harmonics are defined as follows oY n(O8) = Vind) (3.9.9) yer fata oy, 8,9) O 0 Vhen if By = 048 fallons that y ~ th Ms Oy = 0 implies48 3 Anmptopia only that 9 is @ linear combination of the ,Yyq. The equivalent results for # <0 are obtained by interchanging 3 and 3 in the ‘above assertions Proof: By completeness we may express 7a (3.0.11) Different (3.9.12) b= SS bananas Finn (8.9.13) where the non-vanishing factots Gyjay bm £80 be deduced from (3.9.10). Next suppose that 3n = 0. By corupleteness each term in the sum (3.9.12) must vanish. Since neither dy 0¢ ,-; Yin ¥enish for 1 > 4, it follows that each ¢,,, vanishes, ie., 7 = 0. However 1 similar argument applied to (3-9-13) has to be modifieds since vaYou Yanishes identically the ¢,,, are undetermined. A similar argument can be ured if s <0. ‘As an immediate application of this lemma we complete the argument, started in section 3.5, that the Weyl spinor vanishes on % Weis clear that the field y defined by (8.5.3) has spin weight 5 = 2, since Wage cannot earty spin Weight. Also transvecting equation (3.5.5) with of gives By) = 0 on. Thus by the above lemma y= 0 on.Fwhich implies (8.5.6). We can alzo write 3,3 in terms of complex coordinates such as GG. A complex coordinate € is said to be holomorphic if the lines R(E) = const, are orthogonal to the lines &(¢) = const., the spacing of each family ir the same, and a rotation through = converts the R(E)-increasing direction to the 9(¢)- increasing one, ‘Clearly a holomorphic function of « holomorphic coordinate is also 8.9. Spinsscighed spherical harmonics ur ' holomorphic coordinate. An obvious example on the Riemann sphere is € = 2 +iy. Now from section 2.3, Heia tie) ‘Thos if € is holomorphic we cxpect Vaf to be parallel tow, Not ing that m?m, = 0, Mm, ~~] and (0/06) = 1, d8(0/06 vee may 1 Bag, (3.9.14) where we shall assume that P(€,é) is real. (This restriction is easily removed.) Then le = 2m rade 2eedt ds? = —2m, mydeta = ME (3:90:15) One might have expected ¢ = eat 0 to be a lnnspic coordinate onthe type, In thal cae =InG = eee hs ig would also be holomorphic. However Loe d ~ cosee 048 + idé. ‘Thus the lines (€) = const. are lines of constant latitude vith ‘R(E) increasing towards the north pole, while &(€) measures lon- situde, This would be holomorphic if we took the normal to the sphere to point inwards, Current conventions require the normal to point outwards, and so a suitable candidate for coordinate ie holomorphic AE = cosee Odd +idg. (8.9.10) It is however more convenient to take ¢ a8 our holomorphic ‘ordinate g Comparing equations (8.8.12) and (8.9.15) now shows that for a sphere of radius R and outward normal mee 8) ie (3917)148 3 Antonia If we had choten { = ~In@ as our holomorphic coordinate then the me? defined by (3.0.14) would be parallel to the m* defined by (G.0.8), and the foomula corresponding to (3.9.7) could have been read off directly. Instead we procsed as follows. Equation (3.9.5) implies that acting on a acalar of vanishing spin weight, 2 = 6. Then (3.9.14) implies that P = 9¢ since ¢ has no spin weight. We shall require the commutator relation (3.0.8) to hold and x0, BP =30f = 85 =0. Now suppose that has spin weight s. Then P-'y has spin weight 0 and so d(P~*n) = PA( Pr’n)/8, of Bn = P!"8(P-*n)/0¢. We write this equation and its complex conjugate as ay Selo (3.9.18) an FelnP'), n= P ‘These are often more useful than the original representation (8.9.7), and indeed can be used to generalize d to act on objects of, non-integer spin weight. (See exercise 3.9.5 below.) It is important to realise that choosing ¢ rather than In as our holomorphic coor- dinate implies that the 3,3 defined by (8.9.18) are not the same as those defined by (2.9.7). Although the Yj, defined by (3.9.9) and (8.9.19) below are the same, the spin-weighted spherical harmonies ‘abo differ. The reader has been warned! (3.9.5) Beercise, Suppose that A, B have spin weights a,b where +b==L, Show that =f noayes, where d5 = Pi (9.9.6) Bzercse. Verify the commatator relation (9.8) wing the representation (29.18) ‘A representation ofthe standard apherial harmonics is given by ye (2s) mbm i (29.9) (ayreengenn Lama S10 Anyuptti asation of the fl equations 9 where the sum is from max(0,—m) to min(!