This action might not be possible to undo. Are you sure you want to continue?

**Lecture notes Poul Olesen and Jan Ambj rn
**

AND

University of Copenhagen The Niels Bohr Institute Autumn 2000

Preface

The following lecture notes on general relativity, cosmology and classical gauge theory were written in connection with a one year course on eld theory. The notes are introductory, and do not in any way pretend to be comprehensive. Several important topics have been left out. For example, gravitational radiation is not discussed at all. There are two reasons for the brevity of the notes: the alloted time was short, and it was hoped that by making the notes equally short, there is a bigger chance of getting through that general relativity and cosmology are exciting subjects. Sometimes the trouble with exposing the beauty of physics is that one has to walk a very long way so people start to feel that they rather walk in a desert than in a beautiful garden. By cutting down on the exposition of the basic formalism in a very geometric fashion we hope that the spirit of Einstein's theory nevertheless appears. A chapter on classical gauge theories is included. It seemed natural to do so since many of the geometrical concepts are the same as in general relativity. In addition modern cosmology has been closely linked to the study of gauge-Higgs theories since the topological objects like domain walls, cosmic strings and monopoles encountered in gauge-Higgs theories play an important role in the early universe. We thank Kaj Roland for having made most of the problems in section 5.

P. Olesen and J. Ambj rn

Contents

1 General relativity

1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 1.10 1.11 1.12 1.13 1.14 1.15 1.16 1.17 1.18 1.19 1.20 1.21 1.22 1.23 1.24 1.25 1.26 1.27 1.28 2.1 2.2 2.3 2.4 2.5 2.6 The principle of equivalence : : : : : : : : : : : : : : : : : : : : : : : : : : Gravitation and geometry : : : : : : : : : : : : : : : : : : : : : : : : : : : Motion in an arbitrary gravitational eld : : : : : : : : : : : : : : : : : : : The Newton limit : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : The principle of general covariance : : : : : : : : : : : : : : : : : : : : : : Contravariant and covariant tensors : : : : : : : : : : : : : : : : : : : : : : Di erentiation : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : A property of the determinant of g : : : : : : : : : : : : : : : : : : : : : Some special derivatives : : : : : : : : : : : : : : : : : : : : : : : : : : : : Some applications to physics : : : : : : : : : : : : : : : : : : : : : : : : : : Curvature : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : Parallel transport and curvature : : : : : : : : : : : : : : : : : : : : : : : : Properties of the curvature tensor : : : : : : : : : : : : : : : : : : : : : : : The energy-momentum tensor : : : : : : : : : : : : : : : : : : : : : : : : : Einstein's eld equations for gravitation : : : : : : : : : : : : : : : : : : : The time-dependent spherically symmetric metric : : : : : : : : : : : : : : A digression: A simpler method for computing ; : : : : : : : : : : : : : The Christo el symbols for the time-dependent spherically symmetry metric The Ricci tensor : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : The Schwarzschild solution : : : : : : : : : : : : : : : : : : : : : : : : : : : Birkho 's theorem : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : The general relativistic Kepler problem : : : : : : : : : : : : : : : : : : : : De ection of light by a massive body : : : : : : : : : : : : : : : : : : : : : Black holes : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : The spherically symmetric solution in 2+1 dimensions : : : : : : : : : : : : Cosmic strings : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : Kruskal coordinates : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : Appendix: Some remarks on geometry : : : : : : : : : : : : : : : : : : : : The cosmological problem : : : : : : : : : : : : : : : : : : The cosmological standard model : : : : : : : : : : : : : : A geometric interpretation of the Robertson-Walker metric Hubble's law : : : : : : : : : : : : : : : : : : : : : : : : : : Einstein's equations and the Robertson-Walker metric : : : The big bang : : : : : : : : : : : : : : : : : : : : : : : : : 3 3 4 5 7 9 10 11 13 15 16 19 20 22 23 26 29 30 32 33 34 35 36 39 41 43 47 49 50

3

2 Cosmology

: : : : : :

: : : : : :

: : : : : :

: : : : : :

: : : : : :

: : : : : :

: : : : : :

: : : : : :

: : : : : :

53

53 54 59 60 62 64

3 Discussion of the fate of the universe : : : : : : : : : : : : : : : : : Fitting parameters to observations : : : : : : : : : : : : : : : : : : : : : : The cosmic microwave radiation background : : : : : : : : : : : : : : : : : The matter dominated era : : : : : : : : : : : : : : : : : : : : : : : : : : : 2.13. Doppler peaks and all that : : : : : : : : : 2.9.6.1 Motivation for in ation : : : : : : : : : : : : : : : 4.2 The in ationary scenario : : : : : : : : : : : : : : 4.9.3 Monopoles : : : : : : : : : : : : : : : : : : : : : : : 3.6.3 Successes and problems of the in ationary model : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 89 4 Applications : : : : 131 131 131 135 137 5 Problems and constants 139 .2 Generalization to eld theory : : : : : : : : : : : : 3.6 Spontaneous symmetry breaking and the Higgs mechanism 3.13.4 Geometrical interpretation : : : : : : : : : : : : : : : : : : 3.3 The open universe 0 < 1 : : : : : : : : : : : : : : : : : : : : : : : 2.1 Existence of the big bang (the initial singularity) : : : : : : : : : : 2.4 An upper bound on 0 : : : : : : : : : : : : : : : : : : : : : : : : : 2.9.8 Solutions of Yang-Mills equations : : : : : : : : : : : : : : 3.8 2.11 2.1.9 2.1 In ationary Cosmology : : : : : : : : : : : : : : : : : : : 4.12 2.1 The closed universe 0 > 1 : : : : : : : : : : : : : : : : : : : : : : : 2.5 Generalization to non-abelian groups : : : : : : : : : : : : 3.2 The age of the universe according to big bang : : : : : : : : : : : : 2.1.3 Abelian gauge theories : : : : : : : : : : : : : : : : : : : : 3.6.1 Classical mechanics : : : : : : : : : : : : : : : : : : : : : : 3.1.5 Doppler peaks : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 1 64 65 65 66 68 70 70 71 71 72 73 73 75 77 79 80 83 84 84 86 89 89 92 96 97 99 100 104 109 110 112 117 121 125 2.13.3 The peaks : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 2.10 2.2 Taking into account relativistic e ects : : : : : : : : : : : : : : : : 2.9.5 Discussion of the life time of the universe : : : : : : : : : : : : : : : Causality structure of the big bang : : : : : : : : : : : : : : : : : : : : : : In ation : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : Evidence for the cosmological constant from observations of distant supernovae : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : Appendix: Density uctuations.2 Vortices and strings : : : : : : : : : : : : : : : : : : 3.CONTENTS 2.1.1 The classical Jeans analysis : : : : : : : : : : : : : : : : : : : : : : 2.7 2.4 Primary uctuations : : : : : : : : : : : : : : : : : : : : : : : : : : 2.13.7.2 The at universe 0 = 1 : : : : : : : : : : : : : : : : : : : : : : : : 2.1 The action principle : : : : : : : : : : : : : : : : : 3.9 Instantons : : : : : : : : : : : : : : : : : : : : : : : : : : : 4.13.1.7.2 Internal global and local symmetries : : : : : : : : : : : : 3.7.13 3 Gauge theories 3.1 Domain walls : : : : : : : : : : : : : : : : : : : : : 3.7 Solitons : : : : : : : : : : : : : : : : : : : : : : : : : : : : 3.9.

2 CONTENTS .

We have the general equation of motion ~ F = mi~ a (1. because eq.4) ~ = ~ . Einstein's starting point was the principle of equivalence. Newton made experiments where the period of oscillation of a pendulum made up from di erent materials were studied. i. 2 ~ t2 y x 1g we get ~=0 y (1. which can be understood in the context of Newton's mechanics. With mi = mg = m one has the equation of motion ~ =~ x g (1. (1.1) where ~ is the acceleration and mi is the inertial mass { for a given force the acceleration a is smaller the larger the mass is..1 The principle of equivalence Einstein's general theory of relativity is a beautiful piece of art which connects gravitational elds with geometry of space and time and thus provides a scheme in which our universe can be discussed. In a constant gravitational eld the force is given by ~ Fg = mg ~ g (1. the body is more inert the larger mi is.e. Let us consider a constant gravitational eld. and this was accepted to such an extent that most text books today (and at Einstein's time) do not bother to put any indices on the masses.5) Therefore we conclude that an observer living in the y-system sees no e ect of the gravitational eld. eq. if we introduce the coordinate (1.4) shows that the observer is freely falling ( 1 ~ t2 is just 2g 3 . On the other hand. It is clear that a priori the parameter mg is not related to g the inertial mass. Later on many very precise experiments were made which showed that mi = mg to a high accuracy.3) Thus.2) where ~ and mg are constants. (1.Chapter 1 General relativity 1.5) shows that particles move in straight lines as if there was no force. and he found no variation with mi=mg .

we need. In special relativity an event is described by a four vector y = (y0 ~). d 2 = (dy0)2(1 .g. if we make a Lorentz transformation from y to y0 then d 2 = (dy0)2 . All this is true irrespective of any mass because mi = mg . Thus in each of the (in nitely many) freely falling systems of inertia we can apply special relativity with no forces acting on the particle. The proper time has the following physical interpretation: Let us consider a clock (or any physical system which speci es a time. which by de nition marks time by small intervals dt when the clock is at rest. 1.6) i and hence the primed system would depend on which material we consider through the ratio mg =mi . v y Thus.8) says that the speed of light jd~=dtj is equal to one in y all systems (Michelson-Morley's experiment).e.. One can use this statement to obtain some insight into the way in which gravity in uences other physical phenomena by writing down in each of the small elevators some law of physics and then transform it to a general coordinate system. In Newtonian mechanics we therefore know that an observer in a su ciently small freely falling elevator is unable to detect a gravitational eld. (d~)2 = (dy00 )2 . The proper time y d 2 = (dy0)2 . where y0 is the y time.7) is an invariant.2 Gravitation and geometry Let us consider a particle which moves under the in uence of a gravitational eld only. Einstein's principle of equivalence generalizes this to any physical phenomena: In any arbitrary gravita- tional eld it is possible at each space-time point to select locally inertial systems (freely falling small elevators) such that the laws of physics in these are the same as in special relativity. and is easily seen to eliminate the gravitational eld also if other forces (e. (d~ 0 )2 y y (1. When mg = mi the transformation (1.g. namely a particle which is freely falling in an arbitrary gravitational eld. ~ 2) = (dy0)2 v (1. ~ 02) v .4) would have to be replaced by ~ = ~ . If the gravitational eld varies in space we can apply the transformation (1.4) is universal. 2 mg ~ t2 y x 1 m g (1. If mi 6= mg . e. a particle which decays with a certain life time). i.4) in a su ciently small domain. in the rest system. (d~)2 y (1. Thus d = dy In a moving system d 2 = (dy00 )2 (1 .4 GENERAL RELATIVITY the displacement pertinent to a free fall). electrostatic forces) are at work. Since the elevators are a priori small. however. (1.9) 0 = the interval between two ticks on the clock (at rest). to consider an in netesimal four vector dy = (dy0 d~). the coordinate transformation (1. In the next section we shall consider the simplest example.8) For light d = 0. and eq. In the rest system the velocity ~ = d~=dy0 vanishes.

10) means 3 3 X X d 2=. eq. d 2 then means the distance between two points computed by the law of Pythagoras.4). If we denote the space-time coordinates in an arbitrary coordinate system by x .15) d2 Using that the y 's depend on x we have ! ! d dy = d @y dx ( ) 0 = d d d @x d 2x @2 @y (1.10) is valid in any of the freely falling elevators. (1. dy dy (1.10) has an almost obvious geometric interpretation: In a curved space (e. in spite of the fact that this geometry is not valid in general in curved space.10) where = 0 for 6= and 11 = 22 = 33 = 1 00 = . the surface of a sphere) one can introduce local coordinate systems where euclidean geometry is valid. and we have the equation of motion d2y (x) = 0 (1. we should notice that eq.11) =0 =0 To obtain the e ect of gravity. Then eq. (1.10) if we replace by . The local euclidean geometry corresponds to eq.12) For an example. v02 . @x @x dx dx . (1. see the Newtonian case (1.13) d 2 = . where is the Kronecker symbol ( = 0 for 6= = +1 for = ).1. then the y's are functions of the x's. (1. y = y (x): (1.3 Motion in an arbitrary gravitational eld Now let us again consider a particle which moves under the in uence of an arbitrary gravitational eld. (1. In four vector notation we write the proper time as p 5 d 2 = . In the freely falling system special relativity applies. However.13) is the same distance written in arbitrary coordinates. (1. in general the elevators are di erent in di erent space-time points. dy dy (1.g (x)dx dx where @y (1.g.10) we use the summation convention: Whenever an index occurs two times in a product it is to be summed. Eq. (1. Inserting this in eq.10) we get @y @y (1. In eq.3 MOTION IN GRAVITATIONAL FIELD which leads to the formula for the time dilatation dy00 = d = 1 . 1.1.16) = @x dd 2 + @x y dx dx @x d d . Thus.14) g (x) = @x @y = g (x) @x is called the metric tensor.

multiplying eq.19) we have @g = @x a @ 2 y @y + @x @x @x @y @ 2 y @x @x @x (1.14). +g . dx dx = 0 d2 d d where @ 2 y @x . We shall now show that the Christo el symbol can be expressed in terms of the metric tensor. g (x)g (x) = (1. have been expressed in terms of the functional relation between the local freely falling elevators and the arbitrary system x .19) is called the Christo el symbol or the a ne connection (sometimes denoted f g).18) (1. (1.23) where we used the expression (1. (1. Using eq.16) by @x =@y and summing over we get d2x + . @g = 2g . (1.23) we get @g + @g . (1. So far the metric and .14) g depends only on the rst derivative.20) we get @g = @y .14) for the metric.20) ! @y = @ 2 y @x @y = @ 2 y . whereas from eq. (1.21) (1. (1.24) @x @x @x If one wishes g can be thought of as a 4 4 matrix. (1. @y @x + . One can then consider the inverse. (1.25) . @x @x @x @y @x @x @x where we used the chain rule for di erentiation @x @y = @y @x Thus we can express the second derivative of y in terms of . which we denote g .22) (1.21).6 GENERAL RELATIVITY This looks like an equation of motion (because of d2x =d 2) with a force.19) we see that . depends on the second derivatives of y . @y @y @x @x @x @x = g . Therefore let us di erentiate g by use of the de nition (1. From eq. Using this in eq. From (1. = @x @x @y (1. We can remove the factor multiplying the second derivative of x by the following trick: By the rules of di erentiation we have ( = +1 for = = 0 for 6= ) @x @y = @y @x Thus.17) (1. by means of eq.

0 ' 0 since h00 does not depend on time.30) g (x) = + h (x) jh j 1 where h is the correction to the constant metric tensor .31) 2 @x so that .27) 2 @x @x @x g = 1.4 THE NEWTON LIMIT The inverse exists: 7 @x @x (1. namely the Newton limit where all velocities are small relative to the velocity of light. (1. @g . g is time-independent. 00 (1. (1. jd~ =d j << 1. Later we shall see that the second derivative of g are determined through Einstein's eld equations in terms of matter distributions.28) Using eq. and where the x problem is static. (1. (1. (1. At present we shall study a much more modest problem.27) determine the motion of a particle in a gravitational eld provided we know how g (x) depend on the gravitational eld.26) @y @y because the transformations y ! x and x ! y are non-singular coordinate transformations. eq.24) now gives the following relation between .18) and (1. dt = constant d For = i i = 1 2 3. (1. and g.e.29) (1. 1 dt 2 rh (~ ) ' 0 x ~ 00 x d2 2 d (1.32) . Then eq.31) ! d2~ .4 The Newton limit Eqs.00 ' .29) gives .28) becomes by use of eq. 1 @h00 (1. " # 1 g @g + @g . (1. dt 00 d d2 !2 '0 (1.18) then gives (t = x0 ) d2x + .e.00 ' . 1 g @g00 2 @x Since we are interested in small e ects of gravity we write (1.27) and the fact that all time derivatives vanish we have . Eq. Hence the = 0 component of eq. = (1.28) becomes simply d2t ' 0 d2 i.1. i. To the lowest non-trivial approximation eq.

g00 (x) This is not in itself an observable e ect since all clocks in the point will su er the same e ect.37) can be observed.38) falling = .33) and (1.40) g00(x1 ) 1 In the weak eld approximation (1. Recalling that the proper time is the time observed on a freely falling watch.41) . If we consider light which passes from the sun to the earth we have for the sun's potential = .34) we get h00 = constant . 00 dt2 x 2 (1.34) dt2 where (~ ) is the gravitational potential.39) .g (x)dx dx = . GM r Comparing eqs. and requiring that at very large distances from the point mass.35) where G i Newton's constant (G = G=c2 = 7:41 10.36) = .42) R 1 = 2. Thus the time measured in this system in the point x is dt = q d (1. If the clock is a physical system with a frequency .8 and because dt is proportional to d GENERAL RELATIVITY d2~ = 1 rh (~ ) x ~ (1.r (~ ) x ~ x (1. (1. GM = . we can compare two di erent points.33) dt2 2 00 x This equation can immediately be compared to Newton's equation d2~ = .37) Thus we see that su ciently close to a point mass space-time must indeed be slightly curved! The \curvature" indicated by eq. we have d 2 = . (x1 ) (1. For a point mass one has the well known result (1. The Newtonian potential is determined by x Poisson's equation r2 (~ ) = 4 G (~ ) x x (1.g00 (~ )dt where the last expression is valid in a gravitational eld where the clock is (approximately) at rest.(1 + 2 (~ )) x x (1.2 . we get ( = 2 =dt) v u 2 u g00 (x2 ) =t (1. 1 1 = (x2) . However.6 (1.37) we then have which is an observable e ect.29 cm/g). space should be at we get g00(~ ) = . 2:2 10. (1.

and the e ects of gravity can then be obtained by transforming to an arbitrary system. bottom = (980cm=sec )(2260cm) (3 1010cm=sec)2 = 2:46 10. The frequency of light from the sun is thus shifted by 2:2 parts per million relative to light from earthbound sources. namely the principle of general covariance: The general laws of nature are to be expressed by equations which hold good for all systems of coordinates. (1.5 GENERAL COVARIANCE 9 whereas the earth's potential can be ignored relative to . (1. are covariant (i. It then follows that if general covariance is satis ed then the equivalence principle is also satis ed: in each point there are freely falling local elevators in which the laws of nature are those of special relativity. and from general covariance they are thus valid laws in all coordinate systems.42) G should be replaced by G=c2 in ordinary units. (1.43) The experimental value is (2:57 0:26) 10. Einstein introduced (1916) a new principle which leads to a much more systematic way of obtaining the physics of gravity from the physics without gravity. preserve their form) with respect to any substitutions whatever (\generally covariant").5 The principle of general covariance So far we have studied the e ects of gravity by use of the principle of equivalence according to which the physics of special relativity is valid in freely falling local systems of inertia. it should be emphasized that second derivatives of g (x) are not in general assumed to vanish. Such a procedure is in general rather complicated. The equation of motion in a gravitational eld is given by equation (1.15 (1. Another experiment made 1960 consists in emitting light from a tower of height 22.15 in excellent agreement with the prediction.e.45) @x x=x0 = 0 Because of the connection between the Cristo el symbol . We need to specify precisely what we mean by \local elevators". because then from experience with tidal forces we know that gravitational e ects can be observed on a su ciently large scale. At each point x = x0 we can select the elevator such that g (x0) = (= at Minkowski space). and g given by eq. This amounts to saying that we only assume that .41) the frequency shift should be 2 = top .44) 1.6m.1.27) this ensures that to lowest order near each point of space-time we have no e ect from gravitational forces. In eq. and in order to have local straight-line motion we can select the elevator such that @g (x) (1. Taking into account various other e ects the best experimental result is 1:05 0:05 times the predicted value. From (1. However.18). In this connection the \laws of nature" are to be understood as those which are valid in special relativity. The falling light is then observed on the ground. that is. Clearly they should not be too large.

The transformation law (1.e.10 GENERAL RELATIVITY e ects of gravity can be transformed away on a scale. T 1 n 1 m .47) U 0 = @x U @x It should be noticed that since d is invariant.g. The proper time is an example of an invariant. which are suitable for applying the principle of general covariance. In general one can also have a scalar eld (x) de ned in each space-time point. Another quantity is the coordinate di erential dx which transforms like 0 (1. The tensor transformations are linear and homogeneous for the components of the tensors.50) @x0 @x @x0 A general tensor can have arbitrarily many indices. The simplest quantity is a scalar quantity. In other .51) can easily be remembered by noticing that T transforms the same way as if it had been a product of n contravariant and m covariant vectors. which is small relative to the scale of the gravitational eld. Its transformation is given by 01 0 n @x 1 @x m T 0 1 ::: n 1::: m = @x 1 @x n @x0 1 @x0 m T 1::: n 1 ::: m (1.48) From (1. U is a contravariant vector if under a transformation x ! x0 one has U ! U 0 with 0 (1.46) dx0 = @x dx @x Any quantity which transform like (1. 1. For tensors there exist transformations when the coordinates are transformed (x ! x0 ).46) is called a contravariant vector.6 Contravariant and covariant tensors We shall now present a systematic construction of certain quantities. @ 0 (x0) = @ (x) @x (1. it follows from (1. if it is valid in one system. Tensor laws thus follow the principle of general covariance if we ensure that they are valid in special relativity.48) we can form a scalar 0 @x A0 U 0 = @x0 @x A U = A U (1. i.51) @x @x A tensor with upstairs as well as downstairs indices is called a mixed tensor. e. which is invariant under x ! x0. tensors..49) @x We say that by contracting the indices of two vectors we obtain an invariant. A law of nature requiring that all components of a tensor vanish is thus valid in all systems.46) that the four-velocity dx =d = (dx0=d d~ =d ) is a contravariant vector. From a scalar (x) we can form a covariant vector by di erentiation. Hence all components in the x0 -system vanish if they vanish in the x-system.47) and (1. x A covariant vector is de ned by @x A0 = @x0 A (1.

j@x=@x0 j = j det(@x =@x0 )j. A B =T (1. The index of a tensor can be \raised" or \lowered" by means of g or g . e.g.53) @x0 where j@x=@x0 j is the Jacobi determinant.7 DIFFERENTIATION 11 words. denotes the matrix g . Finally let us mention a few simple rules for tensors which follow from the transformation (1. This is related to the fact that di erentiation is de ned by comparing the tensor in two di erent points (and then taking the limit where the points approach one another). from the transformation law (1.54) where in the rst step we used the usual Jacobi transformation. . one obtains @x 2 g (1. g .50)).56) g0 = 1. a product of vectors is a tensor. The symbol is easily seen to be a mixed tensor. @y @y g0 (x0 ) = @x0 @x0 @y @y @x @x = @x @x @x0 @x0 = @x0 @x0 g (x) (1. However. det g @x @x From the transformation law (1. T = g g T .14). which in matrix form reads (g0) = ( @x0 )(g)( @x0 ). Thus the measure of integration d4x is not invariant. in general it is not true that di erentiation of a tensor leads to a new tensor. A B C transforms as a tensor D .52) @x @x Also. g is a contravariant tensor.52).7 Di erentiation We saw in the last section that di erentiation of a scalar leads to a vector (see eq. E.g.53) has the important consequence that (d4x = dx0 dx1dx2 dx3 ) q 4 0 q @x0 4 p 4 g0 d x = g0 @x d x = g d x = invariant (1.51): a sum of two tensors is a tensor. etc.g. where (g) etc. as can be seen e. e. Often the determinant of g occurs. and T is denoted T .55) and a contraction in a tensor is a new tensor with fewer indices. a tensor transforms di erently in the two points.g. However. Eq. ~ T =T (1. The metric tensor g is a covariant tensor. but should be multiplied by pg.g. etc. as is easily seen from the de nition (1. e. a product of two tensors is a new tensor (with more indices). (1.1. By going to a freely falling system we see that the tensors T and T are the same physical object. (1.51).

58) we have that @V V = @x . (1.58) V = @x is a covariant vector. V (1. . and since the left hand side of (1. An important property of the covariant derivative is that it reduces to the ordinary derivative in a freely falling local elevator. This follows from the de nition of locality given . C + . This statement has already been checked directly in eq. .60) d @x @x @x d d Since dx dx d d is a contravariant tensor. @ dx dx invariant = 2 (1. By choosing all sorts of such elds we can.62) @x is a mixed tensor (T ). If we di erentiate a contravariant tensor V the following quantity V = @V + . C .63) @x We leave it to the diligent reader to verify this statement.g. However. (1.57) once more. it follows that the quantity is bracket in eq. . The concept of covariant di erentiation can be generalized to an arbitrary tensor. Alternatively one can use the connection (1. d2 = @ 2 dx dx + @ d2x invariant = d 2 @x @x d d (1. however.61) was derived for a trajectory of a particle falling freely in some gravitational eld.59) @x d 2 Next let us take the trajectory x ( ) to be a path for a particle which falls freely in an arbitrary gravitational eld.61) is a covariant tensor with indices and . to show directly that (1. Hence eq. V (1. and consequently it follows that @ (1. C = @C + . C (1. Then we have from the equation of motion (1. One has e. Now let us di erentiate (1. Inserting (1. V is called the covariant derivative of V .50) .57) d @x d where the invariance follows from the fact that d and d are individually invariant. Starting from a scalar (x) we have invariant = d = @ dx (1. as well as the fact that the metric g is a tensor.27) between . and the metric.18) ! d2 = @ 2 .61) should be valid for x being an arbitrary point in the space-time continuum. (1.60) is an invariant. dx =d is a vector. manage that the curve is completely arbitrary.12 GENERAL RELATIVITY To see what happens let us consider some trajectory x ( ). The result (1.61) is a tensor.60) is a covariant tensor.

27) one can easily see that . vanishes. and these laws.64) that in any point there exists a freely falling elevator where .64) From this we have the very important conclusion that in spite of the notation . because if it was a tensor then it follows from (1. V dx DV (1. i. and from the fact (see eq. (1.65) @x @x @x The rst term on the right hand side is what one would get if . transforms as 0 @x @x 0 @2 . of course. transforms in a more complicated way. and all the terms in (1. This is rather satisfactory since we are supposed to use the laws of special relativity in each elevator. is a tensor. Eq. (x) jx=xo = 0 (1. (mixed tensor notation) it is not true that . 1.27) it follows that . from the fact that the rst derivatives of g can be taken to vanish in a local elevator. In the sum (1. We can then project the covariant derivative to the tangent dx =d .66) Finally let us consider motion along a trajectory x ( ).45) of a local elevator. Since . Since . and consequently it should vanish in all coordinate systems. In this case one can only talk about di erentiation along the curve x ( ). In a local elevator it reduces to the ordinary derivative dV =d .1. Because of (1. From the de nition (1. contain only the ordinary derivatives.68) containing the particular factor ars can then be written arsArs (no summation over r and s). Ars is called the .45) according to which the rst derivatives of the metric tensor can be required to vanish in a local system. It is de ned by det a = X (.64) it follows that in the small elevators the covariant derivative is just the ordinary derivative. To this end we need to di erentiate a determinant. (1. g =0 (1. but the second term shows that .62) is then replaced by dV + .0 = @x @x0 @x0 . is directly related to these derivatives through eq. does not vanish in general. (1. was a tensor.1)P ( 1 2 ::: n ) a 1 1 a2 2 : : : an n (1. it follows that it is not a tensor.67) D d d which is easily shown to be a contravariant vector with index .8 THE DETERMINANT OF g 13 in (1. 1 factors. it follows that the covariant derivative of g vanishes.52) ) that g is a tensor.e.68) where P is even (odd) for even (odd) permutations of the reference sequence 1 2 : : : n. Using the transformation law for the tensor g and the relation (1. Let us therefore start by considering an n n determinant with elements ars. where Ars is a sum of terms which consist of n . + @x @x0 x 0 (1.8 A property of the determinant of g We shall show an interesting property related to the determinant of g .68) we can take an element ars outside a bracket.

and the variation is a product of n .69) The n complements Ar1 : : : Arn are all independent of the elements ar1 : : : arn in the r'th row.14 GENERAL RELATIVITY complement.72) B A @ Ann A1n A2n det a det a : : : det a The correctness of (1.68).70) can be used to construct the inverse determinant or matrix.25) we have g g = 4.70). (1.1. From this general result let us return go g . Using (1.69) and (1.e.68) has only one element from the r'th row and the s'th column). With g = .72) together by use of (1.74) has the alternative form dg = . However. according to the usual rule for di erentiation of a product. det g we have (g is symmetric) dg = g g dg (1. Now each of the n! products in det a contains one and only one element in the r'th row and we have (no sum over r) ar1 Ar1 + ar2 Ar2 + + arnArn = det a (1. consider the variation det a arising from (1. g dg + g dg = 0 so (1.76) .1)sr stands for the elements in the inverse matrix a.1 )sr det a ars (1. and each term in the sum (1.72) is seen by multiplying (1. 1 factors multiplied by ars.71) @ A : C an1 an2 : : : ann is 0 A11 A21 : : : An1 1 A A22 An B det12a det a : : : det 2a C B det a det a det a C C (1.68) of the determinant. Thus. The result is the unit matrix. collecting all terms which are multiplied by ars amounts to collecting the terms in the complement Ars. since two rows have the same elements.e.69) and (1. the new determinant has the value 0. According to (1.68) has n factors. i. 1 factors do not contain ars. i.70) Eqs. as1Ar1 + as2 Ar2 + + asnArn = 0 (s 6= r) (1. Each term in the sum (1. In particular we can keep Ar1 : : : Arn unchanged if we replace the elements in the r'th row with the corresponding elements as1 as2 : : : asn in the s'th row (r 6= s). Thus the inverse of 0 a a ::: a 1 B a11 a12 : : : a1n C B 21 22 B : : : : : : : : : : :2n C (1.75) (1. Because of the construction (1. Ars is independent of the elements in the r'th row and the s'th column (we have already picked the element ars. i. Thus the quantities Ar1 : : : Arn are unchanged if in det a we interchange the elements in the r'th row with other elements. Now let us di erentiate det a.73) where (a. The n .74) since g was de ned as the inverse of g .72) this means that X X det a = Ars ars = (a.71) and (1.g g dg (1.e.

1.9 SOME SPECIAL DERIVATIVES From (1.74) we have

15 (1.77)

@g ln g @g = 1 @x = @@x g @x g This property turns out to be rather useful.

**1.9 Some special derivatives
**

In mathematical physics the curl of a vector is de ned by

~ curl V = @V3 ; @V2 @V1 ; @V3 @V2 ; @V1 @x2 @x3 @x3 @x1 @x1 @x2

!

(1.78)

For the covariant derivative of a vector V we can similarly de ne the curl to be V ; V . Using the expression (1.61) we see that the terms involving the Christo el symbols drop out and we just have @V V ; V = @V ; @x (1.79) @x i.e. the curl is given in terms of just the ordinary derivatives. The divergence of a vector is given by V = @V + ; V @x Using the expression (1.27) we have ; =1g 2 ;

(1.80)

"

@g + @g ; @g @x @x @x

#

1 = 2 g @g @x

(1.81)

Now we can use (1.77) to simplify this to 1 = 1 @ ln g = p @ pg 2 @x g @x (1.82)

Eq. (1.80) can then be written in the elegant form

V

In eq. (1.54) we mentioned that pgd4x is invariant. Gauss' theorem in four dimensions can then be written Z p Z Z @ d4 x g V = d4x @x (pg V ) = d pg V (1.84) where P are 3-\surfaces" (e.g. for = 0 d o = dx1 dx2 dx3). If these surfaces are taken at in nity and if V ! 0 at in nity, one has

1 = p @ (pg V ) g @x

(1.83)

Z

d4xpg V

=0

(1.85)

16

GENERAL RELATIVITY

For tensors similar simpli cations occur. The covariant derivative of a tensor T is given by T = @T + ; T + ; T (1.86) @x Thus the divergence is given by T = @T + ; T + ; T (1.87) @x Using (1.82) we can combine the rst two terms just like we did in eq. (1.83), 1 @ T = pg @x (pg T ) + ; T (1.88) If the tensor is antisymmetric, F = ;F (1.89) then the last term in (1.88) drops out (because ; is symmetric in and ), and we have the simple result 1 @ (1.90) F = pg @x (pg F )

The usual Laplace operator acting on a scalar equals divgrad. In our case there is a similar result. The gradient of the scalar S is just @S=@x , which when multiplied by the metric can be contravariant. We can then apply (1.83) to obtain ! 1 @ pg g @S 2g S pg @x (1.91) @x where 2g generalizes the usual d'Alembert operator @ 2 =(@x )2. In a local elevator 2g reduces to the d'Alembertian.