—m). The spine weighted spherical harmonics become, in this representation mt a eym (EEE) ULE gL at + myer = =o" (BE) nee oo Fe a) eam a= ay (3.9.20) where the sum is from max(0,s~ m) to min(d 4! ~ m). (3.9.7) Problem. Compute the four independent quantities a2¥igt- Consider the set $ of quantities of the form af, where cach‘of'« and 0 is one of the above harmonics. Of what group is S a representation? The operators 5,3 and the spin-weighted sphericel harmonics are explored in some detail in the atticle by Goldberg et al, 1987. However they consider a sphere of radius 2-"” with a positive def ine signature and inward normal. Thus their operator, denoted Dy satisfies Bo = ~2' RB ‘The sign change comes from the signature, and the factor from the change in radius. Since they regard ¢ rather than @ as holomorphic they have effectively interchanged m* as ‘omplex conjugation. It also means that ‘m*, and this causes the "The spin-weighted spherical harmonics defined here are the same 1s those of Penrose and Rindler, 1984. It should be noted how: ver that although a number of earlier papers made use of the harmonics and d there appears to be no consistent notation, 3.10 Asymptotic solution of the field equations ln the later sections we shall need properties of the space-time in ‘neighbourhood of. ‘The information obtained in section 3.7150 bays applies only on. We must therefore develop asymptotic expan sions for the NP scalars. This was first done by Newman and Unti, 1962, and most of their successors have adopted the same technique, With a little foresight and a carefal choice of coor dinates and tetrad however much of the tedious algebra can be avoided. Because almost all of our calculations take place in the physical spacetime we remove the tilde from physical quantities and attach a caret to unphysical quentities, eg, (81.5) Let (Ma) be an asymptotically simple space-time. Suppose there isa I-parameter family of nll hypersurfaces N, intersecting in ents 5,. We may use the parameter to define a scalar field win M via w= const, on M,. On each geodesic generator 7, of N, we choose an affine parameter r. On any one cut of say 5, we choose atbitraty coordinates (2°), a5)... = 2,3. We propagate 2” onto.” by requiring 2” = const. on the generators of % Finally we propagate 2° into M via 2” = const. on each 7, Now (u,7,2") is # coordinate chart in a neighbourhood of % The remaining feedom is une Ay), a relabelling of the null hypersurfaces, 2) 2° 43° = 3"(24), a different choice of 2” on Sy, 3) rev F = o(uy2")r + (uy2°), the sealing and origin for r on each generator 7, We next set up the NP tetrad. Let Yau. (3.101) ‘Then I* isan affinely parametrized geodesic vector, and so we may use the scaling freedom in r to set FY, = 0/0r (3.10.2) Sy At each point pon S, [* is future pointing and orthogonal to S,, ‘There ie precisely one other null direction at p with this property, and we choose n* to be parallel to it. Finally we choose m*, to span 7(5,,)- The freedom remaining We next label the surfaces u = const., 7 = const 1) a boost I+ Al, n++ A-*n, which allows us to rescale r again, B40 Asyplicsltion ofthe ald equations 151 ava jin mo em, with 0 real From the construction we deduce that nV, = 0/0u+ Q0/dr+C°9/62", mV, = "8/03", (10a) where Q, C*, P* represent 6 real functions, the metric compo- nents, (We may always recover the metric from 9,5 = 2amy) — 2m,ym%.) Further C* = 0 on % (NB. Since this is not the Newman-Unti tetrad, the notation has been changed,) ‘Under a spin, ¢-@ =m Din, 4 ¢-#-4iD9, Weshall choose Dé o set e=Z everywhere, However 9 can stil be chosen arbitrarily on any one surface r = const. Now applying the commutators to 2 gives F=O46, pandyarereal, (3.10.4) and the “frame equation Pe i (Fa) Do* =24P* 4 2P (Fo) DP* = pP* oP", (Fe) ape 6c" a aP* ip (Fe) Fn For this and the next sections we are denoting the “unphysical” ‘quantities by a hat. It is easy to see from (3.5.8), (8.4.5) that M-"G,. Since we know that near J, = O(N) we deduce ¥, own"), (3.105) Using the definitions, the rules of section 3.4 and (3.5.8) it is straightforward to show, assuming the NP scalars to be regular near J that @ is O(08), 7, a, 8, = are O(M), 7, my Aare Ol), ¥ is O(8-') and pM DO 4 O19")152 3 Ammonia We choose DV = dM fdr =~? so that p = —1+0(0). Now we Ihave not yet fixed the origin ofthe affine parameter r, and writing pin terms ofr its enay to se that we may choose this origin to set 2-4 O(9?), We next set ¢ = #0+0(0) where 5 = s(u,2") dependent. Substituting these expressions into the field =nd/do we equations (2, b) of appendix B and noting that obtain =a 40(9'}, = 48 O(N"), (8.108) We expect m* and hence P* to be O(M). Setting P° ‘p°(us2")-+ O(0) we find immediately from equation (Fe) that Pv = pO sp°0? + 0(0?), (a.10.7) It we set 6° = p°V, this can be written as 5 = 18" — 5988 +0(0"). (3.108) Writing o = a+-a1M? + 0(00), with a similar expansion for field equations (d,e) reves! A ast? + 010) (3.10.9) Be ‘The leading terms in equation (Fe) then show that 68pe = Boye = ap" + 247", (3.10.10) Given a choive for p we can determine a. We now define 3 acting con a spin # scalar 9 by an ‘consistent with section 3.9. The leading terms in field equation (&) now give a! and so Sy + 26a, By = Fy Been, (10.1) + (98+ a3) + (0°), (3.10.12) Since + = 4-+ 8 we have Be? +010" (32043) 8.10 Asytotic solution of the field euations 153 Also the Bianchi identity (Ba) plus the estimates (3.10.5) imply Y= VIN +010), ¥, = HOt GuN)A* + O(n), (2:10.14) where the superfix denotes an independent quantity, We have expected C" tobe O(1), but our coordinates and frame were chosen that O° vanished en Aasuning "= O(0) equation (t8) revs crv, =— [09° +G9i"]o* FON, (1018) ‘The next group of equations are a little trickier to get started, ‘The Bianchi identity (Be) plus the estimates (3.10.5) reveal ¥, = WIM! — GHA + (0%) (10.6) We expect = O(1) and field equation (f) revesls = [abs — 008+ 198] 0° 4 0(0°) Equation (Fa) then gives Q=- 0" +5240" — HH + HO + O(0"), Next we set Daw MOs MAE 4 O(W), With a similar expression for u. Equations (g/h) immediately give 2° = p= 0. But now the leading, O(f) terms in equation (Fd) reveal that 4? = —!Ap"/p" where A = 9/du. Now pis fee data ‘on.Fand we shall slays choose it to set Ap" = 0. Thus we may conclude that" = 0, and y [Bs-aar+tes]ar +o). (autour ‘The leading O(9) terms in (p) give A! = 48. The leading O(9") terms in (I) give +408, (3.10.18)14 8 Anmptopia O3- ¥8, Finally [Next (gjh) give 3# = ~ma ~ 205, p= —0° ‘the O(?) terms in (q) give Q° =m. Thus = (Aa) = [ma + 309]? +.0(0"), (3:10.19) mam [09 + 2d + HO? + O10"), (8:10.20) Qa m— 3H) + ¥))A + O(0"). (3.10.21) ‘The final group of equations is straightforward. The Bianchi identities (Be,g) reveal ¥, = Hin? — Gut)’ + 0(0%), (3.10.22) B= WI - (BUs)A +010"), (3.10.23) ‘and equation (FY) gives v= Gm) + [159+ ¥) + 9m] oF HOUR. (81020 ‘There are some mote useful equations relating the leading terms in the Weyl scalars. Since the are precisely the x, of sect 3.7 and the « here is the o there we can read some of them off ivectly from (8:7-14-16) as Hat, W=-08s, (8:10.25) Sf + 10949%9] = 0. (8:10.26) We also need one other result, obtained fom the leading order terms in equation (BO), vi. AWS = 395 4 094, (6.10.27) If higher order terms are required it is very easy to. gener ate them once the lower order terms are known, provided the equ For use in the next section we construct the a and m appropriate jons are handled in the order indicated S11 The Don fmamentin costr 185 ‘Then equation (310.10) reveals e=-7 oot, (6.0.29) and (8.10.18) then gives 3.11 ‘The Bondi 4-momentum covector In this, the penultimate section ofthe chapter, a 