**1.10 Some applications to physics
**

Having developed the apparatus of tensors and generalized di erentiation let us enjoy ourselves with some applications to physics. First let us consider again the problem of a particle which falls in a gravitational eld g (x), but this time we apply the principle of general covariance. Thus, rst we must know what happens in special relativity. Here the particle is completely free. Its four-velocity is given by ! y dy = dy0 d~ (1.92) U = d d d where y are the coordinates in special relativity. The equation of motion is simply dU = 0 (1.93) d Now we must make this equation generally covariant. This was done in eq. (1.67), where we showed that d=d should be replaced by D=D , i.e. we have DU = dU + ; U U = 0 (1.94) D d This equation is correct because

1.10 SOME APPLICATIONS TO PHYSICS

17

1) it is valid in any freely falling local elevator (; = 0 locally), 2) it is generally covariant (a vector DU =D vanishes in one system, and hence it vanishes in all systems). From this example we see that once we have been through the somewhat tiresome procedure of setting up tensor calculus, the result (1.94) can be derived in a much simpler way than by applying the equivalence principle directly. To present another example let us consider a small gyroscope which moves in the gravitational eld g (x). In special relativity it is thus not in uenced by any forces. The gyroscope taken to be at rest has an angular momentum. In the limit where it can be ~ considered to be a point we say that it has a spin S . The spin is conserved if no external ~ force acts. We can then introduce the four vector S = (0 S ), and hence dS = 0 (1.95) d Also, since the gyroscope is at rest (the spin is de ned in the rest frame) we have

U S =0

(1.96)

These equations are valid in special relativity. Under the in uence of the gravitational eld they become dS ; ; U S = 0 DS (1.97) D d

SU = 0

(1.98)

Eq. (1.97), for example, describes the precession of a gyroscope carried in a satellite moving in the gravitational eld from the earth. To actually utilize eq. (1.97) we need, however, more accurate result for the metric than the one obtained in the Newtonian limit (see eq. (1.37)). Newton's equation (1.1) can be generalized to special relativity, m dU = f (1.99) d where f is called the four-force. In general relativity this becomes (1.100) m DU = F D where F is a generalized four-force. We now turn to electrodynamics. The Maxwell equations can be expressed in terms ~ of an antisymmetric tensor F , with F 12 = B3 F 23 = B1 F 31 = B2 (B is the magnetic ~ eld), F 01 = E1 F 02 = E2 F 03 = E3 (E is the electric eld). Furthermore, the current ~ ~ is a four-vector, J = ( J ), where is the charge density and J is the current. The equations ~ ~ ~~ ~ ~ ~ E ~ (1.101) div E = rE = r B = curl B = @@t + J

104) then becomes F where e.102) and (1.102) (1.105) (1.108) Now F is antisymmetric. F (1. Eq. +F +F F = @F .109) @x Adding the three terms in (1.'s cancel.J @x B ~ ~ r E = .J (1. @@t ~ (1. (1. Eq. F .106) we have @ 2 (pg F ) = 0 = . The conservation law @ (pg J ) = 0 @x (1. so we can simplify as in eq.104) are of course valid in special relativity as a matter of fact. (1.103)] are the raison d'etre for special relativity.103) ~ ~~ div B = rB = 0 can be expressed as @ F + @ F + @ F =0 (1. .110) @x @x @x In special relativity the four-current is conserved. (1. Now we can apply our tensor apparatus to obtain the e ects of gravitation on electromagnetism.107) =0 (1. they or rather.102) is simple to write in a generally covariant form.108) it is seen that the six terms containing the .104) @x @x @x where the indices are permuted cyclically. @ (pg F ) = . the equivalent eqs. F = . F =g g F which is still antisymmetric. From eq. Eqs.g. and we thus get the simple result @F + @F + @F = 0 (1. . @ (pg J ) @x @x @x because F is antisymmetric. (1.111) (1.101) and (1.104) generally covariant we lower the indices.89). (1.18 become The other Maxwell equations GENERAL RELATIVITY @ F = .pg J @x To make (1. It is easy to see that the same must be true in general relativity.106) (1. (1.112) .

However. T . F = e F dx d where F is a vector since F is a tensor and dx =d is a vector.99)] f = eF dx d In general relativity the generalized four force (1. . In curved space derivatives are replaced by covariant derivatives.118) T = @T .85)] in generally covariant form. The opposite is. the derivatives commute.atness is an illusion.g. where x y z are the Cartesian coordinates 1 In the new system it appears that the metric does not represent at space. however. . where . and g (x) = everywhere. if we di erentiate a vector two times. x ! x0 .g0 (x0 )dx0 dx0 @x @x g 0 ( x0 ) = (1. but who has obtained this eld from at space by applying some complicated mathematical transformation x ! x0 . The spherical coordinates x1 = r x2 = x3 = ' thus look \non. Finally we mention that in special relativity the Lorentz force d~ = e E + ~ B ] p ~ v ~ dt can be written compare with eq. where the hydrogen atom is most easily solved in spherical coordinates. in quantum mechanics.11 CURVATURE precisely corresponds to the vanishing of the four-divergence of J compare (1.d 2 = dr2 + r2 d 2 + r2 sin2 d'2 .at" with g11 = 1 g22 = (x1 )2 g33 = (x1 )2 (sin(x2 ))2 (gij = 0 for i 6= j ).d 2 = (dx)2 + (dy)2 + (dz )2 . (1.114) (1. Thus Such a transformation is very often used e. The question then arises whether we can somehow avoid being fooled by somebody who claims to have an interesting gravitational eld g .116) We can. It is clear that this \non.113) (1. it is clear from the derivation that the apparent non. (1. and then 1 . since we also have . not true. d 2 = .11 Curvature The principle of equivalence (and the principle of general covariance) shows that a gravitational eld can be represented by a metric tensor g (x) which is not everywhere constant.115) 1. always transform the x-coordinate to a di erent system. however. Suppose space is just at Minkowski space. To avoid such a joke we can notice that in at space. T @x d 2 = . 19 with eq.117) @x0 @x0 .at" metric does not represent any gravitational eld. dx dx (1. which in general do not commute. To see this let us notice that (1.100) therefore becomes.1.

From the curvature tensor (1. V @x From this we get (. so we can take T = V . S say. However. since the covariant derivative of a vector V is a tensor. . . V @x V where . 1. . .-terms can be brought to vanish. V (1. (1. Thus (1. The equation satis ed in an arbitrary system is thus dS = . (1. . R = @@x . . .R V (1. . @V . V + .20 GENERAL RELATIVITY we obtain (the ordinary derivatives of V commute and thus drop out) In this equation T is an arbitrary tensor. @2g . . the .122) . and the curvature tensor is thus locally given by the second derivatives of the metric. . ] (1. Now it is clear that nobody can fool us any longer. . .125). Here S can be the spin of a small gyroscope or the velocity vector. satis es the simple equation dS =d = 0.. @@x + . it is zero in any other system. since if space is at. .125) d d The vector S is said to be parallel transported if it satis es (1. . in any local elevator the vector. R is called the RiemannChristo el curvature tensor (or just the curvature tensor). @2g + @2g 1 = 2 @x @x @x @x @x @x @x @x +g . V = @V . S dx (1.122) is a vector.120) @x @x Inserting now (1.121) V = @V . is symmetric in and !) V .119) V = @V .V = . .12 Parallel transport and curvature In section 10 we discussed the \transport" of vectors which fall freely in a gravitational eld and which are not in uenced by any forces.124) If we go to a local elevator. if a tensor vanishes in one system.123) we can construct a covariant tensor. We leave it to the hard-working reader to show that it can be written in the form R = " R g # @2g .123) The quantity R is a mixed tensor because the left-hand side of eq. V . it follows that the curvature tensor vanishes in the at system. Thus.122) is a tensor and because V on the right-hand side of (1.

(1.129) Furthermore.12 PARALLEL TRANSPORT 21 Figure 1.127) .131) . j a nS (P ) . j0 . . . is evaluated in P0 . so we get S (P0 P1 Pf ) = S (P0) .126) where the index 0 on . Next we want to go from P1 to Pf . j0 S (P0) b . j0 + @x 0 0 " 0 0 = S (P1) . S (P0 P2 Pf ) = . means that . . R S (P0) a b (1. . since the vector S can experience di erent gravitational elds on the two paths. In general relativity we do not expect to obtain the same result. 1. j0 S (P0) a (1.1: Parallel transport of a vector from P0 to Pf through two di erent paths P0P1 Pf and P0P2 Pf Let us now consider the transport from a point P0 to a point Pf through two di erent paths. @@x j0 S (P0 ) a b + .130) For the path P0 P2Pf we can obtain S (P0 P2 Pf ) from the above expression just by interchange of a and b. j0 S (P0) b . j0 + @@x j0 a From (1. as shown in g.126).127) in (1.128) we get to second order ) ( @ . j0 . j0 S (P0) a . which is given by . in the point P1.125) S (P1 ) = S (P0) . Subtracting we nd S S (P0 P1 Pf ) . j0 S (P0) a b (1. " j0 S (P0) a (1. .125) we have S (P0 P1 Pf ) = S (P1) .1. which are considered to be in nitesimal.125) we need . In order to apply (1. j S (P ) a"o b S (P0 P1 Pf ) = S (P1) . The change in S can be computed from (1. S (P1 ) can be expressed in terms of S (P0) through (1. jP1 = . The parallelogram has sides a and b . In special relativity the vector would be the same in Pf whether we used the path P0P1Pf or P0P2 Pf . .126) and (1.128) Inserting (1. jP1 S (P1 ) b (1. . @ . j0 S (P0) a b @x " b +.1. . Let us consider the path P0P1Pf .

Thus. If the curvature tensor vanishes we can always de ne everywhere inertial coordinates (\ at coordinates") by parallel transport. we assume that the Riemann tensor vanishes. Contraction of the other indices leads again to R because of (1.132) it is symmetric The contraction (1. Apart from the algebraic properties mentioned above the curvature tensor also satis es important di erential identities.13 Properties of the curvature tensor The covariant form of the curvature tensor (1.138) which is called the Ricci tensor. which is parallel transported through two di erent paths to the same point.132) (1.=@x -terms. it is necessary and su cient for a metric g (x) to be equivalent to a at metric that R = 0 everywhere. Because of (1. (1. These are most easily derived from (1. Di erentiating R with respect to x one obtains third derivatives on the metric as well as . In other words. (1.137) is essentially unique: if we contracted instead the rst two indices or the last two indices we get zero because of (1. We can make a further contraction in order to get a scalar R=g R =g g R (1. and only depends on the space-time point.131) we see that as expected.133) (1.122) for covariant derivatives (the left hand side of (1. if there exists a gravitational eld then in general a vector.139) Contraction of and and and gives zero because of antisymmetry in these indices. We can form the contracted curvature tensor R =g R R =R =R (1.R =+R =. @ . then it follows that S becomes independent of the path. At one point we can prescribe a value of S . S is a eld S (x). (1.136) Eq. these terms vanish .124). does not acquire the same value. R +R R R R R +R =+R =. on the other hand.125).132). However.124) by going to a local elevator where . If.134) (1.135) is a simple consequence of the commutation rule (1.134) and (1.124) has a number of symmetries which can be read o from (1.125).135).137) (1. (1.135) (1.22 GENERAL RELATIVITY where R is the curvature tensor de ned in eq.R =0 ( $ ) ( $ $ ) ( $ ) ( permuted cyclically) ! (1. 1. vanishes.135). This equation simply means that the covariant derivative of the eld S (x) must vanish.122) is clearly antisymmetric in and ).133) and (1.123). From eq. which satis es the di erential equation (1. (1.133). and the value in any other point (in a domain where R vanishes) is then obtained by integrating (1.

R 1 (1.144). (1. Remembering that the covariant derivative of the metric tensor vanishes we nd by contracting and in (1.35) and (1. (1. Thus.14 THE ENERGY-MOMENTUM TENSOR 23 in the local system. Eq. we have from (1. where the covariant derivatives reduce to ordinary derivatives. how do we determine the gravitational eld g (x) ? So far.144) suggests that we should attempt to understand how (~ ) can be generx alized to arbitrary relativistic velocities. In generally covariant language this means R +R +R =0 (1. and cyclically and adding the three expressions one nds that this sum vanishes.140) This equation is valid in the local elevators.144) where (~ ) is the energy density in the limit where all velocities are small relative to the x light velocity. In special relativity the ow of matter is characterized by its density and the four-velocity U (x) of matter at the point x .14 The energy-momentum tensor We now turn to the essential problem facing us: Suppose we are given some distribution of matter e.140) R Contracting and we get and . in our universe) in terms of special relativity. From this we can construct a tensor with two indices T = (x) U (x) U (x) (1.37)) r2 g00(~ ) = . (1. Let us consider a simple situation where we have a ow of matter.1. which for small velocities reduce to the right-hand side of eq.2 R =0 Again.1 g R =0 (1. we should attempt to generalize the right-hand side to a tensor with two indices. we know how to represent some matter distribution (e. since the covariant derivative of the metric vanishes. (1. 1.143) plays a fundamental role in the theory of general relativity.R +R =0 =0 (1. The proper density (x) is then the density measured by an observer moving with the ow.142) R .144) suggests that in more general situations it should be represented by a tensor with two indices (this tensor should somehow at least contain second derivatives of the metric tensor).141) . where we found (eqs. 8 G (~ ) x x (1.g. If we can nd this.145) . The left-hand side of eq. (1. Permuting the indices . as we shall soon see. our universe. we only know how to proceed in the Newton limit (sect.R R . 4). and it is generally covariant and hence it holds in all systems.142) R .143) 2 It turns out that eq.g.

and ~ . (1. v2 which expresses the conservation of a quantity of material with density =(1 . we see that T 00 in eq.152) @T = 0 (1. In di erential form we just have ! @T o = @ ~ =0 v ~ (1.146) is the density measured by a xed observer who sees the matter passing by with a velocity ~ .150) @x @t 1 .146) we see that in the Newton limit v << 1 we get T 00 .149) T 00 The quantity T is called the energy-momentum tensor (of special relativity).vv2 (1.151) 2 @t + ~ rvx 1. ~ 2) x v (1. T 0i is the density of momentum (as seen by a xed observer) and T ij is the current of momentum. v2 @z 1 . (1.153) @x . so eq. v2 where v is the usual velocity ~ = d~ =dx0. T has the property that for closed systems the energy and momentum are conserved.vv2 i T i0 = T 0i = 1 . and that the volume decreases by a factor 1 . Thus we have by combining (1. and where we used the special relativistic v x 0 . and in the absence of any forces this velocity cannot change.24 Let us consider T 00. v2 @x 1!. (d~ )2 = (dx0)2 (1 .146) = d 1 . v2) moving with a velocity ~ (x). v2 + r 1 . v2 @y 1 . v2 @vx v ~ = (1.v where we used eq. v2 @t v since the quantity in the bracket in eq. v We also have ! vx + @ vx 2 + @ vxvy + @ vxvz @T 1 = @ @x @t 1 . (1. GENERAL RELATIVITY ! dx0 2 = 1 (1.148) Recalling that the mass of a small volume of moving material increases by p factor a p 2 1= 1 . v relative to the rest mass.150) and (1. v The other components are just i j T ij = 1v.152) @x 1 . (1. 8 G T00 (x) (1.147) From eq.150). Thus ! @T i = @vi + ~ rvi = 0 ~ (1. relation between d dx v d 2 = (dx0)2 . v2.144) becomes approximately r2 g00(x) = . (1.152) is just the change in the velocity for an observer following the stream of matter. Thus T 00 is simply the relativistic energy density.

149) implies that Z P = d3x T 0 (~ t) x (1.157) To see that T is correct notice that it is a tensor (in special relativity). because (1.159) . To make the above result generally covariant we proceed as done many times before. however. v2) The energy-momentum tensor (1.146) since ~ = 0.14 THE ENERGY-MOMENTUM TENSOR 25 and we say that the energy-momentum tensor is conserved. From (1. In (1.158) becomes replaced by T =0 (1.145) represents an extremely simple physical system.158) @x leads to the special relativistic equations for an ideal uid.156) The (00)-component is the same as in (1. not write down. and the result for the (ij )v component expresses the meaning of isotropy. We can now express T in a frame where the uid moves with velocity ~ (~ t). (1. The conservation equation (1.153) it follows that the total energy-momentum P is conserved.154) is indeed a four-vector (remember that volume d3x decreases by a factor 1 . which we shall. and that it reduces to (1. To nd the energy-momentum tensor we simply use the rest frame where i where we used that dP = Z d3x @T 0 = Z d3x @T = 0 dt @t @x (1. and hence the energy-momentum distribution of such uids is of particular relevance in the theory of gravity. vx T =p + (p + )U U (1. Conservation of energy and momentum in the di erential form (for a closed system) @T = 0 (1.153) d3x @T i = 0 @x if T i vanishes at in nity.146) and (1. Similarly the (i0)-components must vanish due to isotropy. (1.156) in the rest frame.155) Z (T 00 )rest = (T i0)rest = (T oi)rest = 0 (T ij )rest = (T ji)rest = p ij (1.149) it is easily veri ed that P de nedpby eq.146) is again the proper energy density. A somewhat more useful case can be constructed by considering a perfect uid. whereas p is called the pressure.1. From eqs.154) and hence from (1. Ideal uids are often used to approximately describe our universe at large scales (much larger than the size of galaxies). which is de ned as a uid characterized by a velocity eld ~ (x) such that an observer moving v along with the uid sees it as being isotropic around each point.

This is a general property. @ 4 g=@x4 by (length)2 etc. i. The total energy-momentum can (tentatively) be written down in analogy with (1. Annalen der Physik (Leipzig) 49 (1916) 769).160). The energy-momentum tensors constructed above are symmetric. T = T .148). The explicit form of the energy-momentum tensor for an ideal uid interacting with gravity is given by T = p g + (p + )U U (1. where the tensor E should depend on the metric and its rst and second derivatives. 8 G T00 (x) (1. they are incredibly small and of no relevance for large scale phenomena in our universe. which were \derived" by plausibility arguments in 1916 (\Die Grundlage der allgemeinen Relativitatstheorie". however. Also. (1. the so-called Planck length. whereas the righthand side is the (00)-component of the energy-momentum tensor.160) where we can say that the right-hand side represents a gravitational force density.27 erg s one obtains LPlanck ' 10.162) since T is a tensor.163) it is clear that we need to include second derivatives. (1. and since it reduces to (1. 8 G T (x) (1.29 cm/g and h ' 10. P is not a contravariant vector. which can be written as in eq.163) The left-hand side contains the second derivative of the metric tensor.26 Using eq.154).. We return to the Newtonian limit. They . r2 g00 (x) = .15 Einstein's eld equations for gravitation We are now able to proceed to present Einstein's eld equations. From (1.161) which is. not conserved due to the non-vanishing of the right-hand side of eq. (1. However.88) we get 1 @ pg @x (pg T ) = . P = d3 x pg T Z 0 (1. assumed to be valid in the following.165) LPlanck 2 = G3h c Using G=c2 ' 7:41 10.33cm (1. Thus it is natural to generalize this equation to E (x) = . such terms would have to be multiplied by a constant with dimension of a power of length (@ 3 g=@x3 -terms would have to be multiplied by some constant with dimension of length.164) in arbitrary coordinates. Looking around among the fundamental constants of nature there is only one constant with dimension of length. (1.157) in special relativity. and the reason we do not include higher derivatives is primarily because of simplicity.e. Physically non-conservation of P is due to the possibility of exchanging energy and momentum between matter and gravitation.). 1. T GENERAL RELATIVITY (1.166) It therefore follows that if the constants of dimension length to a power are present.

170) we get 1 E00 c(R00 + 2 R) 2c R00 (1. we saw in sect. 13 that there is essentially only one such tensor. (1.174) Such terms occur in superstring theories (one of which could be the \right" theory of quantum gravity). which we shall x by requiring that we get the correct Newton limit.167) is wrong. 2 . (1. and should be linear in the second derivatives of g . R00 i. Because of the symmetries of R .170) E = c(R .172) 2 In the weak eld limit. it should be a tensor which represents the genuine physical contents of gravitational elds. (1. For this we have only one candidate. which is supposed to be very small (big bang). R00 ' 2 R . However.168) Going back go eq. Due to the occurrence of h it is clear that quantum phenomena must be involved.170) Rij 1 gij R (1. (1. and hence such terms could be of relevance in quantum gravity. 2 g R) which is covariantly conserved. 1R R (1.143) it follows. 8 G T00 Since jTij j << jT00j. Eqs.167) The answer is no. namely R .143) shows that the right answer is 1 (1. Could we take E =R ? (1. In sect. (1.164) and (1. Thus eq. 2 g00R) = . Here c is a dimensionless constant. Thus 3 R = Rii . 2 However. (1. Since E has two indices. eq. that R 6= 0 in general. it should depend on the contracted curvature tensor.33 cm. Also. 14 it was shown that the energy-momentum tensor satis es di erential conservation T =0 (1. we also have that Eij is very small. In a non-relativistic system the pressure terms are much smaller than the energy-density. here we shall proceed to consider only scales which are large relative to 10.171) c(R00 . so from (1.164) we therefore also have E =0 (1.1.170) reduce to 1 (1. namely the curvature tensor and quantities derived from it. to the rst approximation gij ' ij . however.15 EINSTEIN'S FIELD EQUATIONS 27 could be relevant in the very early universe.e.164) thus depends on the geometry of space-time.173) 00 2 Inserting this in (1. The Einstein tensor E (x) in eq. so jTij j << jT00 j.169) From eq.

and in this chapter we shall ignore it. (1.182) R . T = T . a small . which allows one to compute the gravitational eld from a given energymomentum distribution. Consequently Einstein's eld equations read 1 R .182) requires that T is symmetric. (1.178). 1 g R. g =. 8 G(T .177) This shows that we obtain eq.8 GT (1. (1. and we then get 1 R = . It should be noticed that consistency of the Einstein equation (1. Thus " # 1 @2g . 8 G T R .180) Eq.182) does not satisfy the Newton limit. By contraction we get R = R .174) we get E00 2c R00 2c(Ri0i0 . (1.175) R 2 @x @x @x @x @x @x @x @x Since all time derivatives vanish we get @2 R0000 0 Riojo 1 @xig00j (1. (1.180) is. eq.2 g R=. @2g . but in the local case there are other non-trivial solutions (due to the non-linearity of eq. 8 G T 2 This is allowed from the point of view of covariant conservation of T . and (1. The cosmological constant must therefore be rather small. where the .176) 2 @x so from (1. R0000 ) cr2 g00 (1.181) Flat Minkowski-space satis es this equation globally and locally. as we shall see later. has profound cosmological e ects. which seems to exist. @2g + @2g (1. since the covariant derivative of g vanishes.1 2 2 or R=8 GT (1.178).178).178) can be generalized by adding the \cosmological term".124). In empty space T = 0. However. . To conclude this section we now have a non-linear set of second order di erential equations (1.28 GENERAL RELATIVITY Now R00 can be obtained from (1. (1.. fully equivalent to eq. (1. This is precisely the contents of Einstein's general relativity. Eq.178) This equation can be written in a di erent form. 2 g T ) (1.180) then gives the vacuum Einstein equation R =0 (1. of course. However. 1 4 R = .R = . (1.163) provided c = 1.179) This can be inserted in eq.-terms are of higher order and hence can be left out in the Newton limit.181)).

C (r t)drdt . which are functions of t and r0. Once we are given at a certain time for all r. In the following. we can write d 2 = A(r t)dt2 . B (r t)dr2 . First.186) becomes 2! 2 = 1 dt02 .186) The \mixed" drdt-term can also be removed by resetting our clocks by use of a new time coordinate t0.187) becomes a perfect di erential with A = @t0 =@t . the function D can easily be removed by introducing a new radial variable.188) @r 2 @t This is a rst order partial di erential equation for .185) d 2 then becomes dependent on new functions A0 B 0 . 2 C (r t)dr] (1. Thus 29 p t r dt ~ d~ = r dr xx d~ 2 = dr2 + r2(d 2 + sin2 d'2) x (1. 3 . 1 C = @t0 =@r.189) A2 4A the proper time (1.1. Thus. The proper time interval can then only depend on quantities which are rotationally invariant. Using that 1 dt02 = A dt2 .187) Here (r t) is an integrating factor de ned such that (1. In particular. for simplicity we only consider non-trivial transformations (r0 = const turns out to be inconsistent with the Einstein equation).184). C dtdr + C 2 dr2 (1.183) (1. D(r t)r2(d 2 + sin2 d'2) In general relativity it is important to realize that the coordinates are arbitrary. B + C dr 2 .16 SPHERICALLY SYMMETRIC METRIC 1. r2 (d 2 + sin2 d'2 ) d 2A 4A It could happen that this transformation produces a constant r0 . we can compute it at any other time. we are free to make transformations x ! x0 in (1. namely where r = x2 + y2 + z2 . we can always try to select such transformations in a way which simpli es the expression (1. Dropping the primes.184).184) d 2 = A(r t)dt2 . This requires the integrability condition 2 @ ( (r t)A(r t)) = . and C 0.16 The time-dependent spherically symmetric metric In order to gain insight in Einstein's equations we shall start by considering a timedependent spherically symmetric gravitational eld g (x). 1 @ ( (r t)C (r t)) (1. 3 q r0 = r D(r t) (1. since A and C are given. which is de ned by 1 dt0(r t) = (r t) A(r t)dt . B (r t)dr2 . r2(d 2 + sin2 d'2) (1. C (r t)drdt .

192) (1.193) g = r4EF sin2 The invariant volume element is thus q r2 sin E (r t)F (r t) dr d d' dt If E and F were constants. @g (1.194) @x @x @x This is a straightforward but tedious method. . we obtain by a partial integration. for the metric (1. and _ _ demand that the trajectory x = x ( ) is obtained by requiring the minimum variation of the integral Z F (x x )d _ i. =1g 2 # @g + @g .E 1 gtt = . Requiring that the variations vanish at the ends of the interval of _ integration. Z Z F (x x )d = _ F (x x )d _ # " Z @F @F x d = 0 (1. E (1. d @ x x d = 0 _ " Now x = d x =d . One method is to use eq.e. F (r t)dr2 .195) = @x x + @ x _ _ Z Z " @F d @F # Fd = @x .191) (1. renaming these functions. (1.e. to use . 1. In order to write down the Einstein eld equations we need rst to know the Christo el symbols . where x = dx ( )=d . A somewhat simpler way consists in obtaining the . The metric thus has the form g = r2 g = r12 whereas the determinant g is grr = F 1 grr = F g'' = r2 sin2 1 g'' = r2 sin2 gtt = .191). i.'s from a variational principle. First let us consider some functional F = F (x x ).190) This form of the metric is called the \standard" form of the metric. this would just be the usual volume element in spherical coordinates.17 A digression: A simpler method for computing .27) directly.30 or. GENERAL RELATIVITY d 2 = E (r t)dt2 . r2(d 2 + sin2 d'2) (1.

dx dx = 0 (1. _ (1. i.18) we see that this is precisely the equation of motion for a freely falling particle in a gravitational eld speci ed by the metric g (x).199) can be obtained from the variational principle Zs g (x) dx dx d = 0 d d (1.199) From (1.200) can be written d = 0. or d2 x + . (1. 31 (1. As a side{remark we mention that use of the Euler-Lagrange equation (1.199).198) @g dx dx + g d2x = 1 @g @x d d d 2 2 @x so after multiplication with g we get ! d2x + 1 g 2 @g . Eq.200) is the distance.R since the square root times d in (1. To return to the main issue of obtaining the Christo el symbols we see that eq. (1. In our case the line element (1.197).194) and (1.e. @g dx d2 2 @x @x d dx dx d d dx = 0 d (1. (1. r2 sin2 '2 d = 0 _ _ _ = d =d and ' = d'=d .200) shows that freely falling particles follow a curve with the shortest \distance".201) d2 d d allows one to obtain the .17 COMPUTING .196) also shows that (1.196) with F given by (1.202) where t_ = dt=d r_ = dr=d .1. r2 _2 .'s from a variational principle using the Euler-Lagrange equations (1.190) therefore leads to the variational principle Zh i E t_2 .200) We leave it to the reader to verify this statement. The curve followed by a freely falling particle is therefore called a geodesic.197) d @F = @F d @ x @x _ Now we can easily check that if we take F = g (x) dx dx d d then the Euler-Lagrange equation becomes " # d g (x) dx = @g (x) dx dx 2d d @x d d Doing the di erentiation on the left-hand side we get or the Euler-Lagrange equation (1.196) (1. F r2 .

207) @t F @r Comparing with (1.r = . dd (2F r_ ) = @E t_2 . 2r sin2 '2 @r @r Proceeding as before we perform the d=d on the left-hand side and obtain 1 r r r + F @F t_r + 21 @F r2 + 21 @E t_2 .201) we see that .r = 21F @F . @F r2 @t @t @t _ or 1 t + 21 @E t_2 + E @E t_r + 21 @F r2 = 0 (1. @F r_ 2 .209) which gives _ (1.2r2 sin cos '2 _ (1. F _2 .208) F F .210) + 2 r _ .r r = .r sin2 .204) E @t @r _ E @t _ From this we get by comparison with (1.205) @t Next let us use the Euler-Lagrange equation (1.203) d @t @t _ Performing the di erentiation on the left-hand side (remember that E depends on r and t.202) gives from (1.r = 21 @F tr rt rr @t F @r r = 1 @E r =. _ (1. sin cos '2 = 0 r _ .197) for the problem (1. In that case (1.196) d (2E t_) = t_2 @E . @F r2 (1. dd (2r2 _) = .202) to the case where x in the Euler-Lagrange equation (1.202) when we vary x = r.18 The Christo el symbols for the time-dependent spherically symmetry metric Let us start by applying eq. (1.197) is the time. . F sin2 '2 = 0 _ F @r _ _ (1.ttr = 21 E .trr = 21 E Proceeding now with the variable we get (since E and F are independent of ) . 2r _2 .'s (remember that mixed terms like t_r occur twice in (1.206) .tt 2F @r .201)) _ GENERAL RELATIVITY @E @t @E = .ttt = 1 2E . which in turn depends on ) _ @r 2E t + 2t_2 @E + 2t_r @E = t_2 @E .32 1.201) the following .'' (1.t rt @r @F (1.