4-(co-)veetor P, will be defined with the following properties: ') P, is defined for each cut of % and if t* is a future-pointing timelike vector field at Ythen P, defines a scalar My = Pt, ) if i chosen as the timelike Killing veetor for the Kerr family of spacetimes then My, is the mass parameter, ©) My 2 0 provided the dominant energy condition holds, 4) if My(1) is defined for a Bondi system of cuts then AM, <0. ‘Thus the Bondi mass M, is to be interpreted as the mass of fan isolated system. It is positive or veto, identifiable with the ‘mass in stationary situations and decreases as the system radi- ates, Thus an isolated system can only radiate a finite amount of By. There is no possibilty of extracting an infinite amount of energy by allowing the “Newtonian potential energy” to decrease indefinitely. In this section only property a) is verified. ‘The re mainder are dealt with in section 3.12. Note that throughout this section a quantity in the physical space-time will not eatty a tilde since most of the calculations will be done there. Instead quanti- ties in the unphysical space-time will be denoted by 8 caret, e,, equation (31.5) We shall use the asymptotic null tetsad developed in section 8.10. Since the frst part of the analysis occurs in the unphysical space-time we need to be able to translate the results. It will be recalled from equation (2.5.8) that we are adopting the conformal ‘transformation formulae By haat, (na)156 3 Anymore “This implies for the frame components (3.10) that Q=97Q, 6 and that D = d/d0. To determine the NP scalars we need to use the formulae of section 8.4, obtaining cr, Braap, (11.2) O-(n-+ 34), a a, (ais) be noted that on fall of these a in the notation of where T= 9-'DN. Tt show scalars vanish except for ¢ = 4,00, 8 section 3.10. Finally (aaa) We now need a concept which has caused some confusion in the past, We would like a constant spinor, but as we saw in section 2.0 thie is only possible in type NV spacetimes. From the peeling theorem we know that an asymptotically simple spacetime is asymptotically of type MN, and so we should be able to construct an asymptotically constant spinor. (8.11.4) pernetion A spinor r* willbe said to be strongly asymptotically constant in Mf 8) nyo ry are regular oF i) Pe c5p <0 at where the conformal Schaviour of + ie defined by r* ‘The equation above is the twistor equation, and with the cated conformal behaviour i is conformally invariant. If imposed {generally it has non-trivial solutions only in type IV spacetimes, bat it can be imposed on certain subspaces, in particular on J, and in this context itis known as the asymptotic twistor equation. SIL The Bondi -monietion covetor 187 (8.11.2) venta Let 44 G04 = 208 b yh, Then z,y are regular on J and if their values there are denoted 2,98 then B= Ne" 40's! +0(0%), on"), where the coeficients are Q-independent and G-By* » F-ayy" on (ais) Proof: ro, = 743, = y is regular and so we may set y yf + Dy! + O(02). Similaely 41, = 9-547, = 0-2 is regular, and so we may set 3 = 22° + 0s! + 0(0). Projecting out the asymptotic equation into components on Snow gives y! = 0 and (3.118), ith any univalent spinor r we may associate @ real null vector tt 2749s symmetric two spinor © ates — Feo Pcaran) (3.11.6) and a real bivector (skew symmetric valence 2 tensor) F, want eapdap (az) We can now proceed to the construction of the Bondi 4 ‘momentum. Let © be # null hypersurface in M which extends toF, and let (0) be the 2-surface = const, in E. We now choose a null tetrad as in the previous section. 1" is tangent to the senerators of 5, n* is orthogonal to the (9) while m*, x span the tangent space to $(). Let un- frees ans158 3 Avymptapia where ‘The Bondi 4-momentum now appears as the Limit as — 0. (6.113) tmoxe Suppone that ris strongly asymptotically constant. Then I = T(G) sists ond defines the generalized Bondi 4-momen- Hata P* vie T= Pr, on % Proof Set shows that nb = lau + Spl (3.11.10) Now ua -(-a)yt on, p= (64 Be +H Thus Fata! = R[-26 — ay +.4(6+ 2 + p28+ wy] (8.11.41) Since dS = O(9") = 0(9-2) we have to show that the integrand (8.11.11) is (6) in order for 1(9) to converge as 1 — 0. Now x has spin ! while y has spin}. If 3 is defined for the unit “phere then the O(M) terms in (3.11.11) are found after making the relevant substitutions to be O(Ayterms = | [-as! = jg" — 2839" 45720! +00] (Oy 2) — WG" — BY! +e 49°" - 7° Gay" ~ jy") tee] 11 The Hand done vvctor 159 here the commutator of again. The O(92) terms are 3 has been used together with (3.11.8) O(@? terms = 3-2 =a! + ply'y! —253y! + pant B°35y! + (28)2"y" 4 5"(-Wd)e! + ec. (.t1.12) ‘The first and fourth terms cancel exch other out and the sixth term vanishes because of (3.11.5). Next note that the second and filth terms become, using (8.11.5) nate! sy'0e! = 2°39" +980! Now spin(z) + spin(g) = 1. Using exercise 3.9.5 these terms will cancel on integration. The seventh term is 2°99(93). Since 4 has spin 2, ¥ has spin 2. Thus using exercise 3.9.5, under integration this term delivers the same value as —30(y#35", of iy'3'g". Now y! has spin } and can be expanded in terms of spin-weighted spherical harmonics Y= Dewt¥, (6,8), where I ranges from 0 to 20 and m ranges from —(I-+ 1) to L However (311.5) implies 0)" = 0 oF O= Dea x ieslvant factor x 2¥;,1 (8. 9)s "his means that ¢,, = 0 for > 0, since the spherical harmonics are independent functions. We ean say nothing ebout the 1 «oeflcients because the corresponding spherical harmonies vanish Now x fnrelevant factor x _a¥),,1(0,8)s vanishes because (3.11.12) wil va ‘ight term abo 9 0 for ¢ > 1. Thus the seventh term in integration. By a similar argument the shes un integration. Ul that romaine is the160 3 Anymetapia third term and ite complex conjugate, which because of equation (8.10.28) contribute ete 4 oan ‘The last term here vanishes also on integration for precisely the reasons given sbove. Thus 10) =—3 fats + S, (3.1.13) exists and is finite It is clear that strong asymptotic constancy requires teak (ania) where the coeficients eg are that, stants, Now problem 3.9.7 showed evant constant factors agp a AT BC+ BE + OR ae THE where the constants A,C are real and B is complex. But com- parison with equation (8.8.7) shows that the freedom in yg” is precisely that of the translation subgroup of the BMS group. Note that the 4 vectors t* implied by independent choices of the constants in (811.15) ate all parallel to n* on Zand so are not independent as space-time vectors. However they are indepen dent vectors when regarded as elements of the Lie algebra of the BMS group. Choosing the constants in (8.11-14) to gener ate the element corresponding to an -translation generates the ‘e-component of P, via (8.11.13). In this sense (8.11.11) furnishes (13s) ty of the Bondi mass 3.12 The posi In this final section of the chapter, we show that the Boi is positive, ie., Pt? > 0 for all fature pointing timelike vectors #4. We need a preliminary lemma. With any univalent spinor 7 S12 The posivity of the Hoi mass 161 we have associated a real null vector #* = 7 spinor (3.12.1) eo dae + CanBuw e1za) The divergence of F,, will be needed several times, and so we compute it now. (8.12.1) vemnea (FMV atin) = (Var aN 9 = Je, )te0) (3.123) Proof: Unfortunately thie calculation is somewhat messy. For the sake of brevity only one half ofthe terms ace shown explicitly, although all are needed to sce some of the cancellations that occur, Clearly aw Jey OO bay bee. = AV Dg + 60 = HV FOALS atin (Fat aMVa\ Fo) +66) + guy 9 POV yn Vogr Vn, wt Fee,162 8 Ant =F Vo sV unt? +P eo gat © ey Duet +O pet at $F eyo ~ THB woF® ~ bryeoOF™— PV yaVaort +00 =~ pDapt® +2 Oporty— t00 = 2 ¢(-3A7,) 42D yotg tee 4b t+ 12M, 264, ‘uring Binstein’s lld equation In section 3.11 we constructed a null hypersurface B which ex: tended to, having afinely parametrized generators with tangent vector [*. E was foliated by two-surfaces (9) on which © was constant. On each 5(f2) we defined