1 @ E + 4F 2 @E @F . sin cos 33 (1. .208). (1.' = cos ' r' 'r r sin The reader can now make a psychological experiment where he/she compares the time it takes to obtain the . From eqs.212) Finally we need only to vary '. 1 @E 2 . This would also provide a check of the correctness of the results (1.216) . 1 @ 2 F + 1 @E @F + 1 @F 2 . (1.'s found in the last section as well as g given by eq. we get the following non-vanishing components of the Ricci tensor.211) (1.82).' = .'' = .'' = . 2F 2 @F + 2EF @E @r @r 2 1 1 = .e. (1.1 + F . r = 1 r .137) we have 1 @ 2 ln g . 1 .'s by means of the variational principle and by the direct application of the expression (1.r = .1 @F rF @t 1 @ 2 E .1.178) or (1. .213) r _ _ sin _ Hence the only remaining non-vanishing Christo el symbols become (1. 1 @E @F + 4EF @r 2F @t2 4F 2 @t 4EF @t @t ! R'' = sin2 R (1. dd (2r2 sin2 ') = 0 ' + 2 r ' + 2 cos _' = 0 (1.192).211). @ . 1.19 The Ricci tensor In order to study the Einstein equation (1.180) we need to know the Ricci tensor R . (1. rF @E 2 2F @r @r ! !2 @r @r2 1 @E + 1 @ F .123). . rF @r 2E @t2 4E 2 @t @t 4EF @t r r 1 = .19 THE RICCI TENSOR i. (1. or _ . 1 @F 2 .194).205).214). 1 @E @F = 2E @r2 4E 2 @r 4EF @r @r ! 1 @F . (1. Rrr R Rtt Rtr = .' = 1 .215) 2 @x Inserting the . @ ln g R = 2 @x @x @x (1. and (1. + .214) .

219) F E rF F dr E dr This gives or d ln F = .216) it follows that we only need to consider Rrr = R = Rtt = 0. 2 g T ) (1. Thus.219) Rtt then vanishes automatically). etc. (1. d ln E dr dr (1.1 1 dF + 1 dE = 0 (1. eq.223) R = .225) dr . (1.190) should become at Minkowski at in nity. except for a mass M situated at r = 0. given some energy-momentum tensor T we can in principle obtain second order di erential equation for the metric functions E and F by use of e.224) 2rE (r) dr Thus.221) is xed to be 1 for r ! 1. 1.222) we get (1. It is seen that Rrr and Rtt contain rather similar terms. from the last eq. Using the result (1.218) tr rF @t etc. except for r = 0 we must satisfy the vacuum Einstein equations R = 0. (1. from R = 0 we get d (rE (r)) = 1 (1. and it is then easily checked that all the time derivatives drop out of the other R's.180). Therefore we form ! Rrr + Rtt = .g.221) We now impose the boundary condition that the metric (1.34 GENERAL RELATIVITY Thus. if R = 0 then Rrr is automatically zero.1 + E (r) + r dE (r) dr 2E Rrr = 2E1(r) d dr(r) + rE1(r) dE (r) 2 dr 1 d R (r) = (1.222) ( Now we can take R and Rrr to vanish (because of (1. Using (1. so the constant in (1. 1 @F = 8 G T (1.220) E (r) F (r) = constant. 1 R = . E (r) = F 1r) (1. Further.20 The Schwarzschild solution We shall now consider the case where space is empty.216) we see that this implies that the time-dependence drops out: From Rtr = 0 we get that F is independent of time. Hence. 8 G (T .217) For example.

190) we get C = . and all time-derivatives disappeared.229) t0 Zt q = f (t) dt (1.g.190) becomes (1. 1 . Birkho 's theorem can also be applied if we have a spherical cavity in a spherical mass distribution.1. from (1. A non-static body can in general emit gravitational waves. and hence GM = 0. Thus. which can be xed by means of the results g00 ! .21 Birkho 's theorem We saw from Rtr = 0 that the function F is time-independent. and we have Birkho 's theorem saying that a spherically symmetric gravitational eld in empty space must be static with a metric given by the Schwarzschild solution. 2MG r dr2 . . t ! t0 where (1.227) d 2 = 1 . if e. r2(d 2 + sin2 d'2 ) (1. the universe was a spherical mass distribution. and the metric is just at Minkowski. E (r t) = f (t) 1 . where the body needs not be static. and gravitation would have no e ect in a spherical cavity.230) Thus the Schwarzschild metric (1. 2MG dt2 . However.228) 1. It is a priori far from obvious that the similar result applies in general relativity. This theorem is similar to Newton's result that the gravitational eld outside a spherically symmetric mass distribution behaves af if the mass was concentrated in the centre. and hence E (r) = 1 .1 + 2GM=r obtained in the Newton limit. 2MG r and the metric (1. In this case there is no mass at the center.21 BIRKHOFF'S THEOREM so 35 rE (r) = r + constant = r + C (1. r This is the Schwarzschild solution.228) always results. the metric would be at. see eqs. 2MG r but f (t) can be removed by a rede nition of time.36) and (1. (1. Thus.2MG. Birkho 's theorem shows that no gravitational radiation can escape from a spherical body into empty spaces.37).222) we then know the functions E (r) and F (r) in terms of the constant C . 2MG r 1 F (r) = 1 .226) From (1. In principle the function E could depend on time.

234) we get 1= 1 .233) d r where the dot denotes di erentiation with respect to . Substituting these results into (1. We also have an equation for r. Hence this apparent singularity is not relevant for the solar system.236) (1. since for a photon d = 0.233) gives 1 .g. (1. (1. We thus have four equations for the four functions t( ) r( ) ( ) '( ).228) for the solar system. 2MG t_ = 0 (1. and (1. then will be xed to the same value for all other 's.22 The general relativistic Kepler problem GENERAL RELATIVITY We shall now discuss some implications of the Schwarzschild solution (1. =2 (1. 2MG r (1. In (1. 1 . Using that E and F can be taken to be time-independent. We shall discuss this phenomena later. Eq.36 1. r2( _2 + sin2 '2 ) _ (1. In eqs.231) d d (r2 sin2 ') = 0 _ (1. (1. and using the result (1. since the solution (1. which is deep in the solar interior.204) . 2MG t_2 . Also.234) Here we assumed that the particle is not a photon.232) d d 1 . We shall return to the photon case later. we get d (r2 _) = r2 sin cos '2 _ (1. (1. eq. Taking e. and _ = 0 for = 0 .228).228) it appears that there is a singularity for a radius r = 2MG. H2 r r 2 2 (1.228) is only valid outside the massive bodies (for the earth the Schwarzschild radius is 1 cm). A great simpli cation occurs because this motion lies in a plane: It is easy to see that if we x at some initial value of . so we can study planetary motion by solving these equations. Here we only remark that for the sun 2M G = 2:95km. r 1 .235) for some = 0 . (1.210). 2MG r r_2 . which we replace by the line element (1.231) implies that this remains true for all .213) we have written down the equations of motion.227).r_2MG . which is called the Schwarzschild radius.207). then eq.232) can then be integrated r2 ' = H _ where H is a constant of integration.238) .237) L2 . 1 = 1 . (1. 2MG t_ = L r where L is another constant.

since 2MGu 3 10. (1.242) we have r0 = .236) would be related to the usual time (not the proper time).242) r It turns out that this substitution is also convenient in general relativity. 2MGu = L2 .7 is the Schwarzschild radius divided by the (average) planetary radius. u2 + 2MGu3 2 H2 2 ' = '0 + Zu u0 q L2 .1.241) 1 . Eq. in practice elliptic integrals are not very useful.) with dr=d' = 0. 1 + 2MG u .22 THE KEPLER PROBLEM 37 Inserting this in (1.236) we thus have (1. if we did not have this term.243) (1. As a matter of fact. and (1. called the aphelia (perihelia).241) gives 1 . In the classical Kepler problem it is convenient to introduce the new variable u= 1 (1.244) (1. i.g. (1. for u = u we also have du=d' = 0. so the inverse function provides us with u = u(') = 1=r(').245) would be the Newtonian result. H 2u02 . 2MG r r r r which is a di erential equation for r = r('). H2 1 . 2MGu) which can be solved for u0.239) _ From (1. We have dr r _ r0 d' = ' (1. and eq. H 2 (1. (1. To do this systematically we notice that in the motion around the sun the orbit reaches a maximum (minimum) distance. and we would therefore like to know r as a function of '.u0=u2. the u2 term. H 2u2(1 . instead of . where r = r+(r.1 H2 du MG + 2H 2 u .e. This suggests that we expand the integral (approximately) around the Newtonian solution. From (1. H4 r02 .246) 2 .240) r = 'r0 = H r0 _ _ r2 2 2 (1. However. H 2(1 .245) This gives the exact result u02 = L H.243) then gives 2 1 u2 .238) we get In applications to orbital motion we are interested in the shape of the orbit. To orient ourselves in this mess let us notice that the term 2MGu3 under the square 1 root is expected to be very small relative to e. u2 + 2MGu3 The integral is of the elliptic type.245) gives ' = '(u). the result (1. 2MG = L2 . except that the constant H in (1.L MGu ) = .

252) can be performed trivially by making the substitution 1 1 u = 1 = 2 (u+ + u. 2MGu) where ' is measured from the perihelia. u.) sin r and we get q (1 + MGu) du (1.247) Subtracting these two equations we get L2 0 = (u+ . (1. It should equal 2MG L2 =H 2 according to the left hand side of eq. Hence (1.251). (1. 1 p '(r) = '(r.248) (1. 1 MG 1 . u+) p where the constant factor outside the integral comes from expanding 1= C .255) .249) (1.) u+ + u. u)(u . u .+ u ) ( + .) 1 + MG 1 + 1 r+ r.250) (1. (1. . The result (1. H 2 (1 + 2MGu + 4M 2 G2u2 ) + H 2 = 0 (1.251) with coe cient C (u+ + u. which are known from astronomical observations. We obtain to second order L2 1 u2 .+ u ) ( + . u)(u . ! (1.248) has been used.). 21 du L2 2. 2MG H 2 (1 + 2MG(u+ + u.) + .) = 2MG H 2 1 + 2MG+uu.245) can now be written (1. H 2 C (u.251) Consider the term linear in u in (1. 2MGu H 2 (1 . '(r. !! ( + ) . u H2 1=r. u.) + 2 (u+ .253) (u. . 2MG(u+ + u. The integral '(r) .) Inserting this in (1. 1 .254) !! Z 1 r 1 r. '(r+) . . 2MGu) .) = 1 + 3 MG r1 + r1 2 + . where eq. 2MGu L2 1 2 2 (1 .249) we get u+ L2 2MG H 2 = 1 + 2MG+uu.)) or " # " #. Thus C 1 . u+ ) H Z 1=r L2 = u+ C (u+ + u. 1 cos 2 2 r+ r. To proceed we expand 1 1 + 2MGu + 4M 2 G2u2 + 1 . '(r.38 GENERAL RELATIVITY which determines the integration constants L and H in terms of u .

257) For Mercury this gives per revolution. Let us consider the free motion along a curve described by x = x (p). However. dx dx = 0 (1.256) Thus. so the perihelia precesses an amount '. Newtonian gravity also gives a precession of Mercury.261) obs so the value (1.1.)j . and observations go back to 1765. Some discussion of this point has been made in the literature.23 DEFLECTION OF LIGHT 39 Now.259) quite considerably. as we did in the previous section. ( ')Newton.259). ! 1 + 1 radians/revolution.23 De ection of light by a massive body So far we have considered the gravitational e ects for a massive body moving around the sun. In each century Mercury makes 415 revolutions. =2 to = =2. d2x + .. where p is some parameter. '(r. 1. From general relativity we thus have ' = 43:03" per century (1. ( ')Newton = 5557:62 0:20" (1. ' = 3 MG r r (1. nobody has found any convincing evidence against the result (1.262) dp2 dp dp Changing to another parameter.259) is obtained as the di erence ( ')obs = ( ')TOT . For an ellipse this equals 2 .e. which is characterized by d = 0 (i. dx dx dq 2 = 0 dp2 dq dq2 dq dq dp (1. ' = 0:103800 + . 2 (1. with p = p(q) and q = q(p).260) whereas the actual observation gives ( ')Tot = 5600:73 0:41" (1. we obtain " # ! d2q dx + d2x + . since it is sensitive to the second order expansion of g00. p ! q.259) This is the most important experimental veri cation of general relativity. One obs might feel unhappy about the subtraction of two large numbers. when r goes from r.263) .254) that goes from = . It should be pointed out that due to perturbations from other planets and due to the rotation of the earth. where ' = 2j'(r+) . because small systematic errors may in uence (1.)j. Therefore we cannot divide by d . Now we wish to nd these e ects for light. We wish to assume the geodesic equation of motion. The total charge in ' per revolution is thus 2j'(r+) . velocity of light is always one in the freely falling elevators). to r+ it follows from (1. '(r.258) whereas observational data give ( ')obs = 43:11 0:45"per century (1.

dp . + Z 1=r Z 1=r 0 0 .268) .e. 2MG r r r (1. 2MG r !3 d' 25 dp (1. the constant of proportionality vanishes. H2 1 .2: A photon passing the sun Thus we see that we get again the geodesic equation of motion if and only if d2q=dp2 = 0. u2(1 .e. follows that the form of the geodesic equation remains only if we make linear transformations of the parameter. However.40 GENERAL RELATIVITY Figure 1. Thus. and at the point of closest approach r = ro one has dr=d' = 0. 2MGu) (1.267) Thus ' = . 1. In most equations this change only leads to trivial alterations. i. + du u2(1 .237) (with (' ! d'=dp and t_ ! dt=dp. r2 4 d 2 + sin2 dp dp 1 . i.234) now becomes 0 = 1 . It.265) we get L2 = H2 1 . u2 q (1.264) since d = 0. 2MGu) 0 ! du u3 . 22 we get the solution ' = '0 + Z 1=r u0 q L2 du 2 H 2 . therefore. (1. eq. Let us consider a light ray passing the sun (see g. and use p as another parameter.266) which is again a elliptic integral which should be expanded. The equations of motion in the last section should therefore be replaced by similar equations with replaced by p. so from (1. (1.265) Proceeding as in sect. 2MG r r 2 0 0 (1.2). so we can take d 2 = constant dp2. u3 0 q2 1 + MG 2 2 u0 . For a photon one then has d =dp = 0. introducing again integration constants as in eqs. if q is a linear function of p. 2MG r 2 ! ! dr 2 dt 2 . H4 r02 .236) and (1. We normalize such that for r = 1 in the incoming direction ' = . u (1 . we get _ 2 2 0 = L2 . respectively). . 2MGuo) . u u0 .

236). we have _ from (1. However.271) A recent observational value is 1:70 0:1000 in good agreement with (1.270) ' = 4MG r0 For the sun we get.269) we get (1. so the de ection is ' = 2'(r0) . 2MG r 2MG (1. In the asymptotic system let the clock be chosen such that dt=d ! 1 for r ! 1.237) and (1. .24 BLACK HOLES Using we get 41 u3 . If there was no de ection this would be . u3 = u + u2 0 0 u2 . The precession of perihelia in sect. 2MG r 2 = L2 . The quantity ' was rst observed in 1919.270) brought Einstein's general relativity into the newspapers. The total change in ' is 2'(r0). In our solar system this seems not to be relevant. it could be that somewhere in our universe there exists an object which is so small that r = 2MG could be reached. that outside a (possibly timedependent) spherically symmetric solution Schwarzschild's solution is valid. u u0 v 0 s 1 u u + MGu @2 . u 1 . 2MG r _ (1. let us consider a particle originating at large distance r ! 1.273) r2 = r _ . 1 . and L t_ = 1 . u2 . Since ' = 0. 21.272) r Now.24 Black holes As already mentioned the Schwarzschild solution looks singular for r = 2MG. + sin. and the excellent agreement with (1. The question is then what happens at the Schwarzschild radius? We know from Birkho 's theorem shown in sect. Then L = 1. u2 u0 + u 0 ! du MGu0 + ' . 1 . with r0 = the solar radius (= 6:95 105 km) ' = 1:75" (1. Hence. from (1. Let us consider radial motion where r and t are functions of proper time . u=u0 A + t = . 22 is far more accurate.271). both from an experimental and a theoretical point of view. ' is in general very di cult to measure with high precision.1. Z 1=r 1.269) 0 u2 1 + u=u0 0 0 Now r = 1=u goes from 1 to r0 and then goes from ro to 1. where space is at (Minkowski) and special relativity is valid. However. (1. and hence 1 t_ = 1 .1 u (1. + 0 q 2 2 1 + MGu + 1 + u u0 .238) that H = 0.

42

GENERAL RELATIVITY

Now we have two times: the proper time is the one measured by an observer falling freely in the gravitational eld, and the time t is the time measured by an observer at rest at large distances. From the second equation (1.273) we get, since pr r_ = ;p2MG (r decreases with ), 3 p 2 (r3=2 ; r0=2) = 0 ; (1.274) 3 2MG with r = r0 for = 0 . The solution (1.274) is valid as long as we are outside the body, so if the radius of the body is less than 2MG, it follows from (1.274) that the freely falling observer passes the Schwarzschild radius without observing any singularity. If the mass is just a point mass, this observer reaches r = 0 in a nite proper time 2 r3=2 (1.275) r=0 = 0 + p 3 2MG 0 Let us now consider this motion from the point of view of the observer at rest at large distances from the massive object. From (1.273) we have

s dr = r_ = ; 2MG 1 ; 2MG (1.276) dt t_ r r This di erential equation can be solved exactly without any problem. However, it is more instructive to consider the two cases where r is much larger than the Schwarzschild radius or where r 2MG. In the rst case we can use 2MG=r 1, and hence (1.276) becomes approximately the same as the proper time result. Hence, at large distances there is approximately no di erence between the proper time and the asymptotic time, which is of course quite reasonable. For r 2MG the situation is quite di erent. Here eq. (1.276) gives to a good approximation s dr = ; 1 2MG (r ; 2MG) ;1 (r ; 2MG) (1.277) dt r r 2MG

with the solution

t (1.278) r 2MG + constant e; 2MG Thus we see that from the point of view of the asymptotic observer, the Schwarzschild radius is never reached (it takes an in nite time to get there). This feature can be put into perspective if we think of the asymptotic observer as a human being on this earth, who observes a distant spherically symmetric body by sending a test body towards it and compare this with the situation as seen by an observer inside the test body. Seen from earth it takes an in nite time for the test body to get there, and during this time the entire revolution of the universe happens. However, the observer in the test body reaches the Schwarzschild radius in a nite time (on his watch), and then he proceeds. He thus sees what happens after t = 1 (on our clock)! A body which collapses to a sphere with radius equal to the Schwarzschild radius is called a black hole. The reason for this is that no particles (including photons) can be emitted from the surface. To give a full discussion of black holes, one needs also to solve the Einstein equations inside the mass distribution, which turns out to be a relatively complicated task, which we shall not attempt. However, the blackness is clear from

1.24 BLACK HOLES

43

(1.278), since if we reverse the time direction it follows that a particle emitted from the surface r = 2MG would require an in nite time t to reach us. For light d = 0, so radial motion of light satis es 2MG dt2 ; dr2 = 0 1; r 1 ; 2MG r i.e. dr = 1 ; 2MG (r 2MG) (1.279) dt r so the velocity of light seen from the asymptotic system approaches zero as r ! 2MG (!). Comparing with eq. (1.277) we see that if light should reach us from the Schwarzschild radius, an in nite time would be needed, so this can never happen. The body is black. Thus, if a body shrinks to its Schwarzschild radius, it disappears from our view. Since g00 is zero for r = 2MG, it follows from the discussion (1.38), (1.39),(1.40) that there is an in nite red shift. Therefore the shrinking body fades out of sight. Can black holes exist? The stability of a star is determined by the balance between the pull of gravity and the pressure from radiation emitted by nuclear fusion. When the light nuclei have been used up, the fusion ceases, and the gravitational pull may win. In some cases it can then happen that the radius asymptotically approaches 2MG (from above), and a black hole is formed. We refer to the literature for a detailed discussion of these interesting questions. We end by some historical remarks: Black Holes were predicted from Newtonian gravity by Laplace (r2) in 1799, in a paper entitled \Proof of the theorem, that the attractive force of a heavenly body could be so large, that light could not ow out of it." The argument is quite simple. Let us consider light as consisting of particles with mass (ultimately you can take the limit ! 0, since the result is independent of ). In modern terminology we would call the photon mass. At the time when Laplace did his work, there were the competing theories of light, namely that it has a wave nature (Fresnel and others), or it has a particle nature (Newton). We know now that both pictures are right in a duality sense. Anyhow, consider a spherical body with mass M , and let us try to shoot a particle with mass away from this body. The condition for escape is that the kinetic energy exceeds the gravitational energy, so 1 v2 > G M=r: (1.280) 2 Thus, after division by , we see that the critical radius is given by rcrit = 2GM=v2: (1.281) Here Laplace mentions that v is the velocity on the surface of the body. In general it can depend on r, but if we assume that v = c. which we have taken to be one, the critical radius is just the Schwarzschild radius! Thus Laplace correctly predicted the existence of black holes, which is a most remarkable historical fact.

**1.25 The spherically symmetric solution in 2+1 dimensions
**

A result analogous to the Schwarzschild solution can be found in two space and one time dimension. It turns out that this result is quite interesting from the point of view of

44

GENERAL RELATIVITY

principles, so for this reason we shall give the derivation. We start from the metric

d

2

= dt2 ; e (r) (dr2 + r2d 2) = dt2 ; e ( x2+y2 ) (dx2 + dy2):

p

(1.282)

Here we used the notation r2 = x2 + y2. For a static metric, the most general rotationally invariant 2-dimensional metric is A(r)dr2 + B (r)(dr2 + r2d 2), as can be seen as in section R q 1.16. Introducing the new variable r0 = exp dr A+B , one obtains the spatial part of r B (1.282). Using the Euler-Lagrange variational principle, we easily obtain d (2t_) = 0 d (2e x) = (x2 + y2) e @ d (2e y) = (x2 + y2) e @ : (1.283) _ _ _ _ _ _ d d @x d @y The last equation follows by symmetry from the second by observing that the metric (1.282) is symmetric in x and y. Hence all results derived from this metric must be symmetric under interchange of x and y. The Christo el symbols can then be easily read o by performing the ;integrations. For example, the second equation leads to 1 @ _ 1@ _ (1.284) x + 2 @x x2 ; 2 @x y2 + @ xy = 0: @y _ _ In this way we obtain 1@ @ 1@ (1.285) ;tt = 0 ;t = 0 ;x = 2 @x ;x = ; 1 @x ;x = 2 @y : yy xy xx 2 The remaining symbols follow by the symmetry between x and y, ;y = 1 @ ;y = ; 1 @ ;y = 1 @ : (1.286) yy 2 @y xx 2 @y xy 2 @x Also, for the logarithm of the metric tensor we have

q log g = 2 ( x2 + y2) = 2 (r):

(1.287)

We are now in the position to compute the Ricci tensor R , for which we have the general formula 2 @; log (1.288) R = 1 @ log g ; @x + ; ; ; 1 ; @ @x g : 2 @x @x 2 We then obtain Rtt = 0 Rti = 0 i = x y: (1.289) Also @2 @ Rxx = @x2 ; @ ;xx + ;x ;x ; ;xx @x @x @ 2 ; @ ;x ; @ ;y + ; ;x + ; ;y ; ;x @ ; ;y @ xx xx = @x2 @x xx x xy x xx @x xx @y @y ! q 2 2 x2 + y2 = Ryy (1.290) = 1 @2+ @2 2 @x @y

The reason for the identi cation in eq. log r trivially satis es eq. The Einstein equation (1.292) where the operator multiplying exp(.1.293) is valid only in the x .282) thus becomes (1. The corresponding energy density is given by = M 2 (x)=pg: (1.dependent function.299) 2 2 2 r2 2 0 as claimed in eq.8 G T with T00 = and all other components T = 0. (1.y @ = 0: (1. .301) d 2 = dt2 . (1. Eq.294) The Einstein equation 1 (1. therefore gives by contraction.298) r2 (r) = .e. 2 g R = . .x @ . Let us consider a point mass M at rest in the point r = 0. In the neighbourhood of r = 0 we have 1 Z d2x r2 log r = 1 I dn r log r = 1 I dn r = 1 Z 2 d = 1 (1. (1.300) (1.297) if we use that r2 log r = 2 2 (x) which shows that log r is the Greens function for the two dimensional Laplace operator.298) can be shown by noticing that for r 6= 0.293) Here 2(x) = (x) (y) is the two dimensional delta function.16 G T00 therefore reduces to (use that in the x . y-coordinates. r.295) R . where g = e2 (r) .24 BLACK HOLES 45 where the last equation follows by symmetry (i.297) now has the solution (r) = .y @2 @ xy xy Rxy = @x@y .x @x . @x .293) is that Z p d2 x g = M: (1. @y + . y-coordinates g = e2 ) (1. (1.8GM log r: The metric (1.16 GM 2(x): It is easy to solve eq. After a few calculations we also nd @ . (1.298). .x .y .298). ) in the last equation is the ( at) Laplacian acting on an r. (1. R = .8GM (dr2 + r2d 2): Here r occuring in the power is measured in some arbitrary units.291) xy xy @y From these equations it now follows that R = Rtt + gxxRxx + gyy Ryy = e. without doing any calculations). ( px2+y2 ) r2 ( x2 + y2) q (1.296) (1.297) (1. Thus.

305) since the original standard angle is. which crosses one another in a single point. there are global gravitational e ects. 4GM ): (1.3: The surprising feature concerning the metric (1. The same applies to a source of light. but the physical angle in the original standard coordinates is just = =(1 . looking like at space! We can introduce a new radial variable r0 by (dr0)2 = r. 4GM ) (1. However. it can move along two straight lines.3 with = 1.301) is that it can be transformed to a simpler form. The two points are therefore the same point. a light ray coming from far away and going far away in the opposite direction corresponds to a change in polar angle 0 = .304) It is important to notice that although this metric looks at. there are no pointwise e ects of gravity. the angle 0 is restricted by 0 0 2 (1 . for 2 this type is called conical. 4GM )2 (r0)2d 2 = dt2 .4GM . In between these 0 2 . of course.46 GENERAL RELATIVITY Figure 1. This is illustrated in gure 1. The two points P and P 0 are in the same distance from the origin. if light originates in P /P 0.303) where we introduced the rescaled angle 0 by 0 = (1 . light follows a straight line. Due to the form (1. 4GM ) : (1.302) By this substitution we get d 2 = dt2 . There then appears a double picture of the source in P=P 0. In the r0 0 coordinates.8GM dr2 or r0 = r1. (1 . A space with a forbidden region of points.4GM =(1 . (dr0)2 . due to the fact that the variation of the angle 0 is restricted to be less than 2 . So an observer \can be in P and P 0" (in the following this point is called P=P 0). Thus.303) of the metric it is clear that in any point (t r 0). (r0)2 (d 0)2 (1. Thus there are no local gravitational elds. restricted to be between 0 and 2 . there is a forbidden region. (dr0)2 . In the conical coordinates.306) . 4GM ) : (1.

y.part of the metric is thus the same as the metric studied in the previous section. 2 R: The metric (1. From (1.292) we see that (1.1. in the next section we shall see that conical spaces can exist even in our three+one dimensional space. The calculation of the Ricci tensor R therefore proceeds like in the previous section.Tzz T = 0 for all other and : (1.292) that 1 Gxx = 0 Gyy = 0 Gzz = .311) R = e. (r) r2 (r) with r2 = @ 2 =@x2 + @ 2 =@y2: The conclusion is thus that the tensor G vanishes.314) is called a cosmic string.314) The constant has dimension energy/length. We shall later see that it is possible to . except for 1 (1.307) The x . 2 gtt R = 1 R: 2 Also.307) is thus consistent with an energy-momentum tensor of the form Ttt = . Using (1.307). It is a line along the z-axis. 1. since Ttt by de nition is positive. where we take the metric to be d 2 = dt2 . and is called the string tension. dz2 . The object described by the energy momentum tensor (1.310) (1. However. not directly possible to do an experiment testing the strange behaviour found above. it is easily shown from the results (1.308) that 1 (1. with support only in the point x = y = 0. of course.290). The pressure is negative. 2g R (1. 1 Ttt = . 2 R while all other components of G vanish. Thus it is.313) and (1.291) and (1.312) Gzz = .Tzz = 2(x) pg : (1. (1.26 Cosmic strings Let us now consider 3+1 dimensions.289) and (1. we get for the tensor 1 G = R . we take the non-vanishing components of T to be proportional to a twodimensional delta function.309) Gtt = Rtt . The metric tensor g having one or both indices equal to z is thus trivial.Gtt = .313) Furthermore. e (r) (dx2 + dy2) with r2 = x2 + y2: (1.24 BLACK HOLES 47 We reemphasize that this solution is valid in two space and one time dimensions.

.... 4G ) 0 = (1 ... . . ..... ... .. . . . . . .... . . .......... ... .. ...4: r0 .. . . .............. (r) = ....318) Here is the angular separation between the two pictures........ ..... .... .... . . .. .. ..... ... ...... . . ........................ .. ... . ........... . .. . .... . ...... ... 1.. ........... .. . ..... ..... ............ .... and l is the distance from the string to the source Q.... . ...... . .. ..... .... ... ......... .. .. .... ... . . . ... ..... .... . .. . . . . . .. . ..... . ............ ....... . ... . . . ...... ............. ... ...g... .... ............ . .......... . ........... ..... .... seen from an observer in \front" of the string.... ....... ...... . where light moves in straight lines. .. ... . In this case the cosmic string is a \defect". r.... . ......... . ..... . .. . .. .. .... . ............. .. .... ... 1.......... . ...... . . ...... ...... .... .. .... ... . .. .................. .. ...... ...... .. ... .. . .. ........ a quasar) bend towards the observer O. .. .... ... . ..... .. .. .......... ............ ... . ... .. . .... . ... ...... .... ...... .... . .. . . .. ... ....... ......... ....... ... ..... .. . ... This is shown in gures 1...... (dr0)2 . ...... ... ......... . ... .. .... ... . ... . ... . ... . ..... ..... . .. .. ... ....... .. ...... . . .. ..... ... ..... ...... .8G log r: (1................ .. we can now solve the Einstein equations.............. ....... .4G =(1 . ....... . . . ......... . ... ........... ......... .. . . ......... .. . ....... . ... .. ..... ... .. . .. .... . . ......... ......... . .. ..... ........ ... which appears . ................ .... .... In g. ... ..... .... .. ... i.. ..... ........ .. ....... . ........ .. ... ... Proceeding as in the previous section.. .. . ........... ... ..... . ......... ...... .... .... . .... ... ...5 we get tan( =2) = l tan( a)=(d + l) where a = 4G : (1...... .......... The conical space leads to the occurence of double pictures of a given light source placed \behind" the cosmic string. ..... ... .... . .. ...... ...... .. .. .. dz2 . ..... .................. ... ... .............. ... ...... ... . . .. .. ...... .. .. ... ....... . ... . ... . .. . . .. ....... ........ ... ..... ............ . .... .. .. 4G ) : (1. .... .. . ... . . ........ ..... ... . ... ....... 1.......... ............. .... ... ..... . ......317) We thus see that again there appears a conical space... .... ... ..... ..... .... .. . ... . ... .. .. .. .. ... ....... ..... .. . .. . .. . ....... .... .... . ..... giving rize to the negative pressure............ . . .. . .......... .. .. ....... .. .... .. .... .. . ..... .. .. . ..... ... ... . .. ......... ........... .. ........ .. ... ...... . . . .... .. .. .. (r0)2(d 0)2 (1...4 and 1.. . ........ .... .. ....... r O d S u l Q Figure 1... . .... . . and d is the distance from the observer to the string........ .... . .. .... .. . ....... .. .. ..... . .... .. ... .... ... .... .......... .... . .. an object which contains false vacuum.... ......... .. ... .4 we see the situation from the point of view of the (r )-coordinates.... . ...... dz2 . .... .... .. ....... ......... ........ ...... .. .... ..... ... ... .5. . From g.. ... ..... .... ............... .. . .. ........... .. .. .. .. . . .......... .. ... .. ... . .. ... .... .......48 GENERAL RELATIVITY .. . . ... 0 u Ou Su 2 a l Q l tg a Q u Figure 1.. . ..... ... . ... .. ... ...... .. ... .... . ...... ..... . ..... ..... ... ......... so the light rays emitted from the source Q (e.... .. ...... .. ...5: construct such an object in certain eld theories.... .315) The metric (1. .........5 we see the situation from the point of view of the conical coordinates (r0 0). .. .... .. . . e. . . ... .. ..... .... ....... .307) can now be written in the two coordinate systems d 2 = dt2 . .. . ... . .. ..... .. .......... where the angle 0 has a range which is less than 2 .. ..... .. .. .. .. .....316) where r0 and 0 are de ned by r0 = r1. .... .......... ... .. .. ...... ... . In g........ ...... .. ...... ..... .... .. . .......... . .. ........ ....................... ............... .... ... .. .. ..... ... .... .. ........ . .. ...8G (dr2 + r2d 2) = dt2 . .... ... . . ..