an integral 18) =f Fattm'as, where Fy, was constructed from a univalent spinor 7. If r was strongly asymptotically constant we saw that 1(0) furnished the Bondi 4-momentum at. We may propagate 7 from.J into D in ‘we show below this can be done 40 easing function of . We need fan arbitrary manner, and that 1() is a monotone non however a definition. (3.12.2) periwition An energy-momentum tensor T*! is said to obey the dominant energy condition if it obeys the weak energy condition ond in addition T's, i+ a future-pointing non-spacelike vector for all future-pointing timelike vectors v S12 The pointy ofthe Hon mass 168 Physically this means that in any orthonormal frame the energy density dominates other terms 7% > for all ab ‘This is satisfied forall known forms of matter; it were not then the speed of sound /(dp/da) could exceed 1 and eausality would be lost. It follows immediately from the definition that Tyw's! > 0, forall pointing. like v* and non-spacclike w*, both vectors being fature- (3.123) tuzonem Suppose that +4 is propagated on 3 by Dr! = 0%, (124) where w= ory, and the aN(@)/a0< 0. man ene cndionhls, Then Prof: Sapo we tone © by, = $(0) ad 5, = 0, sete 2 The verge ete ings 19,)-110,)= f wR, we Now from Iemma (3.12.1) PVR, = gt un —Iayre, where q = 0%6rg, and (8.124) has been used, Now by Binstein's field equation Gy = “8x67, and 30 ('V'R, > 0, which is sillcent to prove the asertion We next consider a compact spacelike hypersurface A bounded fry some $(M). This will have 4 timelike unit normal which we lenote by v*. u* defines a projection operator164 3 Anptopia which annihilates vectors parallel to v. We can use h to define » derivative D, intrinsic to A via D,= ho = V.— HY. (3.25) With these preliminaries we ean now proceed to the next result. (8.124) TmeonEM Let A be a compact spacelike hypersurface with unit timelike nor- ‘mal v*, and boundary $(9,)- Suppose we require the spinor 74 to satiafy the Witten equation De® . (226) in A. Then ifthe dominant energy condition holds 1(8,) 2 0 Proof: The divergence theorem implies that 10,)= f v0" Fd Now V, =D, +9,0°V,, Setting 4g = 0°V.t» we find Vyty = Dyrp + %7p = Yet if the Witten equation holds. Using lemma 3.1211 it is now easy to show that SUR = 16, and the result follows. Following Ludvigsen and Vickers, 1982 we can now prove the positivity of the Bondi mass on a cut S of J: On S we choose a strongly asymptotically constant spinor 7, It is clear from lemma 3.11.2 that this is equivalent to choosing a u-independent solution of dy! = 0, ie, y? has to have the form (3.11.14). Let 5 be the other null hypersurface through S whose generators are orthog- nal to.Z We propagate 4 from S$ to 3 via equation (3.12.4). ‘The surfaces = constant folate 3 and we choose a particular one 5, : =. Let A be any compact spacelike hypersurface 212 The postsity of the Vondi mase 165, with boundary 5,. We now propagate + from $, onto A via ‘the Witten equation (8.12.6). This is a fairly unusual boundary value problem for an elliptic system. However existence, (but not iqueness) is guaranteed by the corollary to theorem 10.6.1 of Hormander, 1976. Now theorem 3.12.4 asserts 1(9,) > 0 pro- vided the dominant energy condition holds, Theorem 3.12.3 now suarantees that 1(0) > 1(0,) > 0. [fone could assert not only ex- istence but also uniqueness of the boundary value problem (3.12.6) fone would have a local energy-momentum fax construction. finally remains to show that the Bondi mass is non-inereasing, Up to a positive factor we may construct My by taking y° = Lin (B.11.10), corresponding to a timelike translation, and 40 My j fey + sanas. (27) Thus fiace +sannas 4 fans +(anja) tones fort so¥4 + (annan +sanyas sng (2.1027), 4 foes sats+(anyany saris ing (81028, 4 flows + (aey(aayas, Now ¥$ has spin —1 and so its contribution to the integral will vanish because of exercise 3.0. Thus AM, =~} f Asazas <0. (3.128) (3.12.5) Problem. Consider « spacetime with coordinates (ur, 9, #) where ~00