2MG (dt02 . and strong interactions equal). ' 8 G l=(d + l): (1. (r t) ! (r0 t0) where r r (1. However. the cosmic string does not have to be a straight line.27 Kruskal coordinates Let us return to the Schwarzschild solution for a point mass M . if an invariant was singular in one coordinate system.e.322) r02 + t02 2MG where T is a constant. an extended object. The metric is non-singular as long as r > 0. the metrics g and g'' do not vanish.e. there is no unambiguous observational evidence for cosmic strings. The string tension then has the value 1022g=cm 1018 tons=m ( 1018elephants=m) (1. A cosmic string is.321){(1. Of course. but can have the form of a curve moving around in space-time. T 2 . This is clearly within the possibilty for observation. aproximately 4 arc seconds.25 KRUSKAL COORDINATES 49 to be in two points in the conical coordinates.228) then becomes 3 3 r (1. which corresponds to r = 0. t02 = T 2 2MG . Such a system is the Kruskal coordinates. and hence should be observed by a concentration of many double pictures along the string.323) is thus singularity free. The Schwarzschild solution (1. for 0 < t0 < T the metric is well de ned as a function of r0. leading to 1030 GeV2.321). Thus. T .319) this leads to an angular separation of order 2 10. it would be singular in all other systems.318). but consists of two sheets r0 > 0 and r0 < 0 joined smoothly at the branch point r0 = 0. In practice the quantity 4G is small relative to one. i. .5. which corresponds to the Schwarzschild radius r = 2GM . 1. However. which has an energy scale of order 1015 GeV. there is no singularity at r0 = t0 . The apparent singularity at r = 2MG is not really a singularity. For t0 ! T p the metric has a singularity at r0 = t02 . r2(d 2 + sin2 d'2) 2 rT where it is understood that r is a function of r02 . electromagnetic.319) A typical cosmic string is generated in the \Grand Uni cation" (characterized by having the strengths of weak.320) in ordinary units. t02 given by the inverse of the functional relation (1. but there exist some pictures of unusual concentration of double pictures along rather crumbled curves. and hence we can expand eq. but these points are to be identi ed. so one can continue d 2 down to negative values for r0 ! The metric (1. So far (1996). i. as long as r02 > t02 . as is indicated by the fact that the invariants R and R R do not exhibit any singularity at the Schwarzschild radius. Also. this does not happen with any invariant. so this indicates that it should be possible to nd a system of coordinates without the Schwarzschild \singularity".322) provides us with a metric. It is seen that even when r0 = 0. Thus we see that the coordinate transformation (1. 1 e 2MG 2r0t0 = tanh t (1. (1. From (1. dr02) . of course.1.323) d 2 = 32G M e. which itself cannot be observed directly. These pictures are candidates for cosmic strings.321) r02 .

133){(1.325) gij = g = r2 2 X @yk @yk k=1 @xi @xj (1. like freely falling elevators. and one should always ask questions which relate to invariants. not globally.R = .328) .R1221 = R2121 R1111 = R1122 = R2211 = R2222 = 0 (1. (1. where ds2 = g11 dx2 + 2g12 dx1 dx2 + g22dx2 1 2 where gij are de ned just like in eq.g. and this points to the fact that in general relativity the coordinates are arbitrary.R2112 = . and turn to the application of this theory to cosmology. of independent components to only one.R = R following from eqs. etc.28 Appendix: Some remarks on geometry In our discussion of Einstein's general relativity. we have often used geometric language (metric. In general it satis es the antisymmetry properties R = . Here the indices take only the values 1 and 2. or it is due to the fact that space-time is \curved". The di erence is rather philosophical. etc. one could interpret this by saying that this is due to the suns gravitational eld which in uences the light-ray. etc. (1. since in any case space is \curved" because of the gravitational eld.50 GENERAL RELATIVITY which does not see any singularity at r = 2GM . The others are then given by (1. To see this. e.135). 1.14).).324) Now the fundamental point of non-Euclidean geometry is that this coordinate system can only be de ned locally. and (1.327) restricts the no. R1212 . when a light-ray bends passing the sun. there exists other systems which cover all space. The curvature tensor has a simple interpretation in two-dimensional geometry. Thus. namely x1 x2 . However. curvature.327) Let us consider the curvature tensor in two dimensions. (1. Historically the general theory of relativity was indeed based on the idea that local inertial systems (\elevators") are analogous to local Euclidean coordinate systems (introduced in non-Euclidean geometry). let us de ne a length between two points (y1 + dy y2 + dy2) and (y1 y2) by the law of Pythagoras. 2 2 ds2 = dy1 + dy2 (1. With these remarks we end our discussion of general relativity.326) For a sphere with radius r we have the coordinates (r ) and g'=0 g'' = r2 sin2 (1.327) as R1212 = . although at the same time we have used physical language.

R=2 is the Gaussian curvature (in two dimensions). Eisenhart. g12 . and vice versa.P. A sphere with r large. so the Ricci tensor is R = g R1212 g and hence the curvature scalar is (1. the larger is jRj and vice versa. Thus. In higher dimensions the curvature tensor has of course more independent components. R = . Now we see that the smaller the radius of the sphere is. g g ) R1212 g 51 R (1. g g ) 2 Using the form (1.1. N.333) 2 g The quantity .329) 2 with g = g11g22 . R1212 = r12 (1.330) (1.26 APPENDIX: GEOMETRY which can be summarized by = (g g .327) that . We refer the reader to the mathematical literature. (1. Princeton.J. (1. for example L.329) we get the full curvature tensor in terms of R.124) for R1212 one nds for the sphere (1. curves little. 1926).331) Inserting this in eq. jRj measures curvature in a natural way. R = 2 R1212 g .332) R = 1 R(g g . Riemannian Geometry (Princeton University Press..

52 GENERAL RELATIVITY .

If we wish to describe the solar system. 1013M .Chapter 2 Cosmology 2. In principle one could measure the energy-momentum tensor T (x) in all points of the universe and then solve the Einstein equations. the cosmological principle should be considered as a rst aid for interpreting the observations. which on a very large scale may be considered to be particles. so we have also clusters of galaxies 2. there is no need to bother about the properties of the universe at intergalactic scales. Various mythologies and religions have several scenarios to o er. and hence for everybody else. In physics it is always important to x a relevant scale. Then we can consider even the galaxies 1 to be \point particles". 53 . astronomy does certainly not provide data enough for obtaining T (x).1 The cosmological problem Speculations about the nature of the universe form an integral part of the history of mankind. This situation is somewhat similar to considering a gas consisting of a large number of particles. Every point is equivalent to any other point in the universe. we then assume that nobody else has this honor. This amount to \building up" the universe from \particles" with diameter 108 . The extent to which the cosmological principle is right is not precisely known. on the other hand. 1 With masses 1011 . In modern physics the canonical scenario is based on astronomical observations described in terms of Einstein's general relativity. which when viewed at a large scale are also particles. However. 3 With masses 1014 . 109 light years. we wish to understand the large scale structure of the universe. 1012M . 2 With masses 1012 . Therefore. It therefore also follows that in the large scale view of the universe we cannot assign any \center". rather than as providing an extremely precise description of the universe. where we do not keep track of the individual particles. the universe appears to be approximately isotropic) around any point. It means that the universe is assumed to be homogeneous and isotropic (because for us. The idea that all points in our (large scale) universe are equivalent is called the cosmological principle. but rather consider the large scale statistical properties. If. we do not need to be concerned with small details like our solar system. One can say that since it turned out that we do not live in the center of the universe. There are also clusters of clusters of galaxies 3 (\superclusters"). 1015M . This means a radical deviation from the medieval idea that the earth is the center of the universe. Now it turns out that galaxies tend to \cluster".

For example. the matter of the universe is at rest in three dimensional space (just like in the formulation of the cosmological principle. this means in reality that on the average the matter of the universe is at rest in the three dimensional space). The line element thus has the form d 2 xi = dt2 . when the universe evolves. i. the main point being that the cosmic time t increases. The existence of t is taken to mean that at a given xed value of t.e. The coordinates x1 x2 x3 remain constant along any geodesic perpendicular to the initial surface So. as well as the assumed existence of a \cosmic standard time".2 The cosmological standard model The cosmological \standard model" (not to be confused with the \standard model" in particle physics) is based on the cosmological principle. so in general we need g01 = g02 = g03 = 0 on the hypersurface. and hence g00 = . d2x + . dx dx = 0 d2 d d . the matter of the universe \follows" this increasing t. a homogeneous and isotropic universe.1) (2. As time t increases. Instead of temperature one could use other scalars. We can describe this co-moving system by means of so-called Gaussian coordinates. gij dxidxj 2 S0 (i j = 1 2 3) (2. and the cosmic standard time could then be taken as a function of the temperature which increases monotonically as the temperature decreases. This time is intimately connected to the evolution of the universe. the temperature is believed to decrease as the universe evolves. where we consider a geodesic along which time increases (see g. This coordinate system is therefore called the co-moving system.54 COSMOLOGY Figure 2. Along such a geodesic we have d 2 = dt2.2) Consider the equation for a geodesic. S0 and S are three dimensional spaces (hypersurfaces) 2. 2. In order that the t-lines are orthogonal to the 3-dimensional hypersurface it is necessary that any four vector (0 a b c) is orthogonal to the vector (1 0 0 0) which is tangent to the t-line.1).1: Gaussian coordinates for a co-moving system.1.

3) @t @xi and since g00 = .5) r .27) that 2 @g0i = @g00 (2. This implies from (1. even though it is rotational invariant. The galaxies do move. we deduce that . Thus. This means that in (2.1) implies that there cannot be a drdt-term in eq. (2. Thus the metric for a co-moving system has the form d 2 = dt2 . and dr2 + r2 (d 2 +sin2 d'2). Thus eq. (2.e. The requirement of isotropy in this point means that the line element has the form d 2 = dt2 . are (on the average) at rest. remain zero on all other surfaces. f1(t r1) = f1(t r2) F (r1 r2) Taking r2 to be xed.1 on any geodesic. (2. Eq. then the ratio of the line elements f1(t r1)dr2 and f1(t r2 )dr2 must remain the same at all times.e. Since we are using co-moving coordinates. this implies f1 (t r) = f (t) L(r) (2. and the only thing which can happen is that the overall scale changes.4) over all space. and ' remain the same. so the coordinates remain the same. The functions f1 and f2 only depend on t and r. the only thing that happens is that distances can be scaled. i. we can construct the three-dimensional space coordinates by putting numbers (x y z) on each galaxy. Thus. the elements g01 g02 g03 of the metric tensor. Now let us consider an arbitrary but xed point in space. Thus. but the distance between the galaxies move too. f1 (t r1)=f1(t r2 ) is independent of time and equal some function F (r1 r2). i. we consider two points r1 1 '1. clusters of galaxies etc. f1 (t r)dr2 .5) since the only rotational invariant at this point are t r dr2 . eq.2 THE STANDARD MODEL 55 When d 2 = dt2 and the coordinates x1 x2 x3 are constants along the geodesic.2. (2. If. the galaxies are dragging along the coordinate mesh. gij (t x1 x2 x3)dxi dxj (2. since only the overall scale can change.5) is the most general form of the line element. we obtain @g0i = 0 @t along the geodesic t-lines.5). which are zero on the initial surface S0 . As time increases the coordinates remain the same but all distances can be scaled with a time-dependent scale factor. at a given but arbitrary time. so each galaxy keeps the same coordinates. since a and/or '-dependence of f1 and f2 would destroy rotational invariance.i00 = 0 for i = 1 2 3. Now we must ensure that as time goes by. f2(t r)r2(d 2 + sin2 d'2) (2. and r2 2 '2 with d = d' = 0 at both points.6) . around which we take spherical coordinates r '. The picture of the development of the universe presented above is thus that at a given cosmic time t the galaxies.4) is thus a consequence of the assumed existence of a cosmic standard time.

g. Z 0 ! 2 !2 ! dt . g(t)r2 @ d 2 + sin2 d 4 d d d _ . (2. Introducing r slightly simpler form for the line element d 2 = dt2 . however. dr = d' = 0 and d 6= 0 at the two points).9) After straightforward calculations we get: . We can now compute the Christo el symbols using the variational principle discussed in section 1. Thus we are led to the line element d 2 = dt2 . g f_ _ Rtr = rg . t = . The same can be shown for the function f2 (take e.trr = fL 2 ! 13 d' 2 A5 = 0 d (2.12) .17. f (t)L(r) dr 2 . g(t) r2(d 2 + sin2 d'2) (2. Later on we have to ensure that this choice of an arbitrary origin agrees with the assumption of a homogeneous space.'' = .e. . 1 .' = . . f (t) L(r) dr2 .10) The Ricci tensor can now be obtained from @.r = . f1 factorizes.8) we selected an arbitrary but xed point as the origin of the spherical coordinates. fL sin2 '' . @@x g @x 2 For reasons which shall soon be clear.8) now represents a line element which is consistent with the standard cosmological model.7) This line element clearly has the property that the coordinates remainq same. It should be remembered. rf (2.56 COSMOLOGY i. g(t) H (r)r2(d 2 + sin2 d'2 ) (2.r = . that in eq.' = .' = 1 r' 'r r .r = 2L rr _ . we start by computing Rtr . Eq.11) (2. (2.8) where we have dropped the prime on r. r = 1 r .'' = . 1 @2 ln R = 2 @x ln g .' = cot ' . sin cos .' = 2g_g t' 't (2.t = 1 r 2 g 2 _ rg . @x + .r = 2ff rt tr rg .r = . and the the 0 = H (r)r we get the only thing which happens is that lengths are scaled.r = .t = 2g_g . f (t) L(r) dr2 .t'' = 2 r2 sin2 g _ L0 . fL 1 .

4ff 2 1 . and the Ricci tensor is more easily evaluated. 3f 2 4f R = 0 for 6= (2.r becomes time independent. 2 g T ) (2.15) so we get g = f (up to a constant which can always be absorbed by trivial rede nitions).18) where prime denotes the derivative with respect to r. .). T = p g + (p + )U U (2. Lf 4f 0 r2 2 _2 1 rL R = . G14).13) The energy-momentum tensor has the form of a perfect uid (see sect. we can easily construct the mixed Ricci tensor. Hence from (2. 2ff .14) where U is given in co-moving coordinates by Ut = 1 Ui = 0 (2. Thus Rtr must vanish. since this means an isotropic and homogeneous universe if the pressure p and the density are only functions of t.17) Inserting f = g the Christo el symbols simplify somewhat (for example. L0 1 _2 R = r2f R = r2 1 + r21 . r4 ff R'' = sin2 R _2 3f Rtt = 2f .2. Since the metric is diagonal. 1 f . L2f .1 + L .19) . 2 f . rfL2 . since Ttr and gtr both vanish. etc. The Einstein equations read 1 R = . i. _2 L0 Rrr = .8) The last equation follows because matter is at rest.16) g f h i d 2 = dt2 . 2L2 . _2 L0 1 Rr r = fL Rrr = .e. f (t) L(r)dr2 + r2(d 2 + sin2 d'2) (2. 2f + 3f 2 4f R = 0 for 6= (2. 2rfL2 .12) g = f_ _ (2. 4 f 2 f fL 2f f 1 R'' = r2f sin2 R'' = R _2 3f Rtt = .8 G(T . rL .2 THE STANDARD MODEL 57 where the dot indicates the time derivative. The line element is thus from (2.

23) (2. (2. ij 3 Ri j 3R 3 . This also follows from the energy-momentum tensor (2.k indicates lm that these quantities are computed in 3 dimensions (time xed!). but has to be imposed by hand.58 The curvature scalar is then trivially obtained. (2.e. 6k (2.l kj .k ij 3.21) rL2 where k is a constant.20) R = .k li 3 . since in the next section we shall derive the result (2.22) r2 r2L 2rL2 The constants on the right-hand side have to be the same because of Rrr = R . since if this is not the case there would be one (or more) preferred direction. This is due to the fact that originally we imposed isotropy relative to a de nite point.22) we get 1 . 3ff .19) we see that in general this is not true. 3 . a(t)2 d (2.2k=a2 for Rrr .27) from simple considerations. (2. kr2 where we have replaced f (t) by a(t)2 for reasons which will be clear later. We also need to ensure that R is homogeneous in space. 1 + L0 = 2k (2.14). L = c .21) is trivially solved. r2f 1 .24) 1 . of course. We can repair this by requiring (see Rrr ) L0 = 2k (2. Inserting this in eq. and (see R ) 1 . COSMOLOGY where c is a constant. L . # " 2 dr + r2(d 2 + sin2 d'2) 2 = dt2 . From eq.26) a(t)2 Readers who are not familiar with di erential geometry in three dimensions can skip these remarks.19) we see that the space part of the Ricci tensor.l kl = 3gik 3Rkj 2 1 2L0 (2. rfL2 Now isotropy clearly requires that Rrr = R = R''. Eq.k = @ @xj ki .25) is given by . (2. Again consulting (2. Thus the Ricci scalar in 3 dimensions is 4 3 R = .k 3 .L0 =ra2L2 = . Here the 3 in front of Rij and . @ @xk ij + 3.19) we see that Rrr = R is a non-trivial requirement. so it must not depend on any other variable but time.e. 4 . i. c + 2k = 2 k r2 i.1kr2 (2. Thus we have found the explicit form of the spatial dependence of the line element. c = 1. Consulting eqs.

Let us consider three-space as embedded in a ctitious four dimensional space.24). It is of interest that the three-dimensional part of the metric (2.27) This means that a(t) is the radius of curvature.30) (2.)~ 2 ) x 2 x Introducing polar coordinates we can write this as 2 2 ds2 = dr2 + r2(d 2 + sin2 d'2) + Rr2 dr r2 . we obtain with = r=R R2 dr2 ds2 = R2 . (2. which is called the Robertson-Walker metric.29) (2. so the element of length ds is given by Pythagoras' theorem.3 A GEOMETRIC INTERPRETATION The Gaussian curvature K in 3 dimensions is given by 3R = .28) (2.31) ~ 2 + z 2 = R2 x where R is the radius.3 A geometric interpretation of the RobertsonWalker metric The metric in co-moving coordinates has a trivial time part. The metric is assumed to be just Euclidean. where the fourth coordinate is called z. x ds2 = (d~ )2 + (d~z2) x 4 In this relation z can be eliminated by use of (2. Di erentiating (2.33) ds2 " 1.6 3K . This suggests that the metric (2.24) has a simple geometric interpretation. whereas the space part is non-trivial. has a geometric interpretation which we shall discuss in the next section.d(~ 2 ) x so we can eliminate dz2 from (2.2. d 2 2 + 2 (d 2 + sin2 d'2) # (2.e. If we measure the distance r relative to the radius R. (2.29) we get dz2 = 2z dz = .32) i. r2 + r2(d 2 + sin2 d'2) = R2 (2.28). so 3K = k a(t)2 59 (2.29) 2 2 ( x2 2 ds2 = (d~ )2 + 4(Rd~.34) . ds2 = (d~ )2 + (dz)2 x Let us consider a hypersphere. 2.

which are determined at a certain time.29) it is x clear that ~ 2 cannot exceed R. again with a x time-dependent scale factor a(t).4 Hubble's law We shall now investigate the frequency of light emitted by some distant object in a universe described by the Robertson-Walker metric (2. dz2 x (2. In this case R does not have the interpretation as the radius of the universe. The existence of a cosmic standard time (which is assumed in the standard model) means that we can choose t = 0 in such a way that all particles are at rest at that time.29) can then be regarded as a balloon.1=R2 . In cosmology we can use the galaxies as the points dragging along the coordinate mesh. where at a given time the galaxies are marked by dots on the three-surface. but the distance between the dots will move too. at a xed time this agrees with (2. and is given by .24) with k = +1. One can picture comoving coordinates the following way: Let us imagine that space is lled with a dense cloud (\ideal uid") of freely falling particles. As the balloon is in ated the dots move. (2. and genuinely non-Euclidean (Gauss-Bolyai-Lobachevski geometry). we cannot use the hypersphere (2. say t = 0.29).1. Instead we can consider the surface . the three-dimensional projection of the hypersphere (2. The curvature is everywhere negative. and is given a xed set of spatial coordinates. in eq. Proceeding as before.24) is negative. ~ 2 + z 2 = R2 x ds2 = (d~ )2 .60 COSMOLOGY If. The co-moving trajectories are then just the paths of the galaxies.34) is always less than one. The space-time coordinate t ~ of any event are de ned by taking ~ to be the x x coordinate of the particle which is just passing by where the event occur.24) dr2 (2. in agreement with intuition). Thus. we put a(t) = R. In this p picture one can say that R is the radius of the universe.36) dt2 = a(t)2 1 . If k in (2. If we do that.34) is precisely the space part of (2. Alternatively one can say that a coordinate mesh is being dragged along by the cloud of particles. for a xed time. The curvature of the hypersphere is always 1=R2 (large radius means small curvature and vice versa. 2. The case k = 0 corresponds to a completely at space with ds2 = d~ 2 .4) that this remains true at all times. then we see that the metric (2.1){(2. one obtains the Robertson-Walker metric (2.35) which is called a pseudo-hypersphere. with time de ned by clocks xed on the mesh. and t as the time read o from the particles clock. so each dot keeps the same coordinates. In this case space is obviously unlimited. and is therefore called the cosmic scale factor. Each particle carries a small clock. since from (2. The interpretation of the Robertson-Walker metric given above has a particularly nice interpretation in terms of the framework of co-moving coordinates. but it still sets the scale. Since light has d = 0 we get for radial motion from (2. Thus. kr2 .27).24) with k = .24). It then follows as in (2.

14 sec) a(t) (i. For increasing t0 . in the rst approximation t0 .39) a(t0 ) a(t1 ) where we used the rather safe approximation a(t + 10. t1 is su ciently small. 1 = a(t0 ) .37) to the frequency of light. and arriving at our galaxy r = 0 at time t = t0. Eq. Thus r1 is time independent. then if z > 0. Hence we have the result that an expanding (contracting) universe is characterized by a red shift (blue shift) of the spectral lines. (2. i. 8 . Suppose that the emitting galaxy is relatively close to us. t ) _ (2. Conventionally this relations is expressed in terms of the red-shift parameter z. (2.e. so let the periods be t0 and t1 as measured by us and on the emitting galaxy.1 r1 (k = .4 HUBBLE'S LAW 61 From this we get for a light wave leaving a galaxy at r = r1 at time t = t1 . 0 > 1 (red shift) or if z < 0 (blue shift).41) and obtain to lowest order a(t0) (t . (2. and where we used 0= 1 = 1 = o.. which to lowest order implies that (t0 . t1)=a(t0 ) r1.e. a(t1) (2.39) then gives for the frequencies a(t ) = t1 = a(t1 ) (2. We wish to relate (2.38) t1 + t1 a(t) t1 a(t) so using that t0 is very small (10. ) z = 0 . kr2 : sinh. t1 is the distance measured by an astronomer (remember the velocity of light is one).1 > sin r1 (k = +1) Z t0 dt Z r1 dr < p (k = 0) = = > r1 0 t1 a(t) 1 .37) In co-moving coordinates galaxies have constant coordinates (r '). as is clear intuitively as well as from eq. we get t0 = t1 (2. respectively. Obviously t0 > t1 in both cases.40) t0 0 This shows that the wave length shifts like the scale factor: The wave length is stretched or shortened at the same rate as the universe expands or contracts. the farther is the object.14 sec !). the universe evolves very little in 10. t1 we expect that a(t0 ) will increase (decrease) more and more. only the scale changes. Thus Z t0 + t0 dt Z t0 dt = (2.14 sec for a typical light signal). and hence z will increase (decrease) more and more.41) we see that since o is the light observed by us after a long travel from the galaxy at r = r1.1) (2. and for z .2. Now.42) a(t0 ) 0 1 where t0 .41) a(t1 ) t1 ) 1 1 0 = 0 1 where is the wave length. From eq. 1 = a(t0a(. it follows that a(t0 ) > a(t1 ) (red shift) or a(t0 ) < a(t1 ) (blue shift). so the larger (smaller) the value of z is. We can then expand eq. (2. since we observe the light later than it was emitted.37).

p)a2 aa + 2_ a (2.41). From (2. At present it is not too well known.4 G( + 3p)a (2. If L is not small.50) . 1 (p . _ eq. 1 (3p + ) 2 1 T r r .24) we know that a(t0)r1 is the physical distance for r1 small (relative to the scale parameter). 1 (1 .15). (2. a . t1 )=a(t0 ) r.43) is the famous Hubble law (1929). Such an interpretation should. Let us notice that for r1 small it follows from the Robertson-Walker line element that a(t0 )r1 is approximately the radial velocity r of the galaxy.43) z H0L H0 = a(t0) a(t0) where L is the distance measured by the astronomer. we shall return to the Einstein equations following from eqs. Using the results (2. and (2.3 a a _2 (2. 2:5) 1010years (2. 2 a2 . by use of the co-moving four velocity (2.42) can be written _ (2.47). The relation (2.49) 2 + 2k = 4 G( . ) (2. and in a sense it marks the beginning of modern cosmology. as the derivation of eq.62 COSMOLOGY the metric (2. which increases with the distance of the galaxy.45) H0 Hubble's discovery of \a roughly linear relation between velocities and distances" had a profound e ect on cosmology. 2 T = .46) T t t . we get with the Robertson-Walker metric 1 (2.5 Einstein's equations and the Robertson-Walker metric So far the scale factor a(t) has not been determined as a function of cosmic time.48) Rr r = R = R'' = .14). since light is certainly in uenced by the gravitational elds. To do that. 2 T = .44) This may be interpreted as a frequency shift due to the Doppler e ect. Observationally it is found that there is a red shift. (2. (2.47) 2 with the and '' components given by the same expression as (2. however. and since (t0 .19). (2.13).14).21) or (2. one needs to go back to the relation (2. Thus (2.43) shows. not be taken too seriously. The Hubble \constant" H0 actually depends on time.43) can be written z vr H0L (2. 2.22) for the function L(r) we then get (f (t) = a(t)2) Rt t = . 2k a a a2 so the Einstein equations read 3a = .

2) that . not on space). dp + pg dt pg(p + )] = 0 (2. (2. inserting (2. (2.56) If one has extreme relativistic particles (like photons) the universe is radiation dominated 1 with p = 3 . and M = a3 is proportional to the mass of the universe. since dM is the change in energy balanced by pdV .54) can also be written d ( a3 ) = .14) is covariantly conserved.53) 1 In eq. Eq.51) In co-moving coordinates we have from (2. (2. and we get i dh (2. The reason we 1 only have two independent equations is that T = 0 is already contained in (R .55) da If an equation of state p = p( ) is given. and the Friedmann equation (2.88) this means @p 1 @ T = @x g + pg @x pg( + p)U U ] + . Eq.58) which is called the Friedmann equation.55) can then be written (multiply by da) dM + pdV = 0. Using eq.51) is then trivial for = r ' (remember that due to the homogeneous space p and can only depend on time. which expresses energy balance during the evolution. Thus. the energy conservation (2. 5 . kr2 (2. For example.57) / a(t).55). and the evolution can then be computed from (2.50) we get a2 + k = 8 3G a2 _ (2. (2.5 EINSTEIN'S EQUATIONS 63 We also have to ensure that the energy-momentum tensor (2.50) can be simpli ed by eliminating a.3pa2 (2. eq.4 (p = 3 ) We notice that the Einstein equations (2. 2 g R) = 0. if the pressure is negligible. (2. (2.54) a(t)3 dp(t) = dt a(t)3 (p(t) + (t)) dt Eq.58). and 1 (2.49) and (2. (2. The last two equations are easy to remember: the quantity a3 is proportional to the \volume" of the universe (for k = +1).55) and (2.49) and (2. p .49) in (2.52) only the time dependence of g is relevant. (1. and we have 1 d .50) are implied by eqs.55) can be solved for = (a).00 = 0.58).52) dt The determinant for the Robertson-Walker metric is given by 4 sin2 g = a(t)6 r . then / a(t).2.3 (p ) (2. 5 The functional equations in the standard model are thus the equation of state p = p( ). (p + )U U = 0 (2.49) and (2.50). which is the work done It is easy to show that eqs.

3pa2 (2. It is seen that sometime in the past. in a plot of a(t) versus t the velocity a is positive.64 COSMOLOGY against the pressure forces. This singularity is the famous big bang.62) which is a consequence of eqs.61).R.61) da It is also convenient to use eq. (2. 1973. and a(t) increases until it reaches its present value.F. the second term (G mass of the universe (1=a)) is the Newtonian potential. a 2=8 G .2.43) is positive. See e. since the three-dimensional co-moving space shrinks to a single point as t ! 0. The occurrence of the singularity means that the universe starts as an explosion. G 4 a3 1 = . Hawking and G.49).58). so a=a > 0. We can now qualitatively derive the following results: If p > 0.1) and vanishes for k = 0.6 The big bang If the pressure p is positive. k (2. _ _ but the acceleration a is negative.4 G( + 3p)a (2. Cambridge Univ. Clearly a vanishing scale factor means that the Robertson-Walker metric becomes singular. (2. Conventionally we take this time to be t = 0. 6 so it is not just a feature of the Robertson-Walker metric. i. Thus.g. which is negative (positive) for k = +1(k = .62) that a=a < 0. many features of the standard model can be understood without speci cation of the equation of state p = p( ). we need to use the Friedmann equation (2. To do this. Ellis. 2 ! 1 a2 . Press. and we have a situation like the one shown in g. Since we observe red shifts (and not blue shifts) it follows from the discussion in section 4 that at present the Hubble \constant" (2. d ( a3 ) = . the function a(t) has reached zero.1 Existence of the big bang (the initial singularity) . 6 2. k _ (2.58) has an \almost Newtonian" interpretation: multiply by 1 . S. The Friedmann equation (2.6.The large-scale structure of space-time. it follows from (2. 2.60) a 3 a2 and the energy conservation (2.W.e. i. Such singularities are known to occur generically in most space-times under very general assumptions. The sum of the \kinetic energy" and the \potential energy" is then the constant total energy.e.60) and (2. 2. 3a = .59) 2_ 3 a 2 and the rst term is the Newtonian kinetic energy of a test particle with unit mass and with coordinate proportional to the scale factor. These considerations should be taken purely as an aid for the memory.55).

45) we see that the upper limit on the age of the universe is between 10 and 25 billion years. which put lower bounds on the age of the universe in a way which is independent of the big bang model.3 . also follows quite simply from a < 0. Since a2 ! 0 for a ! 1.2: Plot of a(t) with a > 0 and a < 0 _ The history of the present universe is thus that it starts with a = 0 and a(t) increases.65) t0 = a _ a(t ) _ _ where H (t0) is the Hubble \constant" at present (see eq. a = 8 3G a2 . We shall return to this point later. Hence it follows from (2. until it reaches its present value a(to ) where t0 is the age of the universe.3 Discussion of the fate of the universe Having looked at the past. stellar evolution and the nuclear abundances. Suppose a(t) continues to grow and ultimately a(t) ! 1.6 THE BIG BANG 65 Figure 2. it follows from _ (2. it follows that a2 ! . let us now turn to the future.6. 2. From this we obtain for the age of the universe Z a(t0 ) da Z a(t0 ) da a(t0 ) 1 < = a(t ) = H (t ) (2.43)). From eq. (2. From eq.63) (2.2. Thus also decreases with increasing a.k (at least _ . _ const. k _ a3 Thus we have that a is smaller today than it was in the past _ 2.64) a(t0) < a(t) _ _ 0 t < t0 This.6.2 The age of the universe according to big bang s (2. 25) 109 yr (2.60) that a=a decreases with a. since there are arguments based on geology. (2. Since decreases at least like a.66) 0 0 0 0 0 This limit is actually an important feature of the big bang model. of course.61) we see that for p 0 the quantity a3 decreases or is a constant (for p = 0) when a(t) increases. t0 < (10 . (2.60) that a is a decreasing function of a.

60) and (2.68) a(t) ! 1 slower than t for t ! 1 and k = 0 For k = +1 it is impossible for a(t) to go to in nity. (2. The 1=a4 behavior is related to the fact that relativistically the energy in a co-moving volume \red shifts" like 1=a. 3 and for galaxies observed now 2=c2 1. and for non-relativistic matter relevant for the present universe. the scale parameter will then drop down again (with a negative and _ blue-shifted spectral lines) and reach a(t) = 0 in a nite time t. For k = . or whether we shall be heading for the disastrous future singularity. For the bene t of the impatient reader we mention that it turns out to be di cult to make predictions. Since a max is always negative. while the volume still scales as 1=a3.61) gives with p = 0 =B (2.61) are incomplete without the speci cation of an equation of state p = p( ). It therefore would be quite interesting to know the sign of k to see whether the universe expands forever. and from (2.60) we see that a(t) _ becomes zero when a(t) has its maximum value amax.1 or k = 0. where a2 = 3=8 G. This is the \closed universe". especially about the future. and hence p=c2 = 1 2=c2. Thus it is only possible for a(t) to go to in nity if k = .1 For k = 0 we have similarly for the \ at universe" (2. . Eq. For a relativistic ideal gas one has (2. In the standard model two such equations are used. (2.69) means that the energy content of a co-moving volume a3 remains constant when a(t) changes.67) (2. We shall discuss this in the next section. Reintroducing the velocity of light. Most of the matter (galaxies and clusters of galaxies) observed today appear to be non-relativistic with kinetic energy much smaller than theprest energy.7 Fitting parameters to observations The general relativistic equations (2. Hence eq.61) gives (2. where 2 is the root-mean-square average velocity of the gas molecules (= the galaxies).1 we thus get for the \open universe" a(t) ! t for t ! 1 and k = . In the early universe one usually assumes the existence of a hot gas of extreme relativistic particles with the kinetic energy much larger than the rest energy.66 COSMOLOGY as fast as 1=a.70) p= 1 3 and eq. (2. namely for relativistic matter relevant for the early universe. the pressure in the energy-momentum tensor enters like p=c2.71) =D a4 where D is a constant of integration.69) a3 where B is a constant of integration. 2. since a2 ! 0 at least like 1=a). In the kinetic 1 theory of gases (here applied to galaxies) p = 3 2.

and hence the fate of the universe. Since crit is given in terms of the Hubble constant. and k = 0. 0:1 . even though there is essentially no luminous matter to observe.74) crit. the Friedmann equation (2. This indicate that a large part of all matter is \dark". (2. If one considers luminous matter (stars) one nds 7 lum. so > 1 = 1 < 1 correspond to k = +1 k = 0 k = .73) today.2. at. studies of orbits of stars in spiral galaxies (for elliptic galaxies similar results do not exist. By studying the orbital motion of stars in a galaxy at the radius where light has essentially disappeared (this radius is called the Holmberg radius). For galaxies one gets dyn. 2. if is very small. In eq. Ann. we get 2 . If = crit.72) a 3 a2 3 crit. but instead stays approximately constant. ' 0:01 so 0:01. However. then the expansion is driven entirely by . Rev. closed universe for k = .29 3 (2.75) allows a determination of the sign of k. and as the universe expands and cools.1. (2. and since could in principle be determined by \measuring" the mass density in the universe. (2. respectively. This is called \ at rotation curves". = crit. respectively) let us use eq.Faber and J. Astrophys.77) dyn. the kinetic energy of the matter redshifts and ultimately it becomes non-relativistic. 0:3 7 S. From h H 1010 yr 1 h 0:4 (2. with a lot of gravitational attraction. (2.76) lum. . and the universe is open. In order to t the crucial parameter k (open. 1) = H 2( . i. 1) (2. and it means that the mass M increases linearly with r.7 FITTING PARAMETERS 67 Now at very early times matter was very hot. 17 (1979) 135.76). Then we have a lot of energy.60) in the form a 2 = H2 = 8 G . Astron. which will counteract the expansion making a closed universe. (2. k 8 G _ (2. stars outside luminous matter) beyond the Holmberg radius p have revealed the interesting and stunning result that does not go as 1= r. We shall try to t the various parameters to observations. one can also estimate the masses of dynamical systems by asking for the gravitational forces necessary to explain the observed motions. Eq.75 has a simple intuitive appeal: Suppose is very large. where M is the mass interior to the orbit of an object (a star) with velocity and orbital radius r. one can measure the mass of the luminous material and one gets (2.1 k = 0 k = +1. Gallagher. there is very little counteraction to expansion.e.72) H i the value of the Hubble \constant" at time t. However. 2h 10 g=cm Introducing = = crit eq. For a system with spherical symmetry one can use Kepler's third law GM=r = 2. perhaps because of the lack of \test bodies".75) a2 3 crit.72) becomes k =8 G ( . Similarly. = crit.

We shall. To describe the situation in very simple terms let us consider photons in thermal equilibrium. If is very small. From geology. The discrepancy between (2. as one can see qualitatively. 3 today. using this and (2.8 The cosmic microwave radiation background In the standard big bang model the universe starts out with a very small scale factor. but it should still ll space around us. slowing down the expansion.81).68 COSMOLOGY so 0:1 .81) a(t0 ) 1 Thus. . the temperature dropped. a2 . Finally we shall determine the parameter D in eq. since the absence of free charged particles means that the interactions of photons with other matter drop o steeply. (2. 2. and we cannot say whether the universe will eventually re-contract or expand forever. and evolution of nuclear abundances one can make arguments which puts a lower limit of 10 billion years on the age of the universe. To get a bound on D. (2. 3 0:9 crit. if t1 is the time of the decoupling. Now is related to the age of the universe. no counteraction from gravity). however.78) Thus. Also. From eq. and the thermal contact between matter and radiation was broken. This turns out to give the bound 8 <3 (2.9 we shall discuss this quantitatively. we need to consider the microwave background radiation.81). As the universe expanded. (2. and thus the life time of the universe is small. and the universe is lled with highly relativistic particles. 0:3. We found in eq.80) crit.79) Consequently we see that we are unfortunately not (yet!) in a position to determine whether the universe is closed ( > 1). The radiation existing at that time has since been much redshifted. (2.71).75) and the bounds (2. or open ( < 1). (2.82) In section 2. we have a lot of gravitational interaction.76) and (2. the free electrons joined the nuclei to bind into neutral atoms.79) we get k 8 G 8 G 2 (2.40) that the frequency changes with the scale factor according to 0 = a(t1 ) (2. At a temperature of around 40000K (the decoupling temperature). we can get the redshift today (t = t0 ) from eq. the Boltzmann factor for a photon satis es e kT1 = e kT0 8 h1 h0 (2.77) gives rise to the search (theoretically and experimentally) for dark matter. Presumably the temperature is very high. not enter this fascinating subject. This limits the three-dimensional curvature k=a2 . so matter and radiation cooled. If is very large. because of eq.77) we have 0:1 3 (2. the converse is true (if = 0 there is only expansion. stellar evolution. at ( = 1).

79) one gets a lower bound on a(t). the big bang predicts that there should be a thermal distribution of photons present today.86) a(t ) (t) Inserting the densities of baryons at t and at present ( (t0 ) 10. (2. (2. i.87) Now using Z 1 x3 dx 4 = 15 (2. left over from the time of the decoupling of photons from other matters. which was also done by a number of people (Gamow. 8 h 03 d 0 ( 0 )d 0 = (2.83). Peebles.8 COSMIC MICROWAVE RADIATION where we have de ned 69 a(t ) (2. After the recombination the universe is matter dominated. 0 . Dicke.89) today From (2.e. In particular we get at the present age of the universe the Planck black body radiation distribution of the photons. 1 we obtain by integrating (2.71) for in the radiation dominated early universe. 1 0 Thus. Herman. =B (2. Thus DMW (all times) MW = a4 In the relativistic era the photons were not the only component to the energy density so D > DMW (2. and we have (2.69).90) Thus the constant D in the radiation dominated density = D=a4 is huge. The presently observed microwave photons contribute a relativistic component MW to .85) a3 from which we get ! a(t) = (t0 ) 1=3 (2. We shall return to this later.83) T0 T1 a(t1 ) 0 If there is a thermal equilibrium.84) with T0 = 2:75 0:05oK. Alpher. = D=a4. In principle the temperature T0 at present can be predicted from eq. in good agreement with the theoretical estimates. Presumably the most signi cant experimental result in cosmology since Hubble's discovery is that such a spectrum is indeed observed (Penzias and Wilson 1965). Recent satellite measurements give a very precise Planck spectrum. as they did in the relativistic era. We can now obtain a lower bound on the constant D in eq.84) over the frequency 5 a4 D > DMW = MW a4 today = (kT0 )4 815h3 (2. and inserting this we get for h = 1 D 6 10114 (2. Roll and Wilkinson) before the discovery of the radiation. we can thus absorb the scale factor in the temperature.88) 0 ex .2.84) h exp kT0 . The experiments give a nice black body radiation spectrum (2.6g=cm3) the most careful prediction was T0 5oK.75) and the bound (2.

at. This is seen to be the case in the above integral.95) or (2. 6. + a0 ) _ 0 0 0 a2 a 0 This equation can easily be solved for time. 0 0 a . and hence the life time is expected to be short/large. 1) Using this today in eq.95) x 0 (2. (2.96) where the inequality (valid for 0 > 0) was already derived from the qualitative arguments in sect.92) (2. and 0 < 1 (closed.9. 1) = H ( . so = B=a3 .96) can easily be performed.0 1)3=2 ( .95) can then easily be performed and we get sin2 2 = (2.91) we get a2 = H 2(1 . 0 + x0 In particular.75) we have k =8 G 2 a2 3 crit( . The intuitive appeal of eq.9 The matter dominated era a2 = 8 G _ a0 k crit. gravity counteracts very much/little. 0+ 1 H0 COSMOLOGY We now consider the matter dominated era with negligible pressure. Here the Friedmann equation (2.96) should be emphasized. The integral in (2. (2. sin ) 0 (2. the present age of the universe is 1 Z 1 q dx t0 = H 0 0 1.97) H0 t = 2( . If 0 is very large/small. respectively).91) (2.98) . open universe. 2. a2 2 a0 3 0 where we divided by a2 (a0 is the scale parameter today) and used 0 3 = a0 a 0 From eq.1 0 a(t) a0 The integral (2.1 The closed universe We de ne the \development angle" by 0 >1 0. (2.70 2.60) becomes (2.93) (2. It is most convenient to distinguish the case 0 > 1 0 = 1.94) (2. 1 Z a=a0 q dx t= H 0 0 1 .

(2.102) or (2.2. 1 t0 = H f ( 0 ) (2.100) 0 q f ( 0 ) = 2( . 01 (2.97) and (2. 1 = 2 1 . (2. and reaches a maximum at 2H0( 0 . The scale parameter increases from zero at = 0 t = 0.0 1)3=2 cos. shows that t0 > 1010 years.104) (2. 2.99) 0 and then it returns to zero at = 2 t = 2tmax. = i . 2 0 .98).97) and the result (2.102).9. as mentioned before. The result (2. the life time is very small.97) that ! 0. 1)3=2 a amax = 0.102) 2 1010 yr (2. Expanding the trigonometric functions one 2 then gets (2. 0 )3=2 (sinh . for 0 ! 1 one has f ( 0 ) ! =2 0 .101) 0 0 0 tmax max = = 0 for 0 1.9. Since evidence from geology etc. we see that if 0 = 1 then we cannot have h = 1. 1 . if 0 is very large. (2.97).1 (2. except that the development angle is imaginary.103) 3h with 0:4 h 1.3 The open universe Formally we can again use the substitution (2. Thus. 1 ' 0:57. The result (2.97) shows that a = a(t) is a cycloid.103) again has the form (2. (2.95) is trivial and gives 0 =1 2=3 a(t) = 3H0t a0 2 and the age of the universe is 2 t0 = 3 H0.103) could also be derived from eqs. Hence we get 0 <1 H0t = 2(1 .98) by noticing that for 0 ! 1 it follows from eq. The function f ( 0 ) is seen to be monotonically decreasing. so f (1) = 3 .100). The present age of the universe is obtained by putting a(t) = a0 in eq.1 2 . but h must be less than 0:67.2 The at universe Here eq. 0 0 (2. 1 (2.9 THE MATTER DOMINATED ERA 71 which together with eq. (2. 2.105) a(t) a0 . ) 0 with cosh . For 0 = 2 one p has f (2) = 2 .

Les Houches Lectures 1985.72 COSMOLOGY Figure 2. 517) The age of the universe is given by (2. with an upper bound 2=4 (obtained for 0 ! 1).107) and since f ( 0) is monotonically decreasing.107) we then have f ( 0) 1 (2.9. so independent of h one has from 2 2 =4. Turner. .100).108) 0h2 0f ( 0) .3: The function 0 f ( 0)2 .106) Again f ( 0) is monotonically decreasing with increasing f (0) = 1. 0 0 1 0 0+2 0 0.78)).4 An upper bound on 0 In section 7 we discussed how to get an upper bound on 0 (see (2. II. p1 . 0 (2. but now f ( 0 ) = 1 . (2. (2. Vol. 2(1 . We mentioned that by the dating of the oldest stars and of the elements it is strongly suggested that the universe is at least 10 billion years old. Also plotted is the straight line h2 for h = 0:4.S. p. From eq. (From M. and the maximum value is 2. Since we have t0 = f ( 0 ) H 0 1010 yr f (h 0 ) (2. )3=2 ln 2 . The best estimates obtained this way actually point to 15 billion years.3). we can now ask what a lower age limit of 1010 yr means for 0 . 2.108) h Now let us notice that the function 0 f ( 0)2 is monotonically increasing (see g.

these old star life times agree. and t0 = 13:5 2:0 Byr from giant branch tting (readers. (2. then we get if we take h = 0:4. This is illustrated in g. k = 3 a2 _ q with the solution a(t) = eHt H = =3 for positive.111) a2 = 8 3G ~a2 . For photons d = 0. Thus. the expansion should somehow be enhanced. to add g on the left hand side of the Einstein equation.e. this can be checked by measuring the Hubble constant H or h very accurately.e. in spite of premature reports in newspapers on the death of the cosmological model. To illustrate the problem for the reader. This is one of the main purposes of the Hubble space telescope. we see that if is positive. It is easy to see (Problem 16) that the old cosmological Einstein equations are satis ed. has been much debated. We end by remarking that the age problem has been there since Hubble's dicovery. At present (1996) the situation is very unclear. If the age limit is 15 billion years. There is thus some indication that the small (including dark matter) is not enough to provide su cient life time from the cosmological model. If is introduced. let us mention a few numbers: The ages of the oldes globular clusters are t0 = 15:8 2:1 Byr. i.9. In principle. 8 G ~ To see that is related to expansion. with no matter.110) p!p=p. As an example.10 Causality structure of the big bang The region of causal contact with a given event is determined by how far photons originating at that event can travel. Thus.2. With 0 = 0:35 and = 0:65 these numbers t. It is left to the reader to ponder as to whether this solution is too easy. based on main sequence turno . suppose t0 = 15 Byr and 2 0 = 0:1. we can do better than that by asking when 0 h2 equals 0 f ( 0)2 . This has led some people to (re-)introduce the cosmological constant. = 0 and p = 0. 0 3 (2. 0h 2 1:5 i. provided we make the replacements 2. An example: Let t0 = 13:5 Byr and h = 2=3. it is possible to have expansion just from the cosmological constant. To get out of this problem. let us take the simple example of at space k = 0. the life time of the universe was shorter than the life time of the planet earth! 2. This is below the reported values from the Hubble telescope (h = 0:7 0:15). Using his value for H . ~ = 8 G .5 Discussion of the life time of the universe ! ~= + 8 G : (2. The Friedmann equation then gives (2.109) It is clear from common sense that the universe must be older than the oldest stars. because then 0 must be less than this value of 0 .10 CAUSALITY STRUCTURE 73 However. however. so the Robertson-Walker metric .3). So within the uncertainty. Then h = 3 0:9 = 0:59. it is easy to solve any age-problem. which. who want to understand these concepts should consult a book on astrophysics).

e. For k = 0 we have from (2.112) we get d = 3 t2 . This is not the case in the early universe.115) d 2t2 1 . t1 t Consequently we see that for t2 small (but t2 2 d a(t2 ) v u 2t2 uq t 8 3 GD t1 ). Consequently the term a2 = D=a2 dominates over the curvature term. (t1t2 )1=3 2 3t2 (2. k _ (2. a2 = 8 3G a2 . where dl is the time-independent distance in co-moving coordinates. i. so that physics in any point can establish itself to be equal to physics in any other point by means of causal contacts. the situation does not appear very self-consistent. at any stage in its development.112) t1 a(t) where the time t1 is the time the photon was emitted. The conclusion from this is that in the early universe the causal distance is much smaller than the scale factor.60). The physical causal distance is thus Z t2 dt d = a(t2 )l = a(t2 ) (2. Although this is an assumption.74 COSMOLOGY gives dt2 = a(t)2dl2 . and this is clearly a di culty with the standard model.117) a = @2 3 and hence from (2.118) . in the early universe 1 (2.116) for t2 small (i.112) we thus get for the physical distance. since the standard model assumes that the universe is.114) p so a(t) is proportional to t in the early universe. ct2 small). From a physical point of view one would like the homogeneous universe to be established by the contact between all points by physical light signals. From (2. and we get quite easily by solving a2 ' (8 G=3)D=a2 _ that v s u u a(t) t2t 83 GD (2. and a is quite small. s ! 2t2 for t2 t1 (2. Thus.113) In the early universe = D=a4. close to the initial singularity the universe consists of many arbitrarily small causally separate regions.113) 12=3 0 s 3 t 8 GB A (2. and although one is free to make such an assumption. To obtain a(t) let us consider the Friedmann equation (2. homogeneous. This is clearly very surprising. In the matter dominated (later) universe the situation is di erent. whereas t2 is the time it was received.e.

Thus the data show that a truly striking degree of isotropy is present in the universe.11 INFLATION 75 for t2 t1. it follows that in _ early times mest have been extraordinarily closer to 1 than it is today.11 In ation We have seen that the standard model has many successful features. The microwave background discussed in section 8 is observed to be very isotropic.e. but has to be put in \by hand" to t the observations. Let us call f the energy density of the false vacuum (relative to the true vacuum). 1): _ (2. which point to di culties for the model if we want to understand it in more physical terms. it is not logically necessary.1 . Now we know that today is at least 0. however.1.2). the system is in a metastable state. it remains equal to one. let us assume that it takes a rather long time for the false vacuum to decay. we saw that certain features emerge from the standard model. = 1. which is not very far from 1.0. So inside the cylinder and piston. At grand Uni cation should be equal to 1 to an accuracy of 49 decimal. Finally. To nd the pressure of the false vacuum. In quantum eld theory one can have negative pressure. 2. Although this seems intuitively natural. . To give some examples: For to be in the allowed range today. Outside we have true vacuum. 9 In ation will be discussed later in the course on quantum eld theory. where the di culties arise .2. in short. The Friedmann equation can be rewritten k = a2 ( . In the last section. \the atness problem". It turns out that the microwave photons originated in causally separate regions if their directions di er by more than a few degrees. so \false" could be replaced by \temporary". the false vacuum decays to a stable \true" vacuum. In recent years a new scenario called \in ation" has emerged. it must have been equal to 1 to an accuracy of 15 decimal places 1 sec after big bang (time of nucleosynthesis). In the big bang model this striking isotropy is not explained. We cannot give a full explanation of this. let us give an intuitive description of what happens. In ation concerns itself with the early universe. and gives a nice framework for discussing the rather few astronomical observations which are available. Given enough time. Suppose. and since a increases if we go backward in time. which is the state with lowest possible energy. let us discuss another puzzle. If is ever exactly equal to one. so throughout the later evolution of the universe the largest regions of causal contact are of the order t. Now move the piston outwards. Flat space. with = p = 0. Also. This fact is not explained by the Standard Model. called the \false vacuum": On short time scales the energy cannot be lowered.3. and in the big bang standard model this could not have been produced by causal processes. at a very early time. However. let us imagine that we enclose it in a cylindrical container with a movable piston. which is larger than a(t). is an unstable equilibrium. we have the energy density f .119) Since k is a constant.9 In the previous work we have assumed that the pressure is positive. since we have not yet discussed elds (the phenomenon of negative pressure is discussed in section 4. the scenario in in ation. This is. where many of the physical di culties have been removed. i.

1: _ (2. so 15 4 (2. You may wonder where the extra energy arising when we pull the piston comes from? It is.125) p=. we always have dU + pdV = 0. In natural units f has dimension (mass)4.e. because the solution of the attractive gravity equation is = 83 G f (2. This is because this energy is (temporarily) the lowest possible one. for t1 < t t2 we have negative pressure and the equation of and the Hubble constant (which is now a true constant) becomes a = = 1034sec. f <0 (2.128) in ation: a 3 f Let us now modify the standard big bang so that for 0 t t1 we have the usual radiation domination ( = D1=a4). However.122) a = + 83 G f a so gravity is now repulsive!! This means that during in ation. and f is time independent as we knew beforehand. Now. the pressure is negative. d ( a3 ) = . where dV is the change in volume.120) a(t) = c1e t + c2e. on the contrary.e.34sec (2. The Friedmann equation for k = 0 is then just .126) a Let us consider the density during in ation. t s s a = 8 G =H _ (2. during the time the false vacuum dominates (2. Hence. we would have to compress it to the size of a proton. gravity does not counteract the expansion. So the change in energy dU is given by dU = f dV . i. the false vacuum keeps its energy density everywhere. gravity is attractive.62)) a = . Let us estimate f . also in the \new" volume. A simple model is to assume that it is the density relevant at the Grand Uni cation scale. with temperature 1015 GeV. ((2. 43 G( + 3p)a: (2.123) where c1 c2 and are constants (remember that f is the constant energy density of the false vacuum).121) Usually this means a < 0. of course.124) f (10 GeV) This is enormous! To get this density in the sun. i. We then get 1 10. Consider eq.127) f f da f da so d f =da vanishes. who has to perform this work againsk negative pressure! The negative pressure has profound consequences. supplied by the agent pulling the piston.76 COSMOLOGY When the piston is pulled.3pa2 = 3 a2 = da3 (2.

t1 10. S.129) The important feature is that even if D1 is of order one. 1). t1 ) 66 then one can have D2 10114 with D1 = 0(1). What about ? We still have Friedmann. then D2 can be very large. J. 317.12 EVIDENCE FOR THE COSMOLOGICAL CONSTANT 77 state (2. J. the energy stored in the false vacuum is released in the form of new particles (in thermodynamics language the released energy is the \latent heat" of the phase transition). Going backwards in time we encounter in ation. 2. p t2 < t < td radiation dominance a t = D2=a4 2=3 td < t matter dominance a t = B=a3 The integration constant D1 and D2 are now determined by requiring consistency at the times t1 and t2. t1 ) > 66 is reasonable. The main point in using distant supernovae is that they are considered to be very good standard candles: they are assumed to be associated with the nuclear detonation of white dwarfs. where the equation of state (2. It is completely outside the scope of these notes to describe the theory of white dwarfs11 .. 10 .and B.120) is satis ed.t1 ) (2. Models. which is the right time for the in ationary period. 11 See e. so is driven towards 1. Riess et al. These new particles then come to thermal equilibrium at the temperature T = 1014 GeV. In this small region photons are in causal contact. Since = Hin we need t2 . Then the evolution looks like: p 0 < t < t1 radiation dominance a t = D1=a4 t1 < t < t2 in ation a etH in . producing an incredibly small region out of our universe. If Hinfl:(t2 . = const. 116 (1998) 1009-this paper is most easily found on the net.130) in ation = 1: Finally. for t2 < t td we return to the relativistic equation of state. Thus. the large value of D2 can be obtained in a quite reasonable way from a decent value of D1. Astrophys. it should be said that in ation occurs in many versions. p. so _ (2. are called \in atory" models. Astrophys. As the in ationary period proceeds all elds get red-shifted away exponentially fast. The in ationary period essentially wipes the slate clear of primordial uctuations (from 0 < t < t1). Gravitation and Cosmology. so D2 = D1 e4Hin ation(t2 . whereas for t > td (td =time of decoupling of photons) we have the usual matter domination ( = B=a3 ).. k = a2 ( . One may ask whether Hin (t2 . 507 (1998) 46-see astro-ph/9805200. and when a(t) blows up exponentially the universe is said to be in ating.32 sec.120). with its enormous contraction (we go backwards!).2. _ During in ation. What happens to the causality problem? Start with the present universe.12 Evidence for the cosmological constant from observations of distant supernovae Recently (1998) evidence for a non-vanishing cosmological constant was found by observing 50 type Ia supernovae out to redshifts of order one10 .g. because of the knowledge of A. In the phase transiton where false goes to true vacuum. Schmidt et al. so there is no problem. which is causally connected. The data was plotted in the magnitude-redshift Hubble diagram. However. a increases extremely rapidly. see astroph/9805201. and some of them are somewhat di erent from the intuitive description given above. Weinberg.

133) M a2 Here M = = crit is the matter density.132) crit 8 G a we obtain the following form for the Friedmann equation. which can be interpreted as a vacuum energy. and = =3H 2 is the relative density associated with the cosmological constant. Turner. and because it has become possible to observe high-z objects. since it exists irrespective of matter. (2. the old uncertainty concerning the determination of the distance in Hubble's relation seem overcome. and the expansion is driven by the last term. and hence the (luminosity) distance can be computed from the ratio of these quantities. the intrinsic luminosity and the observed ux are known. k = H 2( + .136) 3 M 3 6 We expect M to be of order 1/3 (for the most recent survey of the most important cosmological parameters H M :::. there is no serious objection against Ia as a standard candle. because gravitation slows the expansion. In the standard model. with 1 3 q = . aa = 2 M . (2. astro-ph/ 9904051).135) 2 a _ 2 Here tot = M + is the total energy (ignoring the small contribution from radiation). However. From (2. = 1 tot . the supernovae observations showed precisely the opposite! The conclusion is therefore that the expansion speeds up.136) we then get 4=9 + 1=3 = 7=9 2=3 (2. see M. For nearby galaxies it is su cient to use the simple Hubble \law".137) the blow up mechanism. 1): (2. in a late universe the two rst terms on the right hand side are subdominant.134) 0 where q0 is the deceleration parameter now. S.78 COSMOLOGY For a positive .131) a 3 a2 3 . distant galaxies should have larger velocities than predicted by the simple Hubble law.44)). Therefore.111) that the evolution of the velocity a _ changes a 2=8 G . Remembering that 2 _ = 3H and H = a (2. In particular. This is in contrast to the standard model without a cosmological constant. 2 : (2. However. features of the distant sample of Ia's appear to be similar to a nearby sample. From the high-z supernovae observations mentioned above one then nds 4 + 1 1: (2. By further expansion of z = (a(t0 ) . a(t1 ))=a(t1) one obtains the result (solve problem 17) generalizing the simple Hubble law. z = vr = HL. introducing we saw in connection with eq. k + : _ (2. where L is the distance and vr is the radial velocity (see eq. 1 z H0 L0 + 2 H02L2 (1 + q0 ) (2. So far.

we shall give some explanation of what these peaks are. http://xxx. Thus. by observations of the background radiation over a large angular range from satellites.e. It is to be noticed that the last result is at least roughly consistent with in ation. M. With 2=3 and 1=3. thereby causing small uctuations in . An excellent review is12 Wayne T. When talking about uctuations in the CMB (the cosmic microwave background). The conclusion is therefore that the cosmological constant is necessary because of the recent supernovae observations.g. quantum uctuations in the early universe are blown up in scale by an exponential. i. The subject of density uctuations plays a very important role in modern cosmology.13 Appendix: Density uctuations. and we must con ne ourselves to a very introductory discussion. Hu. e. 12 . Ann. then enter the paper number 9508126. H. Therefore the serious student of cosmology should have a good knowledge of the theory of density uctuations.138) 1: These numbers are expected to have an uncertainty of the order twenty percent. can be measured with a few per cent accuracy.lanl. This gives rize to the problem that M we live in a pro table time from the point of view of observations.gov (see astrophysics). Press. due to the presence of galaxies etc. More recent material can be found on the astro-ph archives. For further discussion. Later on. Click on the command PostScript to get the paper itself. Due to lack of time we are not able to give a complete treatment of density uctuations. in an in ationary scenario. by searching for \Doppler peaks" (or just \Doppler"). the universe will go on expanding forever.13 APPENDIX: DENSITY FLUCTUATIONS which is consistent with a total energy tot 79 (2. 30 (1992) 499-542. 2. Review Astron. The abstract of the paper then occurs. e.2. it is clear that a realistic cosmology somehow must break this assumption.gov/abs/astro-ph. Astrophys. When the CMB photons last scatter and decouple from the To get this paper from the Archives. astro-ph/9508126. Cosmology seems to converge rapidly towards a very precise science where observations determine the parameters. uctuations in the density . H :::. one distinguishes between the primary (which we shall discuss). use the address: http://xxx. and the secondary uctuations (which we shall not discuss). The reader is referred to the literature for a detailed study. Carroll and W. Although the standard model makes the a priori assumption that the universe is homogeneous. Doppler peaks and all that We shall now discuss the very important subject of density uctuations. we refer to the recent review by Turner mentioned above. It is realistic to expect that the basic cosmological parameters. since the real universe is inhomogeneous.g. Wandering in the Background: A Cosmic Microwave Background Explorer. we refer to the excellent review by S.lanl. because in the future astronomers will be able to observe fewer and fewer objects. For an account of the life time of the universe with a cosmological constant as well as other problems.