You might also like

• Derivation of Riemannian Curvature

  

  No ratings yet

  Derivation of Riemannian Curvature

  8 pages
• Ryder L H Quantum Field Theory 2nd Edition PDF

  

  0% (1)

  Ryder L H Quantum Field Theory 2nd Edition PDF

  55 pages
• Second Text

  

  100% (1)

  Second Text

  181 pages
• The Kerr Spacetime-Rotating Black Holes in General Relatividad

  

  No ratings yet

  The Kerr Spacetime-Rotating Black Holes in General Relatividad

  190 pages
• Hicks Differential Geometry PDF

  

  No ratings yet

  Hicks Differential Geometry PDF

  95 pages
• Differential Geometry

  

  No ratings yet

  Differential Geometry

  88 pages
• Dreiner H.K., Haber H.E., Martin S.P. Supersymmetry (CUP Draf

  

  No ratings yet

  Dreiner H.K., Haber H.E., Martin S.P. Supersymmetry (CUP Draf

  272 pages
• ALL SPACETIMES BEYOND EINSTEIN by Frederic P. Schuller

  

  No ratings yet

  ALL SPACETIMES BEYOND EINSTEIN by Frederic P. Schuller

  44 pages
• Analysis On Manifold Via Laplacian Canzani

  

  No ratings yet

  Analysis On Manifold Via Laplacian Canzani

  114 pages
• Nakahara GTP Solutions

  

  75% (4)

  Nakahara GTP Solutions

  50 pages
• Tensors Manifolds and Relativity

  

  100% (1)

  Tensors Manifolds and Relativity

  106 pages
• Euler-Maclaurin Formula: 1 Bernoulli Numbers

  

  100% (2)

  Euler-Maclaurin Formula: 1 Bernoulli Numbers

  11 pages
• Tetrad Basis in General Relativity

  

  No ratings yet

  Tetrad Basis in General Relativity

  4 pages
• Octonionic Gravity, Grand-Unification and Modified Dispersion Relations

  

  100% (1)

  Octonionic Gravity, Grand-Unification and Modified Dispersion Relations

  23 pages
• Arithmetic and Geometry I PDF

  

  100% (2)

  Arithmetic and Geometry I PDF

  184 pages
• General Relativity 2012 Harvey Reall Lecture Notes

  

  No ratings yet

  General Relativity 2012 Harvey Reall Lecture Notes

  172 pages
• Genrel PDF

  

  No ratings yet

  Genrel PDF

  347 pages
• Pseudoparticle Solutions of The Yang-Mills Equations

  

  100% (1)

  Pseudoparticle Solutions of The Yang-Mills Equations

  3 pages
• 109 - UnivalentfunctionsandTeichmuellerSpaces PDF

  

  0% (1)

  109 - UnivalentfunctionsandTeichmuellerSpaces PDF

  266 pages
• William L. Burke - Applied Differential Geometry-Cambridge University Press (1985)

  

  No ratings yet

  William L. Burke - Applied Differential Geometry-Cambridge University Press (1985)

  216 pages
• Stewart - Advanced General Relativity

  

  No ratings yet

  Stewart - Advanced General Relativity

  238 pages
• Complex Geometry

  

  No ratings yet

  Complex Geometry

  146 pages
• General Relativity - Lecture Notes

  

  No ratings yet

  General Relativity - Lecture Notes

  3 pages
• Janis-Newman Algorithm: Generating Rotating and NUT Charged Black Holes

  

  No ratings yet

  Janis-Newman Algorithm: Generating Rotating and NUT Charged Black Holes

  68 pages
• Lagrange Equations For N Particles

  

  No ratings yet

  Lagrange Equations For N Particles

  32 pages
• The Elements of Integration and Lebesgue Measure-Bartle

  

  No ratings yet

  The Elements of Integration and Lebesgue Measure-Bartle

  192 pages
• 5059.minimal Surfaces (Courant Lecture Notes in Mathematics) by Tobias H. Colding

  

  No ratings yet

  5059.minimal Surfaces (Courant Lecture Notes in Mathematics) by Tobias H. Colding

  128 pages
• Gauge Theory and Variational Principles: David Bleccker

  

  No ratings yet

  Gauge Theory and Variational Principles: David Bleccker

  10 pages
• Weinberg S Gravitation and Cosmology Principles and Applications of The General Theory of Relativity (Wiley, 1972) (Isbn 0471925675) (685S)

  

  No ratings yet

  Weinberg S Gravitation and Cosmology Principles and Applications of The General Theory of Relativity (Wiley, 1972) (Isbn 0471925675) (685S)

  685 pages
• Riemann - Metric Taylor Expansion

  