If the potential varies with time. ultimately clusters of galaxies form.80 COSMOLOGY baryons at redshift13 z 1000 (corresponding to the distance 6000 h. there is a net frequency shift. causing to decay somewhat. they receive a Doppler shift due to the motion of matter with radial velocity vr . and then perturb around it. the second is the Euler eq. and the third is the classical Newtonian gravity eq.1 The classical Jeans analysis The basic idea is to consider the standard model (with a homogeneous and isotropic universe) as a rst approximation. In the absence of di usion and other forces. pressure. Such a program is quite complex. It is important after recombination. of gravitational instabilities.. if 6= 0. As already mentioned. and we shall therefore only outline what happens if uctuations in the matter part are included. described by the density . however. Consider a non-relativistic gas or liquid. due to Jeans. one also needs to take into account uctuations in the metric.139) The rst equation is the continuity eq. and velocity uctuations. either the universe is vacuum dominated ( + = 1) or curvature dominated ( + 6= 1). Also.13. although they are of course of importance for the interpretation of the observational data. This is called the integrated Sache-Wolfe e ect. . we shall reproduce an analysis. The secondary uctuations are due to the fact that even after recombination. Since. 2. leading to a time dependent . The perturbations are in the form of density. Photons coming out of an overdense region will have a higher temperature. r r2 = 4 G : (2. the metric is coupled to matter in general relativity. the baryons have not completely lost their ability to in uence the photons. in the following we shall not discuss the secondary uctuations. velocity v. It is easy to verify that these equations have the solution 3 a (2. and non-linear e ects become important.140) 0 (t) = ^=a(t) v0 (x t) = (_ (t)=a(t))x 13 14 We have 1 + z = a(tnow )=a(trec ) 4000 0 K/2. these quantities satisfy the following equations: @ =@t + r( v) = 0 @ v=@t + (vr)v = .14 To get some feeling for the development of the density. The inclusion of metric perturbations will be presented without any arguments. Also.1 Mpc). to a gas).(1= )rp . and hence again is time dependent. where the photons still contribute marginally to the density of the universe. When the photons scatter the last time.735 0 K 1400. (=the transcription of the Newton eq. simply because the blueshift obtained by falling down is not exactly cancelled by the redshift from getting out of the potential. since there is still the gravitational e ects. they are imprinted with three primary uctuations: Photons last scattered in a gravitational potential experience a redshift as they move out of the potential. pressure p and gravitational potential .

obtained from (2. Using that the unperturbed quantities actually satisfy the hydrodynamical equations (2.147) ~ To proceed. Z 3 (2. We then get from eqs. ik ~ =a(t) ~ @ v1 =@t + (_ =a)~1 = . Hence we have an adiabatic situation.144) which allows us to eliminate the pressure completely from the above equations. Also. r 1 a a 2 =4 G : r 1 (2.145) (x t) = d k3 eikx=a(t) ~(k t) (2 ) etc. This is permitted as long as the energy density is dominated by non-relativistic particles with small pressure p and as long as all relevant scales are small compared to the characteristic scale a(t). We now study the perturbed equations. so that (x t) = 0 (t) + 1 (x t) 0 (t)(1 + (x t)) v(x t) = v0(x t) + v1(x t) (x t) = 0(x) + 1(x t) p = p: (2. Then we obviously have the uncoupled equations v kv? = 0 vk = k k~21 iku: k (2.148) .g.146) 1 The last equation thus gives the correction to the gravitational potential in terms of the density uctuation. (2. ~1(k t) = 4 Ga(t)2 ~1 (k t)=k2: (2. we have taken into account the expansion of the universe through the scale factor.(r p)= .143) Thus the pressure is given by 2 p = vs 1 (2.i(vs 0 1 a v 1 2 ~ = .13 APPENDIX: DENSITY FLUCTUATIONS 81 Here a(t) is the scale factor.139) by inserting the expressions (2. although the basic equations are Newtonian. which satis es the \force" equation (2. p is small. Thus.49). and ^ is a constant. Thus we use that e.141) Here j j jv1j j 1j are small relative to the corresponding unperturbed quantities. let us perturb around this solution. All quantities are now Fourier transformed.4 Ga(t)2 ~ : k 1 (2.2. Now. 2 vs = (@p=@ )adiabatic : (2.141) and ignoring non-linear terms. we easily obtain @ 1 =@t + 3(_ =a) 1 + (_ =a) (xr) 1 + 0 rv1 = 0 a a @ v1 =@t + (_ =a)v1 + (_ =a)(xr)v1 = . as discussed previously.142) 1 In general these perturbations correspond to sound (pressure) waves. it is convenient to split v1 into a component v? perpendicular to k and a component vk parallel with k. which propagate much faster than thermal perturbations.139).142) @ ~1 =@t + 3(_ =a)~1 + i(1=a) 0k~1 = 0 a v 2 = a)k ~ . The factor 1=a(t) in the exponent takes into account the red shift of the waves. and for the sound velocity vs we have.

e. given by eq. 4 G ~(k t) = 0 _ _ ~(k t) + 2 (2. a = 0. Thus 3 1 (2.141). i. This corresponds to Jeans' classical analysis. = ( )d = aT 4. so we simply have 2 vs = (@p=@ )adiabatic = 1=3 (relativistic gas): (2. with a = 8 k4=15h3 = 7:5641 10.151) which is not a wave. but the general features remain: Below kJeans there is a uctuation which grows inde nitely in time. p = =3. This equation is the fundamental equation for the non-relativistic evolution of the relative density uctuations. The rst eq. (2.15 erg cm. because then the frequency ! becomes imaginary.. In this scenario.154) From the time of electron-positron annihilation (T 1010 K) until recombination (at T 4000 K).i!t) is unbounded.150). _ We shall end this subsection with some remarks on the velocity of sound.146) then splits into the equations COSMOLOGY obtained from the second equation in (2.146) simpli es to _ ~ = k2 u=a: (2. the solution has a component which grows inde nitely. For k > kJeans the solution is just an oscillatory and bounded sound wave. 2 @ v?=@t + (_ =a)v? = 0 @u=@t + (_ =a)u = (.vs =a + 4 Ga 0=k2 ) ~ a a (2.82 The second equation in (2.4 . let us assume that the scale factor is a time independent constant.3degree. The compressional modes vk . and hence exp(.153) p From this we see that if k < kJeans = 4 G 0 a=vs. which is the Hubble constant. (2. where _ 2 !2 = k2vs =a2 . Thus gravity produces an instability for su ciently small k. The damping rate is given by a=a. pressure. it is natural to assume that the main contents of the universe is nonrelativistic ionized hydrogen (baryons) in thermal equilibrium with black-body radiation at temperature T .152) 0 a a2 . the thermal energy. To make ourselves familiar with this equation. (2.143). and entropy can safely be ignored relative to the same quantities for the photons. 0 )= 0 (see eq.i!t). and k is the Boltzmann constant. (2. for sufciently large distances.148) has the simple decaying solution. For a relativistic gas. on the other hand.155) = nmH + aT 4 p = 3 aT 4 s = 43aT : nk Here the constant a (not to be confused with the scale factor a(t) ) is given from the R di erential Planck distribution by integrating over the frequency. the mathematics is more complicated. (2. satisfy the wave equation ! 2 a ~(k t) + vs jkj2 . v? / 1=a(t) (2.152) is of the form ~ / exp(. In the general case with a time dependent scale factor.150) The rst of the uncoupled equations (2.149) by inserting (2. = ( . Then the solution of eq.141)).149) where is the fractional change of density de ned in eq. whereas for k > kJeans the solution is that of a damped harmonic oscillator. 4 G 0 : (2.

we do not have the time to derive this result. which is a genuine general relativistic correction.159) In these equations the dot refers to derivative with respect to conformal time15 . and s is the speci c entropy. we need to take into account also perturbations of the metric.160) . so R = 0. time does not permit such a discussion here. s is constant. k2 ~: (2. so n / T 3. r2 d 2 : (2. Let us remark that the essential di erences c 15 . as already known. _ _ F = 3 ~c + (3R=(1 + R))H ~c . (2. It is important that the velocity of sound depends on the baryon to photon ratio. and hence vs = 1= 3. (2.2. and using the variation of the rst two equations (2. here the Fourier transform is de ned as in eq. kr2 .155) with respect to temperature we obtain the sound velocity ! 4 aT 4 ( T=T ) 2 = @p 3 = = 1 (2. Since we consider an adiabatic situation. of course there are no baryons. since we are perturbing around the standard cosmological model where v = 0. because if the sound waves can be observed. dr2 d 2 = a( )2 d 2 . In a more realistic treatment we would have to take into account such e ects. D.2 in this thesis) mentioned in the beginning. which has no Newtonian analogy. = dt=a(t): Also. Again. (2. 5. 1 .145)) _ 2 ~(k t) + (R=(1 + R))H ~(k t) + vs k2 ~ = F _ (2. since the scale factor becomes a common factor.145) with k=a(t) ! k in the exponent in eq.156) vs @ s=const (3nmH + 4aT 4) T=T 3(1 + R) where baryon R= 3 4 : (2. n is the number density of the hydrogen gas.13 APPENDIX: DENSITY FLUCTUATIONS 83 In the very early universe. Instead of eq.161) As usual.162) This follows from perturbation theory around the Robinson-Walker metric. The Euler equations (2. and c is the perturbation of the spatial curvature.2 Taking into account relativistic e ects Z (2.: Conformal time makes the Robinson-Walker metric more symmetric. To the rst approximation it is related to the Newtonian potential.13. ~ is the Fourier transform of the Newtonian potential (de ned through Poisson's equation). and the interested student is referred to the Ph. we shall be able to gain information on this essential parameter.152) one now obtains (in contrast to before. Unfortunately.157) p 2. thesis by Hu (see sect. (2.139) used above are not relativistic. In principle.158) where F is the so-called \forcing".

the damping term now vanishes if the sound velocity is constant. with oscillations between . and instead the e ects of gravity are combined in the forcing F . The resulting gravitational potential from Poisson's equation (last eq. 6R and -2.165) Thus the solution reduces to ~(k ) = (1 + 3R) ~ cos(vsk ) . 3(1 + R) ~: (2.166) The adiabatic initial conditions thus produce a pure cosine oscillation. These imprints are characterized by the peculiar18 gravitational 16 One has 4 G ~ = 4 G ~ = . A derivation of eq. since it is related to through Poisson's equation.164) This simple result shows that the zero point of oscillation (0) = . 1 1 .k2 ~: (2.k 2 ~ . (2.158) then reduces to 2 ~ + vs k2 = . In this situation.158) is that the term with 4 G is absent in the the latter. Thus.e. Eq.1Mpc away from us. with the forcing F given by the last term in eq. One can therefore use eq. (2. the last scattering of the photons)17 . (2. 3(1 + R) ~: (2.3(1 + R) is shifted with the baryon contents.152) and (2. It can be shown that if one has adiabatic (constant entropy) initial conditions then _ ~(0) = 0 ~(0) = .163) This is a simple harmonic oscillator. This leaves us with only the photonbaryon plasma and heavy dark matter.159) as a rst approximation16. baryons provide only a small fraction of the energy density.158).4 Primary uctuations The photons decouple from the baryons at a redshift around z 103.2 ~: (2. the amplitude of oscillation increases with the baryon contents. Likewise. and are therefore ignored. let us assume that this time dependence is negligible. where Poisson's equation was used. 18 The word \peculiar" refers to quantities which deviate from the standard model. vs = 0. (2. (2. The solution is thus _ ~(k ) = ( ~(0) + 3(1 + R) ~) cos(vsk ) + ~(0) sin(vsk )=vsk .158) is given in section 5.4 .3 The peaks In general we expect that is time dependent.e. the photon-baryon system is so tight that it can be treated as a single uid. we shall give a qualitative discussion of the various uctuations imprinted on the background radiation from the time of the decoupling (i.146)) can then be used as a rst input to obtain the evolution of the photon-baryon system.13. 2. the velocity=0 in the standard model. Also. (2. corresponding to a distance of order 6000h. Massless or light neutrinos do not give an important contribution. and also that the velocity of sound is independent of time. driven by the gravitational force (assumed time independent). 2. The rst two terms in the forcing F are due to metric perturbations.152) for CDM with zero pressure (the universe is matter dominated). called cold dark matter (CDM). As a simple model. This gives eq.13. so vr is a peculiar velocity 17 Before proceeding.84 COSMOLOGY between the equations (2. The idea is that before recombination.2 of the thesis by Hu mentioned in the beginning. i.

Thus (2.13 APPENDIX: DENSITY FLUCTUATIONS 85 potential . Thus. This can be expressed through the integral 19 T=T = Z _ (x( ) )d (2.vr (2.168). = . for the same reason in eq. the density uctuation . where we derived the change in frequencies in a peculiar potential.40) and (2. 1. This is called the integrated Sachs-Wolfe (ISW) e ect. the redshift from moving out does not cancel the blueshift entirely. For example. and the radial peculiar velocity vr .g.171) Here x is the comoving vector from us to the surface of last scattering of the photons. (1. The frequency is related to temperature e. The e ects on the photons are: If the last scattering is in a potential well < 0.Since the peculiar velocity is quite small (of order 10. they will be Doppler shifted. The last item refers to the following relation between the relative density uctuation and the temperature uctuation . and hence there is a net blueshift. Adding up everything19 . corresponding to a red-shift. the temperature shift is given by (x tr ) T (x tr )=T (x tr ) = 1(x tr ) . discussed in sect. see eq. T=T = =3: (2. The dot means derivative with respect to conformal time for a xed position.82). and tr is the time of recombination. when the photon enters the potential it acquires a blueshift. Its physical interpretation is simple: The integral over _ measures the frequency shift that the photon experiences when it enters and exits a time dependent potential. so n nb .170) where x( ) is the trajectory of the photon. Now. If the last scattering is on matter moving with a peculiar velocity vr > 0 away from us. If the photons emerge from an overdense region. they have a higher temperature. and if the potential satis es _ > 0.169) where vr is the radial velocity with vr > 0 when the source moves away. (2. the photons will be redshifted when getting out of the well. The latter follows because baryons dominate the density.41). so T / 1=3 . After recombination. . it also gives rise to a shift in temperature. we use the non-relativistic Doppler formula. (2. v(x tr ) x=jxj + (x tr )=3: (2. simply because the frquency is shifted in proportion to velocity . when the baryons are essentially decoupled from the photons from the point of view of Compton scattering.2. which should be added to eq.4. n /T The rst item is a simple consequence of the redshift in a gravitational potential.167) This can be understood from the fact that the photons are tightly coupled to the baryons.171).168) 1 =T1 = 0 =T0 so = 0 = T=T0 = : Concerning the Doppler e ect mentioned above. there is still a gravitational e ect from the baryons. b 3 from the integrated Planck distribution.5). T=T = =3 = =3. by the combined use of the relations (2.

Therefore.176) to the so-called Doppler peaks (2.166) with R = 0. The spectrum corresponding to (2.172).171). but only mention that any relevant function F is expanded as 20 2. This will be done subsequently.86 COSMOLOGY We shall now discuss how to observe the uctuations included in eq.177) j (k )j2 R=0 = j ~ + =3j2 + vr2 = 1 j ~j2 9 Of course. what we can do is to observe this quantity as a function of k. and hence their expectation value vanishes. the smaller is the angle . where Pm is essentially di erent from zero. i.13. we expect the various elds in this equation to be random. If and then comes a number of peaks. one has k l.173) . < vr >=< jvj2 > =3.172) j (k )j2 R=0 vr =0 = j ~ + ~=3j2 = 1 j ~j2 cos2 (vsk r ) 9 where we used eq. rougly speaking. it follows that for low k( l) the spectrum is at.176) We now have the curious situation that when we add the Doppler term (2. m)!=(4 (l + m)! Pm (cos ) exp(im ): (2. we get for R = 0 and hence vs = 1= 3 D 2E 1 D ~ 2E 2 vr = 9 j j sin (vsk ): (2. so we get by use of eq. The average of the square of the Fourier transform of the rst and third terms in eq. The non-trivial quantities are the expectation values of of powers of the relevant elds. and compare20 with the theoretical expression (2. Some people call these peaks \Doppler peaks". (2.174) l l The polynomials Pm have the property that the larger l is.spectrum. in doing so we should include the Doppler shift also. A more precise relation involves the curvature. This expression is evaluated at the xed conformal time r corresponding to recombination. although they are not related to the Doppler e ect.172). The qualitative features of the spectrum exhibited in eq. large l's probe small distances.172) are quite interesting: ~ has a scale-invariant k. using that ~ and vk are uncorrelated (< ~ vk >= 0).e. p l Ylm ( ) = (2l + 1)(l .5 Doppler peaks F( )= 1 l XX where Ylm are the spherical harmonics related to the associated Legendre functions.171) is for R = 0 given by D E D E D E (2. (2. since the observations refer to various angles. We now include the Doppler e ect. Thus. l=0 m=. it is most convenient to perform an expansion in spherical harmonics.166) ~ vk (k) = (1 + 3R)vs(k=k) ~ sin(vsk ): (2. From an observational point of view. In general.172) therefore shows oscillatory behaviour with respect to k (and therefore also the quantity l occuring in the expansion in spherical harmonics). We shall not describe the technology needed to do this.175) Averaging. (2.l alm Ylm ( ) (2. using that the mean of the angular average of a vector gives a factor 1/3 2 because of thepthree directions available. D E D E D E D E (2. corresponding to the second term in (2. (2. In the comparison with observations. (2. From the continuity eqution we have _ ~ vk (k) = i(k=k2 ) ~. We need the velocity of the compressional modes. since k is the variable which enters in the Fourier transform.171). we have that k 1/distance. On the other hand.

so an increase in n will increase the slope of the oscillations. alternatively. It turns out that eq. Using just the simple expressions (2. this is an unrealistic expression. one needs to take into account the forcing F in eq. and since it has not been taken into account that R = R( ) depends on time. we obtain D j (k 1 R R2 + 9(1++3R) 8 1 0 193 0 < = k A . since a lowering of H will delay matter-radiation equality. This opens up very fascinating perspectives for the future of cosmology. (2. in the extreme relativistic limit there are no peaks! Therefore we expect a large sensitivity of the peaks with respect to R and other parameters.166) and (2. The European satellite \Planck" is expected to provide the necessary accurate data around the year 2005. do not cancel out. If R 6= 0. it is obvious that the peaks depend on the (power) spectrum of through the k. (2. corresponding p to vs = 1= 3.166) and (2. Of course. Usually the power spectrum < ~2 > is parametrized by a power kn. respectively. Thus. presumably the books on cosmology must be somewhat revised! . Also. (2. will be sensitive to the baryon contents of the universe. and then perform detailed numerical studies of the dependence of the peaks on H R :::. The peaks are also sensitive to the value of the Hubble constant. Increasing the baryon contents makes the peaks higher. the variable l).158). The result of such studies is that the peaks are sensitive to these parameters to within a few per cent.158) can be solved rather accurately by use of the WBK method.2. (2. the cosine and sine oscillations coming from eqs. After that. but we shall not discuss this. occuring when R > 0.175).158) has not been fully included.178) 5 1 + 3R + (1 + 3R)R cos2 @ q r : 3(1 + R) 3(1 + R) r E D E )j2 = j ~j2 " which clearly shows that there are temperature uctuations as function of k (or.175). since the forcing F ( ) on the right hand side of eq. 6R(1 + R) cos @ q r k A(2.13 APPENDIX: DENSITY FLUCTUATIONS 87 then all peaks disappear! Again we remind the reader that in these calculations we took R = 0. so oscillations are produced! It is therefore clear the resulting peaks. To see the e ects of the various cosmological parameters.dependence of ~.

88 COSMOLOGY .

3) The total work we have to perform. Then the rst term in the integrand gives zero.V 0 (x) (3.6) L(x x) = T . mx x .1) If we force the particle follow a de nite path x(t) which di ers by an in nitesimal amount x(t) from xcl (t) we have to use a force F (x t) = mx + V 0(x): (3.Chapter 3 Gauge theories 3. The particle will follow a path xcl (t) according to Newtons second law: mx = .2) (3. xV 0(x) = (T . The second term ] in the integrand can be expressed in terms of the variation of the kinetic energy T = mx2 =2 and the variation of _ the potential energy V (x): mx x .V 0(x). such that the force F (x) acting on the particle is . and the integral _ Z t2 _ S x( )] = dtL(x(t) x(t)) t1 89 . labelled by a coordinate x.1 The action principle 3. is given by Z t2 t1 dt xF (x t) = Z t2 t1 d _ _ _ dt m dt (x x) . To simplify things we assume that the dimension of space is one.e. V ) _ _ The quantity (3. i.1.4) (3. moving the particle along the path x(t). This means that we can de ne a potential energy V (x).5) (3. x(t) = 0 for t = t1 and t = t2.1 Classical mechanics Consider a classical particle moving under the action of a conservative force. xV 0(x)] Let us now assume we force the particle to move in such a way that the path x(t) at times t1 and t2 coincides with the classical path xcl (t). V _ which depends only on x and x is called the Lagrangian.

A great advantage of the action "principle" is the easy with which one can change from the cartesian coordinates xi entering in Newtons law to more general coordinates qi .10) is stationary for the path qicl(t) followed by the classical dynamical system. Then a partial integration of the rst term in (3. F x: (3.mx .8) We will again assume that the variation of the path is such that the end points are kept xed. V is calculated from the kinetic and potential energy T andV . But ]=0 is nothing but Newton's equation of motion. V 0 (x) x): _ _ (3. Notice that the action is a functional of the path x(t). the classical equations of motions can be derived by the requirement of stationarity of the action. and to systems consisting of many particles.3) we now see that Z t2 S x( )] = . V) = Z t2 t1 (mx x .7) This shows that the classical solutions to the equations of motions are stationary points of the action under appropriate boundary conditions (namely x(t) = 0 at the end points) Conversely. We can as above convert this requirement into a di erential equation by considering in nitesimal variations of the path q ! q + q. changing from cartesian to polar coordinates) the path singled out by the classical equations of motion will still be an extremum of the action S q( )].9) must vanish.9) If we impose the requirement that the variation of the action vanish for all in nitesimal variations x(t) of the path x(t) where the end points are kept xed. Since the lagrangian does not change its value under a change xi ! qi (i. V 0(x)] x(t) (3. where the end points are xed: qi (t1) = qi (t2) = 0.11) (3. From (3.e.8) gives no boundary terms and we are left with Z t2 S x( )] = . We have S = = In addition Z t2 t Z 1t2 t1 dt L(qi q_i) # " @L q + @L q_ dt @q i @ q_ i i i (3. expressed in terms of the generalized coordinates qi and generalized velocities q_i .12) d d d q_i = dt (qi + qi) . In this way we can derive Euler-Lagrange's equations of motion for a dynamical system described by generalized coordinates q = (q1 ::: qn) and a lagrangian L(q q_). The above derivation can clearly be generalized to a particle moving in a higher dimensional space. where L = T . we conclude that the bracket ] in (3. dt (qi) = dt ( qi ) . In fact. The action principle states that the action t1 S q( )] = Z t2 t1 dt L(q q) (3. the change in the action S when we vary the path from x(t) to x(t) + x(t) is given by t1 S x( )] = Z t2 t1 ( T .90 GAUGE THEORIES is called the action.

1 CLASSICAL MECHANICS and a partial integration of the last term in (3.13). dt @ q_ qi : (3. We know of course that this energy is conserved whenever the particle is "free". L(q q_i) (3.18) we call the dynamical system a canonical system. the motion of which can be described by n generalized coordinates qi. where we are not assuming any quadratic dependence of q_i in L.e. we have of course 1 (3. @L = 0: (3. Let us now emphasize the role of E by introducing the hamiltonian H (q p) of the dynamical system described by L(q q_). @L @q @q i i . S = 0 for arbitrary variations qi .3. or q_i = q_i (qi pi) (3. L(q q_i(q p)) (3. For such a system we can introduce the hamiltonian H (q p) E (q q_(q p) = q_i (q p)pi . i = 1 :: n.13) t1 i i Stationarity of S. To the generalized coordinates qi we introduce the corresponding generalized momenta: ( (3.17) pi = pi(q q_) @L@ q_ q_) : qi In the case (3. even in the completely general case. it follows immediately from Euler-Lagrange's equations (3.14) that E (q q_) is conserved in time when q(t) is a solution to (3. If the n equations (3. can be solved for the n generalized velocities: Z t2 " @L d @L # dt @q . i. However.19) With these de nitions it is easy to check that @H = q_ (q p) i @pi @H = .16) L(qi q_i) = T (q_i) . i. V (qi) 1 and E = 2 mq_i2 + V (qi) is the energy of the particle.e.16) where the qi 's are cartesian coordinates it is clear that pi = mq_i are just the ordinary momentum components of the particle.15) qi In the case where the qi 's are cartesian coordinates and we consider a particle of mass m moving in a potential V (q). follows the classical equation of motion.14) dt @ q_i @qi Let us now consider the quantity ( E (q q_) = q_i @L@q_ q_) . leads to ]=0 in (3. Suppose that we consider a general dynamical system. V (qi) = 2 mq_i2 .17) which de nes pi as functions of the qi 's and q_i 's.14).11) leads to _ S q()] = 91 Euler-Lagrange's equations of motion: d @L .

This eld can (at a xed x time) be viewed as a map: ~ ! r (t ~ ) x x (3. is speci ed by a eld r (t ~ ). The action can of course be described in terms of the new variables and since L(q q_) = q_ip(q q_)i . H (q p(q q_)) (3.22) we have Z t2 S q( ) p( )] = dt(piq_i(q p) .21). @H _ @q i (3. H (q p)) (3. t1 3.23) where the action is now a functional of paths (q(t) p(t)) in phase space.24) from space to the set where r takes values. (3. leads to the hamiltonian equations of motion. In each point of space one or more physical quantities are de ned and in the quantum theory the continuum of space-time leads to numerous problems as will be described in the second part of this course. In this way the r (~ ) are identi ed with our generalized coordinates qi .25) . In the simplest case of a scalar eld.21) which de ne a path (qi(t) pi(t) in the 2n-dimensional space of q's and p's. the set (~ r) playing the role of the indices i i = 1 :: n. The dynamics x of the classical systems considered in the last section was speci ed by the path qicl (t) in con guration space.20) and (3. elasticity etc. which was an extremum of the action under certain xed boundary conditions. where the number of degrees of freedom is in nite. In the hamiltonian formalism this space is called the phase space and it describes the possible dynamical states of the classical system.92 GAUGE THEORIES and we see that lagrangian equations of motion are equivalent the hamiltonian equations of motion: q_i = @H @pi pi = . In the same way a eld r (t ~ ) de nes a path x t ! r (t ) (3. It is easy to check that the action principle. often denoted target space. this is just the the set of real or complex numbers. However. We shall discuss a number of such theories later. Let us rst de ne what we mean by a (local) eld theory. A classical con guration of the system at a given time. demanding the action to be stationary for arbitrary in nitesimal variations of the path where the end points are kept xed. Classical physics is in addition to classical mechanics of systems with a nite number of degrees of freedom also the eld theory of electromagnetism. r can also be a vector taking values in Rn or C n if the discrete index r takes values from 1 to n.20) (3.2 Generalization to eld theory One great advantage of the above (lagrangian) formulation of classical mechanics is the ease with which it allows generalization. For electromagnetism we would identify x r with the vector potential A . t. At the classical level the lagrangian formalism allows us to deal with the in nite number of degrees of freedom in a straight forward manner.1.

In the following we will be interested in relativistic invariant eld theories.31) @ r @ (@ r ) V R3 t1 . or target space. a functional on the elds.26) In this case the lagrangian will itself be a functional of the eld con gurations (contrary to the situation in the last section). t1 V ) is the integration region. r x As was the case when we discussed classical mechanics of a nite number of degrees of freedom. The simplest way to construct such theories will be to take the lagrangian density to be an invariant under Lorentz transformation.30) = @ r r Here. Z t2 S r ] = dt L r (t ) _ r (t )]: (3. cl(t ~ ). the action will be given by a time integral over the lagrangian. @ @ L = 0: (3. which should be speci ed at . is the solution to the equations of motion.28) and we can now write the action as Z S r ( )] = d4x L( r (x) @ r (x)) (3. In the following we will x alway use the form (3. @ r (x) = @ r (x)] and performed a partial integration assuming that r (x) vanishes at the boundary . and we will now derive the classical equations of motion from the action principle which we will assume to be valid: For the given eld theory there exists an action S ]. but will depend only on the elds at a given time: Z L r (t ) _ r(t )] = d3x L( r (t ~ ) @i r (t ~ ) _ r (t ~ )): x x x (3.3. In the case where the elds r transform as scalars under Lorentz transformations we have: L = L( r @ r ) (3. such that the extremum of the action. in the second term we used: @ r (x)] = @ r (x) + r (x)] . This gives the Euler-Lagrange equations of motion: @ L .1 CLASSICAL MECHANICS 93 in con guration space.29) where x (t ~ ) and = (t2 . We now work out the variation of the action under an in nitesimal change of the elds: ! r+ r r # Z " @L @ L @ (x) d4 x S r] = @ r r (x) + @ (@ r#) r Z " @L @ . restrict ourselves to theories which at most contain rst order time derivatives of the elds (in order to get at most second order time derivatives in the equations of motion) the functional form of L is quite restrictive.27) Here V denotes a xed volume in and @i the partial derivatives with respect to the space coordinates xi. The function L is called the lagrangian density.29) for the action since Lorentz invariance is manifest except for the boundary conditions. If we in addition. as was the case in the last section. The action principle states that S has to be zero for arbitrary eld variations r . @ f @ (@ L ) g r (x)d4 x: (3.

and is given by H r (t ) @i r (t ) r (t )] = Z V d3x H( r (t ~ ) @i r (t ~ ) r (t ~ )): x x x (3.37) The energy-momentum tensor satisfy the continuity equation. Let us nally.32) of r that all terms cancel and @ T = 0: (3. It is.36) In order to prove the conservation of energy we calculate the divergence of T @T = .(@ r) r+ L (3.38) is a four-vector. It follows from the Euler-Lagrange equations (3.33) Contrary to the lagrangian density. It is evident that the hamiltonian density is the 00-component of this tensor: H = T00 : (3. . for reasons to become clear.34) One can now show that H is conserved in time by any eld con guration r (t ~ ) which x satis es Euler-Lagrange's equations of motion (3.27). (@ @ L @ (@ r+ @ (@ r ) r) ) : The terms in the bracket arise from di erentiation through the eld variables in L.32) 0 (t ~ ) @ (@ ) r H( r @i r r) = _ r r . L( r @ r ): (3.(@ @ r ) r .94 GAUGE THEORIES and the corresponding hamiltonian density Corresponding to the lagrangian density L we can de ne a conjugate momentum (t ~ ): x r (t ~ ) r x r = @L x (3.31) and the de nition (3. d3x@ i T0i = 0: V (3.35) which by construction transform like a tensor of second rank. as was the case for the lagrangian (3. Rather than doing that.39) provided the elds vanish su ciently fast at the boundary of V .(@ @ ( L @ = . Further the tensor nature of T ensures that Z P = d3xT 0 (3. H is not Lorentz invariant.31). check that the four-momentum is conserved for eld con gurations which satis es the Euler-Lagrange equations: V @0P = Z V Z d3x@ 0 T00 = . as promised. called the energy-momentum tensor. The hamiltonian of the system is the functional of the elds and the momenta r at a given time. Since P0 = H we can identify it with the energy and momentum of the dynamical system. @ r @ r + @ L @ r r) r )@ r +@ r . it is easier to generalize the problem: Let us de ne the quantity T = .

1 CLASSICAL MECHANICS 95 Let us end this section by working out some details in the case were the lagrangian is that of a single real eld and has the form 1 L( ) = . while the rest is various potential energy terms. which is divergence free ( @ t = 0) independent of whether or not the eld satis es the equations of motion.40) 2 We have = @ (@ L ) = .45) k0 = ~ 2 + m2 ~ (t ~ ) = e k x and the general wave solution is a superposition (t ~ ) = Re x Z d3k a. The Euler-Lagrange equation for the lagrangian (3.41) 2 2 T is symmetric as is required by general arguments (recall the discussion in general relativity).44) Later in this chapter we will discuss various potentials V ( ) where one can solve (3. The total energy will be Z E (t) = d3xE ( (t ~ ) x (3. x k + x k k (3. since is involves the time derivative of the eld ( the analogue of q_i).3.@ @ and from the de nition of T as we get T ( ) = .@ + L 1 @ @ + 1 m2 2 + V ( ) : T ( ) = +(@ )(@ ) .40). Let us here assume that the potential V ( ) = 0. which in more complicated theories will di er from the T de ned above by a tensor t . 2 @ @ . It is su cient for the lagrangian (3. The plane wave solutions are q i~ ~ ik0 t kx k (3.(~ ) ~ (t ~ ) + a+ (~ ) ~ (t ~ )]: k . (3. The energy density is given by 1 (3.43) As we have seen E (t) is actually independent of time if (t ~ ) is a solution of the equations x of motion.V 0( ) (3. but it should be mentioned here that the de nition of T given above does not ensure the symmetry. 1 m 2 2 .46) By the particle{wave duality in quantum mechanics this describes the propagation of a particle with mass m.44) with special boundary conditions. In that case (3. V ( ) (3.42) T00 = E = 2 (@t )2 + 1 (@i )2 + 1 m2 2 + V ( ) 2 2 The rst term can be viewed as the kinetic part of the energy density.@ @ + m2 ) = .44) is the wave equation for a free particle of mass m. .40) is (. but in general one would have to use the so{called Belinfante-Rosenfeld de nition of T .