  No ratings yet

  Riemann - Metric Taylor Expansion

  30 pages
• Clifford Algebra and The Projective Model of Hyperbolic Spaces PDF

  

  100% (1)

  Clifford Algebra and The Projective Model of Hyperbolic Spaces PDF

  11 pages
• W. Galleas - The Bethe Ansatz Equations For Reflecting Magnons

  

  No ratings yet

  W. Galleas - The Bethe Ansatz Equations For Reflecting Magnons

  18 pages
• Differentiable Manifolds and Tensors

  

  No ratings yet

  Differentiable Manifolds and Tensors

  55 pages
• Munkres J Analysis On Manifolds

  

  No ratings yet

  Munkres J Analysis On Manifolds

  190 pages
• De Rham Cohomology and A Variational Principle: Why Electromagnetism Is Mathematically Trivial

  

  No ratings yet

  De Rham Cohomology and A Variational Principle: Why Electromagnetism Is Mathematically Trivial

  8 pages
• Differential Geometry by Andrews PDF

  

  No ratings yet

  Differential Geometry by Andrews PDF

  176 pages
• Vector Elds

  

  No ratings yet

  Vector Elds

  18 pages
• X Global Lorentzian Geometry Beem

  

  No ratings yet

  X Global Lorentzian Geometry Beem

  331 pages
• Milnor Thom

  

  No ratings yet

  Milnor Thom

  18 pages
• Applications of Differential Geometry To Physics: Cambridge Part III Maths

  

  100% (1)

  Applications of Differential Geometry To Physics: Cambridge Part III Maths

  29 pages
• TRISTAN NEEDHAM The Geometry of Harmonic Functions

  

  No ratings yet

  TRISTAN NEEDHAM The Geometry of Harmonic Functions

  17 pages
• Dommelen Quantum Mechanics For Engineers

  

  No ratings yet

  Dommelen Quantum Mechanics For Engineers

  1,623 pages
• Coupled 1st Order ODEs

  

  No ratings yet

  Coupled 1st Order ODEs

  28 pages
• Moving Frame Method

  

  No ratings yet

  Moving Frame Method

  61 pages
• Basic Category Theory - Catsmoeder

  

  No ratings yet

  Basic Category Theory - Catsmoeder

  88 pages
• Geometric Calculus

  

  No ratings yet

  Geometric Calculus

  7 pages
• Spectral Theory in Riemannian Geometry

  

  No ratings yet

  Spectral Theory in Riemannian Geometry

  197 pages
• Selected Topics in Malliavin Calculus, Springer

  

  No ratings yet

  Selected Topics in Malliavin Calculus, Springer

  178 pages
• Solution Mnaual of Topic in Topology

  

  No ratings yet

  Solution Mnaual of Topic in Topology

  19 pages
• Maciejko Representation of Lorentz and Poincare Groups

  

  No ratings yet

  Maciejko Representation of Lorentz and Poincare Groups

  11 pages
• MATHEMATICS of Gerenal Theory of Relativity

  

  No ratings yet

  MATHEMATICS of Gerenal Theory of Relativity

  87 pages
• Grassmann PDF

  

  No ratings yet

  Grassmann PDF

  207 pages
• Geometric Harmonic Analysis IV - Dorina Mitrea & Irina Mitrea & Marius Mitrea

  

  No ratings yet

  Geometric Harmonic Analysis IV - Dorina Mitrea & Irina Mitrea & Marius Mitrea

  1,004 pages
• An Introduction ToTensor Calculus and Relativity - Derek F. Lawden

  

  100% (1)

  An Introduction ToTensor Calculus and Relativity - Derek F. Lawden

  199 pages
• Position and Motion of Celestial Bodies Toshio Fukushima Z Library

  

  100% (1)

  Position and Motion of Celestial Bodies Toshio Fukushima Z Library

  334 pages
• Symmetries Integrability Diff Eqs Peter Olver

  

  No ratings yet

  Symmetries Integrability Diff Eqs Peter Olver

  360 pages
• Lecture Notes On Smooth Manifolds

  

  100% (2)

  Lecture Notes On Smooth Manifolds

  142 pages
• Lecture Notes GR2

  

  No ratings yet

  Lecture Notes GR2

  126 pages
• DG - Chap1 Manifold

  

  No ratings yet

  DG - Chap1 Manifold

  7 pages