47) could be one with the following lagrangian density (written in Minkowskian space-time): L( ) = . The abelian O(2) group can be viewed as describing rotations plus re ections in the plane and it acts on two-component real vectors. V ( t ): For O(1) the eld is just an ordinary scalar eld. Nevertheless we can write the action on in a way similar to (3. Group elements g may be represented by N N orthogonal matrices O(g). transforming in a de nite way under the action of certain symmetry groups1.@ y@ .2 Internal global and local symmetries GAUGE THEORIES The eld theory models relevant for Nature seem to involve elds which can be grouped into vectors.e.47) The unitary matrices are precisely the ones which leave invariant the scalar product ( a) Consequently.e. where Ot denotes the transposed matrix.50) The higher U (N ) and O(N ) groups are all non-abelian.96 3. Let us consider the group U (N ).1 = O(g)t. they have group elements which do not commute. it is worth to recall that no fundamental scalar elds have yet been identi ed by experiments.49) L( ) = . and the action on an N -component complex vector will be as (summation over repeated indices): ! U (g) ( y a ! U (g)ab b) a: (3.48) As another example of internal symmetry groups let us mention the orthogonal groups O(N ). They act on N -component real vectors.1 = U y. matrices satisfying O(g). i. and are characterized by the fact that they leave invariant the scalar product t a a In this case an invariant lagrangian can be written as follows: 1 (3. We have SO(2) = U (1) since U (1) acts in the same way on complex numbers = 1 + i 2 when we write them as two-component vectors. 1 . For any hermitian N N matrix H we have that U = exp(iH ) is a unitary matrix. an action which is invariant under (3. matrices satisfying U . 2 @ t @ . These symmetries will then be symmetries of the lagrangian. A typical example to have in mind is one where the the eld components transform according to the unitary group U (N ). For U (1) the group elements g are just phase factors acting on by ordinary multiplication: ! U (g) = ei (g) (3. i. A lot of fundamental fermionic particles has indeed been observed. However. we could as well have used fermion elds.50) by introducing the Lie algebra of the group. i. V ( y ) (3.e. The elements g of the unitary group may be represented by N N unitary matrices U (g). as far as the discussion below only is concerned with general transformation properties of matter elds and the introduction of gauge elds and local invariance. The space Although we will in the following discuss the transformation properties of scalar elds. The subgroup which only describes the rotation is denoted SO(2).

3. The Lie algebra of the group O(N ) is the space of all purely imaginary antisymmetric matrices2.56) It would indeed be better in this case to drop the factor i in the de nition (3. while the Lie algebras of SO(N ) are identical to those of O(N ).51) where the real numbers a(H ) can be viewed as the coordinates for H in the given basis.48) can be written as L0( ) = . it is the closure under the commutator operation which make the hermitian matrices a Lie-algebra. SO(N ) and O(N ) di ers by a discrete group of re ections with determinant -1. The Lie algebras of SU (N ) consist of the hermitian. . It is alway possible to chose the generators "orthonormal": Tr T a T b T a T b = ab : (3. V ( ) 2 (3. a gauge eld.52) which has some similarity to (3.47) is ! ei a (g)T a (3.50). as given in (3.3 ABELIAN GAUGE THEORIES 97 of hermitian matrices form the Lie algebra of the unitary group. traceless N N matrices. Since it is a vector space (of dimension N 2 ) we can choose a basis T a and write H = a(H )T a (3. It is convenient to start with a separate treatment of the abelian case.53) In fact. For later us we record that the basis vectors T a will satisfy a relation T a T b] = icabc T c (3. If we insist that the theory should be invariant under these local transformation we are lead to introduce a new eld. in order to get a uni ed treatment we keep the i. The "di erence" between the U (N ) and SU (N ) is roughly speaking the U (1) subgroup where all elements are phase factors times the unit matrix. They are characterized by the constraint that the matrices representing the group elements have determinant equal one. 3. where the group is just U (1).55) A symmetry of this kind is called a local gauge symmetry. The above symmetries can be made local. in which case the Lie algebra would consist of all real antisymmetric matrices.52). The T a's are called the generators of the algebra and the constants cabc the structure constants. With this notation the action of the group element g on . However. The term @ y@ is no longer invariant under such transformations.3 Abelian gauge theories For the abelian group U (1) the invariant lagrangian (3. It commutes with the rest (SU (N )) and is in this way "trivial".(@ ) (@ ) . By local we mean that the group elements entering in the transformation (3.47) depend on the space-time point x: (x) ! U (g(x)) (x) (3. which compensates for the derivative terms acting on the transformation matrix U (g(x)).54) Associated with the groups U (N ) and O(N ) are important subgroups called the special unitary groups SU (N ) and the special orthogonal groups SO(N ).

The way we have introduced it in (3. 4 F F where F = @ A . If we demand invariance under local gauge transformations (3.jkjt) ei kiei = 0: (3.55). In the case of free electrodynamic. If we can not ignore the eld the conclusion of (3.63) tells us that only transverse polarized free photons propagate. This leads to only two dynamical degrees of freedom for A .58) we can choose A0 = 0. uniquely determine the possible kinetic term of the gauge eld to be 1 L(A ) = . V ( ) (3. .62) The most immediate consequence of local gauge invariance is that it seems to imply that the photon is massless: A term m2A A is not gauge invariant.57)-(3. but we still have the following invariance left: Ai(x) ! Ai(x)+ @i (x) where (xi ) is independent of x0 = t. where we can ignore .61) it has no dynamics of its own since there is no second order time derivative acting on it.60) We say that D transform covariantly and the following lagrangian will now be invariant under local gauge transformations: L = . iA it is easy to check that the transformation of D (x) under a local gauge transformation (3. which transforms together with (x): (x) ! ei (x) (x) A (x) ! A (x) + @ (x) If we introduce the covariant derivative (3.61) The eld A (x) is nothing but the photon eld.58) is as follows: D (x) ! ei (x) D (x): (3. The requirement of invariance under (3.59) D = @ .98 GAUGE THEORIES and it is invariant under (3. @ A (3. this can be achieved by the introduction of a gauge eld A (x).58) and the requirement that only second order derivatives should be present in the equation of motions.58) (3.57) (3. In fact: From gauge invariance (3.63) might not be true: The photon might acquire a mass by the so-called Higgs mechanism which will be discussed later.50). We can use this residual invariance to choose @i Ai(x)jt=0 = 0. we have the equations of motion: @ 2 A . @ @ A = 0 ) @0 (@i Ai) = 0 if A0 = 0: It follows that @i Ai = 0 at all t if we x it to zero at t = 0.(D ) (D ) . We are left with: @ 2 Ai = 0 @i Ai = 0 ) Ai = ei(ki xi.63) (3.

Tdx (x) (3.65) ' ( (x) + dx @ (x)) . Let A0 be the gauge transformed connection: By de nition (x) ! ei (x) (x) 0(x) in every point x and we have the following diagram: Tdx 0(x) (1 + idxA0 (x)) 0(x) = (1 + idx A0(x))ei (x) (x) ei (x+dx) Tdx (x) = ei (x+dx) (1 + idxA(x)) (x) ' ei (x) (1 + idx(@ (x) + A(x))) (x) From this equation we conclude: A0 (x) = A (x) + @ (x) (3. The obvious similarity between the above construction and the construction of covariant derivatives in general relativity extends further. iA ) (x) and conclude D = @ .4 Geometrical interpretation 99 In this section we will show that local gauge invariance has a lot in common with the coordinate invariance of general relativity.58).68) V = 2 da R V : k . The local gauge invariance (3. 1 (3.58) makes @ meaningless as a measure of the change in .64) = exp(i dx A(x)) (x) In the same way as in general relativity we de ne the covariant derivative by comparing the eld Tdx( )(x) translated from x to x + dx with the actual eld value (x + dx) in x + dx: dx D (x + dx) . Disagreement in result was the signature of a curved space and we derived the following formula for the di erence. Recall that the Riemann tensor measured the curvature of space by parallel transport of a (tangent) vector two di erent ways around a in nitesimal parallelogram.4 GEOMETRICAL INTERPRETATION 3.66) We can now derive the transformation properties of A under the assumption that Tdx (x) should be a eld obeying the gauge transformations in x + dx. Of special importance is the construction of curvature. iA (3.67) allows us to compare values of (x) in two neighbour points x and x + dx because it translates (x) to a quantity Tdx (x) which transforms in the same way as (x + dx).67) This is of course just the gauge transformation (3.57)-(3. We can only have a theory with meaningful derivatives if we have a rule connecting (x + dx) and (x) : A connection A which de nes translation T of : x ! x + dx = Tdx(x) (x) ! Tdx( )(x) = (1 + i dx A(x) + 0(dx)2) (x) (3. The derivative built from these two quantities will therefore also transform correctly: it is a covariant derivative measuring the \true" di erence between (x) and (x + dx). ( (x) + idx A (x) (x)) = dx (@ . but now derived from geometrical considerations: The connection A with the transformation properties (3.3.

iF (3. where da is the area element (dx x . The idea of iF as a kind of curvature is further corroborated by comparing the two formulas for commutations of covariant derivatives in gauge theories and general relativity: V . On each ber (which is a copy of V ) we have the action of the group U (1) and it is on the ber bundle that A acts as a connection and de nes parallel transport.1: The two ways to get from x to x + x + dx. As for the abelian elds we want to extend the symmetry (3.1).1 is given by: (x)PT T x Tdx (x) .70) Of course a direct calculation of the commutator shows that the curvature F is nothing but the electromagnetic eld tensor @ A . In order to answer this question one need the mathematical construct of a ber bundle.2. This is illustrated in g.5 Generalization to non-abelian groups In the non-abelian case the scalar eld will be a N component vector.100 GAUGE THEORIES Figure 3. dx x ) of the parallelogram (see g. 1 da D D ] (x) (3. 3. Given the base space Rd we can in every point x erect the vector space V to which (x) belongs (here V is just the complex numbers C ). From the de nition of Tdx it is easy to perform the same construction for gauge theories.48) is no longer invariant and that the derivative @ has bad transformation properties. Locally the space has the structure Rd V . 3.R V (3. let us only outline the idea.69) 2 If we denote the curvature by iF we have comparison with (3. @ A .V = .71) D D ] = . We nd that the di erence PT by the parallel transports shown on g. 3.72) At this point one could ask: in what space is F the curvature and A a connection. It is not the purpose to enter into a detailed discussion of this mathematical construction.47) to a local symmetry where the transformations U (g(x)) can depend on x.68) D D ] = . By allowing such local symmetries it is clear that the lagrangian (3. Tdx T x (x) = . 3. We want to repair it by the same recipe as was used in general relativity and .iF (3.

1 (g(x)) (3. As in the abelian case we have the diagram: (Tdx )0(x) = (1 + idxA0 (x))U (g(x)) (x) k or U (g(x + dx))Tdx (x) = U (g(x + dx))(1 + idxA(x)) (x) idxA0 (x) = U (g(x + dx))U . This requirement xes the transformation properties of A(x) exactly as in the abelian case.5 NON-ABELIAN GROUPS 101 Figure 3.78) It is often useful to consider in nitesimal transformations: U (g(x)) = ei a(x)T a 1 + i a (x)T a + O( 2) (3.1 (g(x)) This leads to the following transformation law for the connection A : A0 (x) = U (g(x))A (x) U .3.1 (g(x)) . Let us denote the gauge transformed of (x) by 0(x) and the gauge transformed of A (x) by A0 (x). for abelian gauge theories: the introduction of a connection and the concept of covariant di erentiation. In this case the connection A (x) will take values in the Lie algebra: A (x) = Aa (x)T a (3. i@ U (g(x))U . cabc bAc (x) + @ a (x) .75) where the group element g(x) can be a function of x.76) can in this case be written as A0 a (x) = Aa (x) .74) A vector eld (x) will transform as (x) ! U (g(x)) (x) (3. 1] + iU (g(x)) dxA(x)] U .2: path in the ber bundle Rd C de ned by the connection A .1 (g(x)) .73) and parallel transport by dx is de ned as Tdx (x) = (1 + idxA(x)) (x) ' eidxA(x) (x): (3. We now demand that the translated eld Tdx (x) transforms as a eld in x + dx.77) (3.76) (3.

igA and in the commutator of F ! @ A . ig A A ].86) g We have introduced the coupling constant between the gauge eld and the scalar eld as the relative coupling strength between the two terms in (3.86).102 If we de ne the covariant derivative as GAUGE THEORIES dx D (x) = (x + dx) .81) (3.82) (3. 1 da D D ] (x) 2 We conclude that the curvature.83) F =@ A . be physical observables. and we therefore get the following transformation law for F (3.1 (g(x)) We see that F belongs to the Lie algebra and we can (if needed) expand it in components: From the construction it is clear that F transforms covariantly (i. A lagrangian like (3.iA A ] F = Fa Ta F (x) ! U (g(x))F (x)U .80) (x)PT T xTdx .1 and we nd (3. by themselves. 41 2 Tr F F (3. @ A .@ A . This emphasizes that the . iA (3. (D )y (D ) .79) D D ] = . 41 2 Tr F 2 .and scalar elds then becomes: L = .iF Inserting the de nition of the covariant derivative D we get (3.85) g The total lagrangian for the interaction of gauge.48) will now be invariant under local gauge transformations provided we replace @ by D . By a scaling A ! gA it will appear in the more conventional place in the covariant derivative D = @ . Note however that in the nonabelian case the components are not gauge invariant and consequently they cannot.e. as itself). which we denote by . Tdx T x = .iF will be given by D = @ . V ( y ) (3.84) As in the abelian case we will call F the eld tensor. If we want to include the dynamics of the gauge eld A it is necessary to add a kinetic term involving the second order time derivative. Tdx (x) or D will by construction transform covariantly under a gauge transformation The curvature can be found as the di erence in parallel transport along the two path shown in g. 3. As in the abelian case this term is essentially unique if we want no derivatives higher than second order: L(A) = .

We then get a vector ( ) de ned as a function of . . We have emphasized the concept of covariant derivative and parallel transport and the connection to the same concepts in general relativity.88) where xi belong to the curve and xn = x( ). We can now solve (3. However. it is possible to nd a better representation of ( ). We have the following equation: D ( )=0 (3. By repeated application of (3. de ned by (3.93) d2 0 0 1 1+i Z (3.90) Note again the similarity of this expression with the one of parallel transport of a vector in general relativity: dV . dx V = 0 d d For a given eld (x) the covariant derivative along the curve x( ) is nothing but dx D (x) d and we see that the statement that (x) is covariant constant along a given curve x( ) means that (x( )) is obtained by parallel transport of a given vector (x( 0 )) along the curve.3. Example 1: Parallel transport (the integrated version) We have de ned the in nitesimal version of parallel transport by Tdx (x + dx) = (1 + idxA(x)) (x) eidxA(x) (x) (3.5 NON-ABELIAN GROUPS 103 non-abelian part of F gives rise to a non-trivial self-coupling of the gauge eld.89) D = dd . iA(x) dx d has to give zero when acting on ( ). which is not present in abelian gauge theories like ordinary electromagnetism.90). By construction. the covariant derivative along the curve.91) (3.87) Suppose we parallel transport the vector 0 along a curve ! x( ) from x0 by applying this de nition many times along the curve. The formal solution is (we state this without proof and invite the reader to verify it): (x( )) = g(C A) Zx(0)) ( g (C A) P exp i dxA(x) C C P exp i Z dxA(x) d dx A (x( )) + Z 0 Z d dx 1 i2 d 1 d 2 d A (x( 1 )) dx A (x( 2 )) + (3.92) . Let us end this section by elaborating on the concept of parallel transport.87) we get ( ) = nlim !1 " n Y k=0 eidxk A(xk ) # 0 (3.

we leave it as an exercise to show that I 1 (3. 3.104 GAUGE THEORIES In these formulas C denotes the curve ! x( ).92) is called the path ordered exponential.96) (3. Rather surprisingly the shape of the potentials relevant for a correct description of Nature is not the simple one of g.93) can be written as an ordinary integral P exp i Z Let C : x0 ! x0 denote a closed curve and (x0 )PT denote the vector (x0 ) parallel transported around the curve. If the curvature F is di erent from zero (x0 )PT will in general be di erent from (x0 ). If L denotes the linear size of the curve and da the area tensor.94) P exp i dx A(x) = 1 + 2 iF da + O(L3 ) C and therefore (in agreement with our former formulas for parallel transport) C dxA(x) = ei R C dxA(x) i (x0 )PT .1 (g(x0)) It is now possible to construct two classes of gauge invariant objects in the theory: Z C :x!y dxA(x) (x) dxA(x) (3. From (3. 3. If we for simplicity . but rather the one of g.97) (3.95) Let us nally address the question of constructing gauge invariant quantities. Although nice from this point of view. C Tr P exp i I 3.91) it follows that a gauge transformation (3. If the group is abelian A(x1 ) and A(x2 ) commute and (3.3b. The group element (3.99) "Tr" denotes the trace of a matrix. We use the notation of example 1. A characteristic feature of such potentials is that the classical ground state will be degenerate.1 (g(x0)) (3.3a. (x0 ) ' 2 F da (x0 ): (3. One can show that the above objects in the quantized theory constitute a complete set of observables in the theory.98) I I C C dxA(x) U . they are not alway convenient observables from the point of view of practical calculations.76) A(x) ! R A0(x) induces the following transformation on the path ordered integral P exp C dxA(x) along the curve C P exp i Z In particular we get for a closed curve C : x0 ! x0 P exp i C dxA0 (x) dxA0 (x) y(y ) = U (g(x( ))P exp i = U (g(x0 )P exp i P exp i Z C dxA(x) U .99) In (3.6 Spontaneous symmetry breaking and the Higgs mechanism Until now we have not been very speci c as to the form of the potential V ( ).

Note that V = SO(N )=SO(N . V the orbit of 0. However. while we for SO(2) have a theory which is equivalent to the U (1) theory. e. consider theories with a N -component real vector eld a(x) and internal symmetry group SO(N )3 we will assume the potential can be written as V ( t ) = 2 t + ( t )2: (3.g.e f j 9 g 2 G : = g 0g. The choice of one 2 V as the vacuum will break the SO(N ) symmetry of the theory. There might still be some invariance left.3. i.3: Two possible shapes of the potential V ( ).102) Let us rst state in qualitative terms what will happen and later illustrate it in the simplest cases of SO(2) and SO(3). 3. For the cases of interest here. It is therefore given by (x) = 0. 1) (all rotations orthogonal to the direction given by 0). 3. 1). The energy (in dimension d = 3 + 1) will be Z 1 E = d3x 2 @0 t @0 + 1 @i t@i + V ( t ) (3. and H the subgroup which leaves 0 invariant (the stabilizer of 0). will form a subgroup SO(N . will be vacuum states.100) 2 > 0 corresponds to the situation in g. 3 . We call the space of all vacuum states V .3b. where 0 is independent of space-time and a minimum of the potential V .101) 2 The classical vacuum con guration is one which corresponds to minimum energy. If the theory prefers one of such states as its vacuum we call it spontaneous symmetry breaking.3a. g 2 SO(N ). Some of the uctuations will always increase the Recall that for O(1) we have an ordinary real scalar eld. if 0 6= 0 (i.e. V will be independent of 0 . < 0) then every other eld con guration which we get by acting with the group SO(N ) will also be a vacuum: All states = O(g) 0. Then V = G=H . 1). 0 a vacuum state. while 2 < 0 corresponds to the situation in g. We then say that G acts transitively on the space of vacuum states. We say the symmetry is broken to SO(N . In the case of SO(N ) we see that all group elements which leave a certain vacuum con guration 0 invariant.6 SYMMETRY BREAKING/HIGGS MECHANISM 105 Figure 3. in the case of spontaneous symmetry breaking one should expand the elds around 0 : (x) = 0 + (x) (3. This is true in general: Let G be the symmetry group. In order to understand the excitations above the ground state one should expand around this ground state.

A glance on g. 2 . but instead some of the gauge particles. 1) = S N . but with one important change: there will be no Goldstone bosons. We summarize the above discussion by the statement that spontaneous breaking of a continuous symmetry is always associated with the appearance of a certain number of massless particles. These uctuations correspond to changes of 0 which are orthogonal to the space V of vacuum states. the so-called Goldstone bosons. Strictly speaking it is impossible to have a phase transition in a system with only a nite number of degrees of freedom.104) = ei 0 0 2 Because of the invariance of (3. 1 Goldstone bosons. and spontaneous breaking of symmetry may occur in quantum eld theory as well as in classical eld theory.e. but no Goldstone bosons will be present. They cannot have a term m2 2 in the lagrangian and must correspond to massless excitations. We will illustrate this so-called Higgs mechanism. From the arguments above it should be clear that the Goldstone bosons can only be associated with the breaking of continuous symmetry. the uctuations which follow the tangent plane of V cost no energy (by the very de nition of V ).3b reveals the reason: Since the barrier between the two vacua is of nite height there will always be tunneling and the wave function of the ground state will indeed be symmetric. If 2 < 0 the possible vacuum states are given by 2=. It is also worth emphasizing that in ordinary quantum mechanics. Example 2: Spontaneous symmetry breaking and the Higgs mechanism for U (1).106 GAUGE THEORIES energy. It can also be spontaneously broken. In the case of SO(N ) we get N .105) 2 (x) 2 L = . 1)=2 or from realizing that SO(N )=SO(N . . in the simple but very important case of the abelian gauge group U (1).103) under global gauge transformations we can assume that the chosen vacuum 0 is real (= v).1. A good analogue can be found the concept of phase transitions in statistical mechanics. They are the massive uctuations. the ground state will always have this symmetry. which only involves a nite number of degrees of freedom. One has to go to the "in nite" volume limit in order to have a sharp transition from one phase to another. ( . This follows either from the dimension of SO(N ) being N (N . We consider the lagrangian )2 (3. If we consider the situation where the internal symmetry group has been enlarged to a local symmetry one can still have the phenomenon of spontaneous symmetry breaking. the reason being the in nite number of degrees of freedom. i. equal the dimension of G=H (=dimG-dimH ). We now have to expand around the vacuum and we displace the elds: ! ! 1 v + p s(x) : (x) = 0 + (x) = 0 (3.103) If 2 > 0 the classical vacuum is = 0 and there is no spontaneous symmetry breaking. i. the ones associated with generators in the directions of V = G=H will acquire a mass. These massless excitations are called Goldstone bosons and we see that their number must be equal the dimension of the space V .@ @ .e they will induce terms like m2 2 in the lagrangian. In eld theory the situation can be di erent. In the case of O(1) we have only a discrete symmetry ! . However. although it from a classical point of view can be broken. (3. 3.

This is not true for the massive particle). Rather that doing this we can directly take advantage of the local gauge invariance of (3. 4 F F . We will get a kinetic term like in (3. It is remarkable that the massless Goldstone boson has disappeared and instead a mass term for the U (1) gauge eld has appeared. This is needed for the standard model. Under the gauge transformation (3. Next. 2 v(s2 + 2 )s .109) the A eld transform as (3.eld and . ( )2 (3.108) and for any (x) rotate (x) to a real value. Such a gauge choice is called a unitary gauge choice: p (x) ! ei (x) (x) = j (x)j = v + (x)= 2 (3. Let us now demand local gauge symmetry.105). and the photon is the gauge particle. it can nevertheless appear via the Higgs mechanism.106) and a coupling between the A . 4 (s2 + 2 )2 : 107 (3. Observe also that the degrees of freedom are unchanged: the degree of freedom associated with the massless Goldstone boson has reappeared as a longitudinal degree a freedom associated with the massive gauge particle (recall that the massless particle only had two transverse degrees of freedom.109) The gauge parameter (x) to choose is obviously just the phase of (x) . ieA ) . it is important even in non-relativistic physics. The Higgs phenomenon described in the above example plays an important role.106) (3. In a superconductor the Cooper pairs of electrons play the role of the condensate. We can now expand the eld as in (3. it is then possible to remove completely the massless (x) eld by a local gauge transformation depending on (x). 2 (@ )2 + 2 v2 2 ] + Linteraction (3. Inside the superconductor the photon is indeed massive (The Meissner e ect).107) In accordance with our general discussion we get one massive excitation with mass m2 = s 4 v2 above the vacuum (the s-"particle") and one massless excitation (the particle.110) A (x) ! A0 (x) = A (x) + 1 @ (x) e and by insertion in (3. After some algebra we get: L = . and introduce the associated abelian gauge eld A (x).3. (@ + ieA ) (@ . 2 v2 s2 + Lint 2 p Lint = .111) where Linteraction contains cubic and quartic interactions in and A0 . First it seems (almost) the only way in which gauge particles can acquire a mass and still provide us with a renormalizable theory with acceptable properties.108) we get the following lagrangian in terms of A0 : 1 1 L = .6 SYMMETRY BREAKING/HIGGS MECHANISM We can rewrite L in terms of the displaced elds. and we have expanded j (x)j around its vacuum value.108) If 2 < 0 we have the situation with spontaneous symmetry breaking (the Higgs mechanism). 1 (@ s)2 + (@ )2 ] . Aspects of this will be discussed in the the second part of this course. . Although it is forbidden by gauge invariance to put in by hand such a mass term. where gauge particles like the Z and the W have masses. The lagrangian will be 1 L = . 2 . However. which is the Goldstone boson). 4 F F + e2v2 A0 2 ] .

where (x) = 0.2. From the general arguments given above we expect the particle excitations in this spontaneously broken vacuum to be as follows: One massive Higgs particle. As in example 2 the excitations above the spontaneously broken vacuum are best revealed by using the unitary gauge: For any a (x) we use the local gauge invariance to make a rotation4 a (x) ! a3 ( 0 + '(x)).119) 2 = g2 2 mW (3. 4 F 2 . It is convenient to make a decomposition of the vector eld. m2 jW j2 .@ A (3. Let us now verify this by an explicit calculation.114) D a = @ a + g abc V b c : (3. as will be clear later. If we normalize the vacuum energy to be zero in the spontaneously broken phase we can write the invariant Lagrangian as follows: 1 (3. j j = 0 and V a (x) = 0 is a classical vacuum con guration. 1 (D )a (D )a . 2 m2 '2 (3. 8 ( a a . 1 (@ ')2 .116) 2 After a somewhat tedious. 1 jD W .115) Clearly a con guration where a (x) is constant. These elds transform under SO(3) transformations as described above.120) 0 2 = 2 mH (3. The generators themselves can be expressed in terms of the "-symbol: T a = i" a : ~ To each T a there exists a gauge eld V a (a = 1 2 3) which form an \isospin vector" V .3 and it is zero in all other cases. but we will later encounter 6 . 4 Ga Ga . @ V a + g abc V b V c (3.108 GAUGE THEORIES The next example of spontaneous symmetry breaking is slightly more complicated. two massive vector bosons. which respects this decomposition.112) where "abc is the usual "-symbol: it is 1 if a b c is an even (odd) permutation of 1.117) H 2 ~ ~ . since we will reserve A for other purposes.113) in terms of the new variables: 1 1 L(' A W ) = . In addition we have a spin-0 eld a which also transform like an isovector. Example 3: Spontaneous symmetry breaking and the Higgs mechanism for SO(3). We introduce: 1 A = V 3 W = p (V 1 + iV 2) (3. It is a gauge-Higgs model based on the non-abelian gauge group SO(3). igF W W + Lint (' W ) W 2 where F = @ A .118) ~ = @ . The model to be considered is the so-called Georgi-Glashow model. D W j2 . but elementary calculation we can express the lagrangian (3. situations where the condition cannot be met everywhere The group elements in SO(3) are 3 3 real orthogonal matrices of unit determinant. The three generators T1 T2 T3 satisfy T a T b ] = i"abc T c (3. igA D (3. 2 )2 0 2 Ga = @ V a . We denote the vector eld V rather than the usual A . and one massless vector boson. Due to the invariance under global gauge transformations we can without any further assumptions take the direction of this constant a to be the threedirection in isospin space.113) L( V ) = .121) 0 4 This can clearly be done in a contractible region of space.

by the replacement @ ! @ . iA . 2 (2g 0 ' + g2 '2 )jW j2 . the charge associated with the generator of the unbroken U (1) group. F denotes the dual eld tensor: 1 ~ F = 2" F and if we recall that the magnetic eld is given by (3.e. The eld con gurations will approach its vacuum values at in nity. and in this section we will consider static solitons only. Unfortunately we have no space to cover this interesting topic. for topological reasons. The remarkable fact is that they come out again as solitons after the collision. associated with the unbroken gauge group SO(2) U (1).e. 3. but the boundary conditions are such that no matter how we interpolate the elds inwards they have to di er from their vacuum values somewhere in the interior. It might not be possible to rotate globally to the unitary gauge if (x) has zeroes. . which are of topological origin. there are no monopoles. which can be viewed as monopoles and where (x) = 0 at the center of the monopole. In addition to this coupling we nd an anomalous magnetic moment coupling igF W W .7 Solitons The non-trivial vacuum structure associated with spontaneous symmetry breaking leads to many solutions of the classical equations. By translational invariance we can consider solitons which move. and therefore classical solutions. The reason is to be found in the complicated vacuum structure. and that the massive Higgs eld ' is neutral. The coupling is a consequence of the non-abelian group structure of SO(3) and has drastic and important consequences when the theory is quantized: It is responsible for asymptotic freedom of the theory. i. We see that their behaviour is very much like particles.123) Bi = 1 "ijk F jk 2 the other equation tell us that @i Bi = 0. g4 (W W . and since they are localized we can collide two of them. We conclude that A is a massless vector eld. We call them solitons.3. W is a massive charged vector eld. The theory contains solutions the the classical eld equations. which allows us to specify non-trivial boundary conditions at in nity. 8 (4 0 '3 + '4 ) 2 109 The equations of motion for the A eld is as follows: ~ @ F =j @ F =0 (3. which can be viewed is the electromagnetic current induced by the charged W -particles. Notice nally that the massive charged vector boson W is not coupled to electromagnetism by the minimal coupling. W W )2 . are stable under variational attempts to lower the energy further. It satis es the usual current conservation ~ @ j = 0. i. In this way we can view the Georgi-Glashow model as electromagnetism embedded in a larger non-abelian group (SO(3)). They will be local minima of the classical energy. The loop hole appears if (x) = 0 at some points.7 SOLITONS and 1 Lint = . In this way it is possible to get "lumps" of energy which.122) where j (A W ) is a complicated expression. In the next section we shall see that this is a wrong conclusion.

2 2 0 (3.7. If the 2 term is negative in the lagrangian. If the theory is de ned in three space dimensions we assume that the eld (x y z) is independent of y and z. The eld theory which eventually is going to extend the standard model will probably be a gauge-Higgs theory where some. The following three examples can be viewed as solitons in one.3b and we have spontaneous symmetry breaking. re ection symmetry ! . i.125) (3. If we normalize the energy to be zero in the (broken) vacuum state.1 Domain walls E=1 d 2 dx i. 3.128) 0 dx2 . of these solitons can be present.127) To nd this solution one can solve the equation of motion. but they are all important in cosmology in a negative sense. there exists a (x) with the property (x) ! + 0 for x ! +1 (x) ! . 2) : (3. i. is \spontaneously broken". In the following we will assume that (x) is a function of only one variable x.1 : (3. Further some of them may act as seeds for galaxy formation.: !2 + 2. 0 for x ! .126) : Thus there are two ground states with degenerate energies.e. two and three dimensions.e. namely that of O(1). This is the reason for the name "domain walls".124) E is a sum over squares. d = 0 and dx with the trivial solution = const. which in this case is a nonlinear equation: d2 = 4 ( 2 . Since they are heavy (as we shall see). and = 0 2 = 2 0 (3. O(2) and O(3). if not all. Whatever structure is present in the x-direction therefore extends in the whole y . z plane. At the same time they have increasingly complicated internal symmetries. Let us consider a 4 theory with O(1) symmetry. they can even in uence the density of the universe in an unwanted manner. and the symmetry $ . we have the situation shown in g. vortex solutions and monopole solutions. There exists a classical solution which interpolates between these vacua. we have the following density in one dimension 3. The two rst types are quite important in solid state physics too.110 GAUGE THEORIES and give three examples of such solitons: Domain walls. We get the lowest energy by requiring that each square vanishes.e. .

Z d p = 2 x + k (k = const:) 2. since E is given by a Q-term ( xed) plus a term which is an integral over a square.133) k e2p2 0x . 1 d + p2 ( 2 . 1 (k = const:) = 0 2 2 0x (3.135) 0 3 where m denote the lowest mass excitation above the vacuum = 0. when x goes from . 2 0 0 2 . The minimum for E respecting this boundary condition (i. 2) : (3. p2 d 3 . Also notice that other con gurations which satisfy the same boundary conditions obey the inequality E Ewall : (3. 3.p2 ( 2 .4.1 .129) 2 p Z +1 d p Z +1 d 3 E= dx E = 1 dx dx + 2 ( 2 . The ground state is either spin up or down. which is m2 = 8 2. . 3. The quantity Q is called the topological \charge" and is Q = (+1) . and it is simpler to use a trick (due to Bogomol'nyi).1 .e. 0 From this point of view the soliton is heavy and non-perturbative in the coupling constant .1 " # 2 p2 2Q + 1 Z +1 dx d + p2 ( 2 . 2 dx dx 3 .7 SOLITONS 111 This is a bit complicated. E = 2 dx 0 dx 3 The integrated energy is " ! # 2 0 ! (3. 2 0 Z +1 " # ! = from which p 1 ln 2 0 0+ 0.1 to +1. The soliton which interpolates between the two ground states then looks as shown in g. (.1) = 2 0 : (3. We attempt to write E as a sum over squares and total derivatives in a di erent way.132) 0 dx This equation can easily be solved by standard methods.134) +1 ke which satis es ! 0 for x ! 1. (3.130) 0 0 2 . so d = . The energy is given by p 1 3 Ewall = 2 2 2Q = 12 m : (3. 2) 2 = 3 (3. The reason is given in g. 2) . A similar situation occurs in ferromagnetic spin systems.3.136) This solution is called a domain wall. 2) 2 .131) Q thus measures how we go from one vacuum to the other. respecting that Q 6= 0) is now trivial.1 dx where we used that the second term in the energy density is a total derivative and hence can easily be integrated by use of the boundary condition that ! 0 for x ! 1.5 the \topology" thus changes.

138) to (x) = ei (x) 0 Ai(x) = @i (x) (3. In that 0 3. y plane.and the Higgs elds are functions of the two space variables x and y only. This is why we talk about vortices and strings.112 GAUGE THEORIES Figure 3.139) and we assumed that it was possible by a unitary gauge choice to transform away the phase factor (x).2 Vortices and strings = ei 2 0 =. which was the Goldstone boson). The lagrangian we have in mind is the usual U (1) Higgs lagrangian 1 L( ) = . ( )2 (3. it will extend in the z direction. 2 . We will assume that the gauge. In this section we consider the U (1) Higgs model. (D ) D .5: A domain wall between two domains with di erent spin orientations. 4 F F .137) Some aspects of spontaneous symmetry breaking was discussed in example 2. In fact it was obviously possible if (x) 6= 0 everywhere. 2 .138) 2 and it was broken. If the system is considered as de ned in three space dimensions this means that whatever structure is present in the x . The particle spectrum after the spontaneous breaking consisted of one massive excitation (the s-particle) with m2 = 4 2 and one massless excitation (the s 0 particle. If we included the gauge eld the invariance of the vacuum is enlarged from that of (3.4: A domain wall Figure 3. Recall that if there was no gauge eld the possible vacuum states for 2 < 0 were given by: (3.7.

the only condition is that it should change by 2 integer when we go around in a circle. case the uctuations around the vacuum (x) = 0 . (x) is single valued. One has to check this explicitly.6: If we start at point "P" and move anti-clock wise round. They are not necessarily physical acceptable. As boundary condition we take ! 0 ein for r ! 1 : (3.142) 0 and we observe that a necessary condition for nite E is Di (x) ! 0 for r ! 1 If we combine this with the boundary condition (3. but the phase need not be. where we plot the complex number going around at "r = 1". 3. (x) goes back to its original value.141) we get (3.144) eAi ! n@i (x y) for r ! 1: .139).3.141) This is a topological condition in the sense illustrated in g.143) (3. while the Goldstone boson was "eaten" by the vector particle which became massive with a mass mV : m2 = 2e2 2 m2 = 4 2 (3. but its phase changes by 2 . 2)2 (3. Again it should be stressed that it is the degeneracy of the vacuum. is the requirement that the energy relative to the vacuum ( (x) = 0) should be nite.6. Since we are in two dimensions it is convenient to introduce polar coordinates r .140) V 0 H 0 We shall now show that it is possible in two-dimensional space to impose topological nontrivial boundary conditions of the form (3. which makes it possible to impose such non-trivial boundary conditions.7 SOLITONS 113 Figure 3. An additional condition to be satis ed if con gurations with such boundary conditions should be interesting. Ai (x) = 0 resulted in a massive excitation. The static energy relative to the vacuum is given by Z 1 2 E = dxdy 4 Fij + jDi (x)j2 + ( . the Higgs particle with the mass mH equal to the mass of the s-particle.

In this sense these sectors are "topological".141) plus the requirement of a nite energy relative to the vacuum leads indeed to a vacuum condition like (3. Let us only note that the non-trivial ux conditions translates into a topological classi cation of the U (1) bundles on the boundary S 1. and this -function (a so-called Dirac string) would exactly maintain the ux quantization (3. y-plane: for r ! 0 the phase still has to change by 2 when one goes around a small circle. i. only determined by the requirement that Di = 0. But if we try to do it all the way round the circle at in nity. Thus. It should nally be mentioned that the gauge eld so far has been a kind of spectator in these topological considerations.114 GAUGE THEORIES At r = 1 Ai(x) is thus a pure gauge. Recall that if the gauge group G acts transitively on V we can write (x) = g(x) 0 and in this way we get an identi cation V ' G=H .141)and (3. For the case in question the map will be from S 1 ! S 1 . mentioned earlier. but it is quantized (use Stoke's theorem) Z I (3. classes where the maps can be continuously deformed into each other. However. winds around n times as one move around a circle at in nity.145) Flux = F12 dxdy = Ai dxi = n ( (2 ) .145). It cannot hold everywhere in the x . to formulate the topological aspects one need the notation of ber bundles.e. in contradiction with (3. has a drastic consequence. which again can be identi ed with S 1 in an obvious way. i. Let the asymptotic condition for be (x y) = 0 ei (x y) . associated .145) is a topological invariant on each class of bundles. However. They play in fact an equally important role of their own.146) These mappings falls in di erent homotopy classes. However. we encounter a discontinuity when returning to the starting point. the boundary conditions are classi ed by a mapping from the boundary of space (here the circle S 1 at in nity) into the vacuum manifold V . From a topological point of view. and such maps are from the point of view of homotopy characterized by their winding number.139) at in nity. Let us now consider a sector where the magnetic ux is di erent from zero. where H is the subgroup which leaves 0 invariant. (0)) = 2n : e e circle at 1 The physical acceptable boundary conditions split in sectors characterized by a quantized magnetic ux and there is no way in which a con guration from one sector can be continuously deformed into a con guration of another sector if we restrict ourselves to con gurations having only nite energy relative to the vacuum. which seems to contain no information about the ux in the interior.e. Locally at in nity we could gauge the boundary conditions (3. and we conclude that a boundary condition like (3. We see that the di erent quantized ux sectors correspond precisely to this topological classi cation. It is a beautiful and remarkable property of topology that one is able identify many of the "global" properties of the system just from the knowledge of the "vacuum structure" in nity. according to (3. The ux of magnetic eld (in the z-direction) is not zero. and that the ux (3.145). this requirement has no meaning and the phase gets confused and doesn't know what to do unless = 0 for r = 0. In our case G = U (1) and H = 1 and we can identify V with U (1). In summary the mapping is x 2 boundary ! g(x) 2 G=H: (3. if the circle shrinks to a point.141). the number of times one circle covers the other under the map. the fact that the phase of (x).144) to the trivial vacuum (x) = 0 and A (x) = 0.

For large r we must be close to the vacuum and quadratic uctuations around the vacuum values can be analysed. From our discussion a natural ansatz is as follows: (r ) = ei 0 f (r) Ai = "ijrxj a(r) (3. but it is not too di cult to solve them numerically by assuming rotational symmetry.147) where ~ is the current in the Maxwell sense.1. (@i .149) .3. ieAi )2 = 2 (j j2 . @k Ak j j2 (3. since otherwise the ux quantization cannot be R right ( Hz dxdy = 2 =e obviously requires Hz 6= 0) as is shown in g. 3. They are quite complex. 3. | It is not possible to solve the equations analytically. and in general there is a non-vanishing magnetic eld near r = 0.7 SOLITONS 115 Figure 3. One nds for r >> 1=mH and r >> 1=mW : Hz (r) j (r)j 0 + const:e const:e.7 So far we have not yet solved the equation of motions.7: The magnetic eld associated with a vortex with the topology at 1 (phase changes by 2 ) is the requirement that must vanish in a point (r = 0).mV r .mH r (3. @j Ai ! H for r ! 0. Since must change from r = 1 to r = 0 it follows that from the equations of motion that although Ai is a pure gauge at 1 it cannot remain so. and where f (r) and a(r) ! 1 for r ! 1 and f (0) = a(0) = 0. 2 ) 0 ) + 2e2 @iF ik = jk = ie( @k . The result is given qualitatively in g. Note that a(r) Hr for r ! 0 means that H (r) = @i Aj . This zero is called a topological zero and there are topological theorems which link the "winding number" of the Higgs eld at in nity to the zero's in the interior for much more general situations than the one considered here. It is easy to solve the equations numerically.148) where ij is de ned by 11 = 22 = 0 12 = +1 21 = .8.

Here we can apply the Bogomol'nyi trick and the equations of motion reduce to rst order equations.152): E = m2 =e2. 2 2 : (3. Unfortunately we can still not solve the equations analytically even in this 5 .151) + e 2 F12 . e2 )(j j2 .152) e2 where we used the ux quantization as well as the fact that Dj ! 0 for r ! 1.116 GAUGE THEORIES Figure 3. Without actually solving the equation of motions we can convince ourselves that the vortices are heavy. We try to repeat the Bogomol'nyi trick used for domain walls and rewrite the static energy density as a convenient sum of squares plus total derivatives. localized to r < max(1=mH 1=mV ).e. 2)2 : 0 0 Thus the energy is a sum over squares plus total derivatives. mH = mV it is possible 5 to have equality in (3. 2) 2 0 2 (3. so that 2 0 Z Z E = dxdy E e 2 F12 dxdy = 2 0 Z dxdy@i( Dj ) = 0 : (3. i ij @i( +Dj ) + ( . For bound for the integrated energy e2 2 we get a lower 2 = mV (3.153) Since mV is a typical mass excitation of the system we see that the vortex energy is non-perturbative in the coupling constant e2 . V The case = e2 =2 is special.8: The magnitude of the magnetic eld and the scalar eld as a function of the distance from the vortex center We see that the vortex viewed as a two dimensional object indeed is a "lump" of energy.150) 0 This can be written as (after somewhat tedious calculations) i 1h E = j(D1 + iD2 ) j2 + 2 F12 + e(j j2 . Note that if = e2 =2. i. The energy density is (in two dimensions) 1 2 E = jD1 j2 + jD2 j2 + 2 F12 + j j2 .

j j = 0 in the core of the tube and j j ! 0 outside the tube. and it is the intrinsic tension of the tube. We refer to example 3 for a discussion of the SO(3) theory. which can in a certain sense be given the interpretation as a monopole. and it is even possible to have a condensate of such vortices. The value = e2 =2 is special in other respects. the monopoles will also be very heavy and this has important consequences for cosmological considerations.3.3 Monopoles In the two former examples the solitons were 1.and 2-dimensional objects. i. After the transition vortex lines (\cosmic strings") can occur. They appear in a natural way in the so-called Grand Uni ed Theories (GUT). .7. Such strings might be very important in a cosmological context where they are denoted "cosmic strings". which is also the ground state of the present universe. They are like defects. we have: Energy 1 =2 2 (3. This model is itself not very important. In type II superconductors it is indeed possible to have stable vortices of magnetic ux tubes. The gravitational e ects are large and these strings are expected to act as seeds for galaxies by attracting gas particles. This means that we must expect monopoles to be present in a cosmological context.155) (6 km cosmic string has mass equal to the mass of the earth). mH = mV . It turns out that they are very heavy. but the construction of monopoles will carry over to generic GUT models in the same way as the vortex construction for U (1) led to cosmic strings.e.154) 0 2 0 : unit length in z-direction This quantity is called the string tension and is conventionally denoted 2 1 0 ( 0 = \the Regge slope"). and a scalar eld a which transforms case. and we call it a string. since inside they have the old vacuum ( = 0).7 SOLITONS 117 Viewed as a three-dimensional system the vortex discussed above is an in nitely long tube in the z-direction characterized by having a magnetic ux Hz inside the tube and Hz 0 outside. String tension = 2 1 0 1022g/cm (3. while j j has the dual behaviour. 3. The idea is that in the early universe one had a di erent high temperature phase where the minimum for the Higgs potential was like in g. We end our discussions of solitons with an example where the soliton is a genuine three dimensional extended object. a corresponding eld tensor Ga . We shall consider an SO(3) model (the Georgi-Glashow model). In fact the "monopole problem" was one of the main motivations for the suggesting the "in ationary scenario" in cosmology. At lower temperatures there is a phase transition (Tcrit 1029degree Kelvin ! 1016GeV) where the new ground state becomes 0 6= 0. Let us only recall that we have a gauge eld V a . The two dimensional gauge-Higgs model described here plays an important role in non-relativistic superconductors. 3. As was the case for the cosmic strings. The diameter of the tube is of order 1=mH 1=mV and in the simplest case = e2=2. For mH very large the tube becomes almost a line. In fact the e ective Landau-Ginzburg theory for such superconductors are precisely the static gauge-Higgs model we have considered here. as already mentioned.3a (symmetry restoration): j j = 0. which have been proposed as generalizations of the standard model. The special value = e2 =2 corresponds to the border between type II and type I superconductors.

160) = g1 3 "abc a@ b@ c + @ A . as for the vortex. Let us now assume that Di = 0 for r su ciently large.118 GAUGE THEORIES as an "isospin" vector. . where j (x)j = 0. We have for a given (x) in the region r > R0: D = 0 ) V a = g1 2 "abc b@ c + 1 a A (3. In such a region the SO(3) theory is locally indistinguishable from conventional electrodynamics. it is the non-trivial vacuum structure of the theory which allows us to impose a nontrivial boundary condition at in nity.162) We recognize Maxwells equations and have reached the important conclusion that in a region of space where j (x)j = 0 and D = 0 the only non-zero component of the gauge eld strength Ga is F = Ga a = 0. This is a remarkable situation. which has a nite energy above classical vacuum. Since the static energy of any con guration is: ~ (xi ) = (sin cos n sin sin E= Z d3x " 1 (Ga )2 + 1 (D a)2 + 4 ij 2 i 8 2.158) a necessary condition for a nite energy is that Di ! 0 for r ! 1. We shall now show that there exists a solution to the classical equations of motion. and that the spontaneously broken vacuum was characterized by a Higgs eld. This condition is strictly speaking not true for the monopole we have in mind.156) has to be nite. 2 2 a 0 # (3. which is regular everywhere and which allows the interpretation as a magnetic monopole.159) 0 0 where A is arbitrary. Again. In fact we impose what can be considered as the three dimensional generalization of the simplest vortex boundary condition in two dimensions: a ! na(xi ) 0 r!1 (3. A calculation shows that the non-abelian eld tensor corresponding to this Via is given by: Ga = F a 0 F (3. say r > R0. x cos ) = ~ (3.161) 0 By the use of the general eld equations and Di = 0 it can be shown (and we encourage the reader to ll in the details) that: ~ @ F =0 @ F =0 (3.156) where na (xi) is a unit vector. Usually one has to put in electric charges and monopoles by hand as point like objects. but it will be satis ed up to exponentially small corrections as we shall see. @ A (3.157) r We now follow the philosophy outlined in the vortex case and require that the energy di erence between the vacuum and a static solution satisfying the boundary condition (3. This component satis es maxwells equations and can be viewed as the gauge eld associated with local gauge rotations about (x).

and this part is responsible for the non-trivial ux: From (3. In our case G = SO(3) and H = SO(2) and S 2 = SO(3)=SO(2). This last term has no total ux out a sphere of radius r > R0. as is seen from (3. as is clear from (3.161).159)-(3. where H is the subgroup which leaves 0 invariant. the boundary conditions in this case can be classi ed as homotopy classes of the mappings S 2 (boundary) ! S 2 (= G=H ) (3.161) we get R R g = g Bi dSi = dSi "ijk "abc a @j b @k c =(2 3 ). The generalization to a mapping where the sphere at in nity is mapped k times on the sphere representing the vacuum manifold is (r ') ! na (r ') 0 na (r ') = (sin cos k sin sin k cos ): Next we turn to the calculation of the nontrivial magnetic ux in the case where the Higgs eld satis es the non-trivial boundary condition (3. part of the gauge potential and part of the eld strength is determined by the Higgs eld.156). The situation is the same here.164) Bi = 1 "ijkFjk = g r3i + "ijk@j Ak 2 The rst term on the rhs is a monopole eld of charge G = 1=g and is due to the phase of the Higgs eld. The component Ai = Via a = 0 is arbitrary. 6 Recall the general discussion: V = G=H .165) = 2 Bi dSi = 4 S (r) r>R0 g The situation is the same as for the vortex: The requirement of nite energy solutions leads to topological distinct boundary conditions and a quantized magnetic ux of 4n =g. 0 But it can be shown that the winding number for the map x ! (x)=j (x)j from the sphere S 2 in space to the sphere S 2 in "isospin" space is precisely 4 times this integral.156) is the simplest non-trivial realization. In fact one gets the same result for all (x) belonging to the same homotopy class as (3. while the last term is the standard expression for an abelian magnetic eld. 7 Let us note for completeness the following more general result: From (3.3. and therefore (3. .163) where we have used that the vacuum manifold V in our case can be identi ed with the sphere6 .156). These mappings are classi ed by their winding number as for the vortex and the boundary condition (3.165) is valid for much more general (x) than the function given by (3.161) we get for the magnetic eld for r > R0: 1x (3. Let us rst compare the boundary conditions: In the same way as the boundary conditions in the case of a vortex could be classi ed as homotopy classes of mappings from S 1 (the boundary) to S 1 = G=H (the vacuum manifold).159). The total magnetic ux associated with the con guration is therefore7 Z (3.156) and (3. This is not the case here. This just re ects that the Higgs eld is neutral and does not interact with the electromagnetic potential corresponding to rotations about . where the mapping is 1-1.7 SOLITONS 119 In the case of an abelian vortex we saw that Di = 0 for r > R0 and a non-trivial winding of were su cient to ensure a non-trivial magnetic ux.156). In the vortex case the gauge potential was completely determined at the boundary by the boundary conditions of the Higgs eld. However.

This form of the energy is useful for the application of Bogomol'nyi's method.e. It is possible to rewrite the energy expression (3. we get a rst order equation Ga = "ijk Dk a .166) and (3. one cannot solve analytically the equations of motion with the imposed boundary conditions (3. As for the vortex.170) shows that the mass of the monopole is non-perturbative heavy compared to typical mass excitations of the system. where Di = 0. 2 2+ a 0 1 2 a a ijk Gij Dk # : (3. we see that (3. i. Without proof we note that the mass of the lowest energy monopole for general couplings and g can be written m Mmonopole = 4 g2 W F ( =g2) where F is a slowly varying function 1 F ( =g2) < 1:5 for =g2 ranging from 0 to in nity. one can solve numerically the equations by an ansatz compatible with the boundary conditions (3.170) E 0 d3x@i Bi = 0 dSiBi = 4n 0 = 4 nmW g g2 We have used that the charged W -particle has a mass mW = g 0.158) as follows8: E= Z d3 x " 1 Ga . and the ij rst order equations can be solved explicitly.167) we get: Z Z (3. We can further conclude that the monopoles are heavy. From (3. One exception is the Prasad-Sommerfeld-Bogomol'nyi limit.169) Si = 0Bi where Bi is the magnetic eld. In the limit where . since it diverges in the perturbative limit g ! 0.159) (with Ai = 0): a 8 = na 0h(r) j Via = "aij n f (r) r (3. If G denote the charge of the monopole we have gG = n which is essentially Dirac's quantization condition.120 GAUGE THEORIES where the n is associated with the winding number n of the Higgs eld at in nity. Recall that the same was the case for the vortex. and (3.167) (3. However. This limit is called the Prasad-Sommerfeld-Bogomol'nyi limit.156) and (3.171) =g2 ! 0 we can minimize the energy by requiring that the rst square shall vanish.168) (3.156). 4 ij ijk Dk a 2+ 8 2.166) The last term can be written as a total divergence 1 Ga D a = @ S i i 2 ijk ij k Si = 1 ijk Ga a : jk 2 If we are in the region r > R0 .

3.8 YANG-MILLS EQUATIONS

121

where h(r) f (r) ! 0 for r ! 0 and h(r) f (r) ! 1 for r ! 1. The behaviour for large r can be analysed and the result is that f and h approach 1 exponentially fast when r > max 1=mW 1=mH ]:

f (r) ' 1 ; const:e;mW r

h(r) ' 1 ; const:e;mH r :

(3.172)

As in the vortex case we have emphasized the topology associated with the Higgs eld as the source to the non-trivial classical solution: If we impose a winding number on the Higgs eld at in nity it cannot maintain j (x)j = 0 everywhere. If we move to smaller and smaller r it is, by topology reasons, impossible to change the winding number. At some point the phase has to change in nitely fast, unless j (x)j ! 0 and in this way makes the phase unde ned. Taking j (x)j ! 0 costs energy and makes the monopole heavy. On the other hand it is precisely the ability of the Higgs eld to go to zero which allows us to avoid the point-like singularity of the usual monopole and in the end to have a nite energy. As for the vortex, the gauge eld has been acting like a spectator, determined by Di = 0. However, the gauge eld associated with the monopole also has a well de ned topology, as was the case for the vortex, and one can show that the quantized ux precisely distinguish topological di erent abelian ber bundles associated a sphere outside the core of the monopole. Unfortunately we have no space to discuss this interesting aspect further.

**3.8 Solutions of Yang-Mills equations
**

In the last section we discussed topological non-trivial solutions to the classical eld equations of gauge-Higgs systems in one, two and three space dimensions. They were all a result of a non-trivial vacuum structure. The requirement of a nite energy above the classical vacuum led to a classi cation of the allowed boundary conditions in homotopy classes of the map x 2 B ! g(x) 0 2 V (3.173) In (3.173) B denotes the boundary, which in our cases were S 0 S 1 and S 2, while V denotes the space of vacuum con gurations. The equations of motion were non-linear equations, and as we have seen the pure non-abelian gauge theories, where no Higgs elds are present, are themselves non-linear theories. In the vortex and monopole cases we mentioned that in addition to the topological classi cation of the Higgs elds, we also had a topological classi cation of the gauge elds. One could from these considerations be tempted to look for soliton solutions in pure non-abelian gauge theories. It can be proven (see example below) that only in four dimensional space, i.e. four space dimensions and one minkowskian time direction, is it possible to have static nite energy solutions. The same theorem will (slightly modi ed) tell us that it is only possible to have static nite energy solutions in a purely scalar theory, i.e. a theory without gauge elds, when the dimension of space is d = 1. This is in agreement with our previous results: In d = 1 we found domain walls in a purely scalar theory, but the vortex and monopole examples in d = 2 and d = 3 used a combined gauge-Higgs system.

Example 4: Derrick's scaling argument

122

The action can be written as

GAUGE THEORIES

1 Z dd x (F a )2 ; (F a )2 ) S = dt L(t) L(t) = 2 0i ij The equations of motion tell us that S is stationary at a solution, but if we further assume that A correspond to a time-independent solution of nite energy, the Lagrange function L(t) must be independent of t and stationary: L A] = 0: Aa (xi )

Z

We can now make a rescaling:

Ai(x) ! Ai( x)

This implies

Aa (x) ! A0 ( x): 0

Z

( 2 )2 Z dd ( x)(F a ( x))2 ! 0i d Z Z 2 2 a a dd x(Fij (x))2 ! ( d) dd ( x)(Fij ( x))2

dd x(F0ai (x))2

The condition that L is stationary reads: Z 1 Z a 0 = dL A] = (4 ; d) 3;d 2 2 dd x(F0ai (x))2 ; 1 dd x(Fij (x))2 d 2 Z 0 = dL A] = 2 dd x(F0ai (x))2 d

Only for d = 4 and the "electric" eld F0ai = 0 is there any possibility to have a stationary solution. We note without proof that even in the non-stationary case four dimensions is still a necessity for a nite energy solution. We encourage the reader to repeat the proof for a pure scalar theory and show that the dimension of space d must be one in order to support stable, static non-trivial solutions of nite energy.

When the dimension of space is equal to four there are non-trivial static nite energy solutions to the classical equations of motion for non-abelian gauge theories. The energy in ve dimensions is Z 1 a E = d4x 2 (F0ai )2 + 1 (Fij )2 (3.174) 4 where i j = 1 2 3 4. According to example 4 the term F02i must be taken as zero in order to have a nite energy solution. It is further necessary to have

a Fij (x) ! 0 for

jxj ! 1:

(3.175) (3.176)

In order to simplify the discussion we will simply assume for the moment that

Fij (x) = 0 for jxj > R:

This will be not be strictly true for the solutions. If Fij (x) = 0 in some open, connected region of space the gauge potential Ai(x) must be a gauge transformation of Ai (x) = 0. This is obvious in the abelian case, but can be proven in the non-abelian case too by the

3.8 YANG-MILLS EQUATIONS

123

following argument: Fij had the interpretation of curvature. Fij = 0 therefore implies that the path ordered integral

U (x) P exp(i

Z

C :x0 !x

dxi Ai)

is independent of the chosen curve C from a xed point x0 to x. From the de nition of the path ordered integral we have by di erentiation along a chosen curve ! x( ):

**dU (x( )) ; iA (x( )) dxi ( ) U (x( )) = 0 i d d Since this is valid for any curve C we conclude that @iU ; iAiU = 0 or Ai(x) = ;i(@i U (x))U ;1 (x)
**

(3.177)

In order to simplify the discussion even further we will in the following assume that the non-abelian gauge group is the simplest one: SU (2), but the results will be valid for any compact non-abelian group. This non-abelian gauge theory is called the Yang-Mills theory after Yang and Mills, who rst suggested the theory in 1954, as a theory for isospin. The generators of the Lie algebra are the Pauli matrices a a = 1 2 3 and any group element can be written as follows:

U = ei

a a

= cos j j + i j aj sin j j = u4 + iua

a a

a

u2 + u2 = 1 a 4

(3.178)

From a topological point of view we can therefore consider SU (2) S 3.. Let us assume that the boundary conditions (3.176) are satis ed. For r > R the gauge potential Ai(x) is a pure gauge as in (3.177). Each choice of a sphere ( S 3 ) with radius r > R the corresponding U (x) will de ne a mapping:

U : x 2 S 3(r) ! U (x) 2 SU (2) S 3

(3.179)

As was the case for the maps S 1 ! S 1 and S 2 ! S 2 these maps can be divided into homotopy classes according to their winding number. By continuity this winding number must be independent of the radius r > R and as was the case for the vortex and monopole we see that the requirement of nite energy solutions leads to a classi cation of the allowed boundary conditions according to topology. We will now show, as for the vortex and monopole, that a non-trivial boundary condition implies a lower bound on the energy. The method will again be the one of Bogomol'nyi, where we rewrite the energy as positive terms plus boundary terms, which have a topological interpretation. If we introduce the dual eld tensor: a ~a 1 (3.180) Fij = 2 "ijklFkl where "ijkl is the obvious generalization of the " symbol already introduced in two- and a ~a three dimensions, we have (Fij )2 = (Fij )2 and we can write: 1 Z d4x(F a )2 = 1 Z d4x 1 (F a ; F a )2 + F a F a : ~ E = 4g 2 ij ~ij ij 4g2 2 ij ij (3.181)

9 The Bianchi identities can be proven be recalling that . since U (x) is just in the form which allows the identi cation of SU (2) with S 3 (see (3. A lower bound like (3. but tedious calculations.181). it is di cult to solve the Yang-Mills equations themselves.182) j l k l ~ We can now write the four-dimensional integral of F F as a three-dimensional "surface" integral over S 3(r) r R: Z Z 1 a ~a (3.184) is saturated if we can nd a so-called self dual con guration satisfying a ~a Fij (x) = Fij (x): (3. The bound (3.185). the details of which we will skip.1 (x) U (x) = x4 +xij x + j Note that when jxj ! 1 Ai reduces to a pure gauge. These singularities could make E = 1. One can check by explicit calculation that 1 a ~a Fij Fij = @i Ki Ki = 2"ijklAa @k Aa + 3 cabcAb Ac : (3. The implication of this result is that the energy of a solution with non-trivial winding number n must satisfy: 2 E 8 g2n (3. and the winding number of this gauge is obviously n = 1. but can also be viewed as a consequence of the Bianchi identities: ~ Di Fij ] = 0 (3.187) ~ These identities are satis ed for any eld con guration Ai in the same way as @ F = 0 9. . The Yang-Mills equations can readily be derived from the lagrangian L(A) = .(F a )2 =4: D F ]=0 (3. is the abelian case It is in fact possible to nd a explicit solution Aa (x). as our boundary conditions demand. and is in principle easier to solve. Any con guration Ai which satis es (3.iF ij = Di Dj ] and that the covariant derivatives satisfy the Jacobi identities Di Dj Dk ]] + Dk Di Dj ]] + Dj Dk Di ]] = 0. The last equality sign follows follows from trivial. The solution at r > R is just a a gauge potential Ai(x) corresponding to a pure gauge where the gauge function U (x) has a winding number n. Note also the arbitrary scale parameter . The classical Yang-Mills equations are scale invariant.186) will automatically satisfy the Yang-Mills equations (3.184) Again we see that the mass of the soliton is non-perturbative in the coupling constant.183) d4x 4 Fij Fij = 3 dSiKi = 8 2n S (r) there the n is just the winding number characterizing the boundary conditions. but it might not be possible to extend the solution all the way to r = 0 without meeting singularities.184) does not ensure the existence of a solution. and if Ai (x) is a solution the same will be the case for Ai ( x).186) This equation is only a rst order di erential equation.185) Although they might look innocent.178)).124 GAUGE THEORIES The last term is a topological term. but one can take advantage of the Bogomol'nyi rewriting of the energy E in (3. We just state the result in the i case n = 1: i i 2 (3. This follows from the fact that the energy is a local minimum.188) Ai = x2 x 2 (@iU (x))U .

1 dimensional space.9 INSTANTONS 3. From this point of view they behave in many ways like classical particles. When we step one dimension down with the same theory. q0 )2 dt and we just have ordinary classical mechanics of one variable.189) S = dt 2 ( dq(t) )2 . by using the static solitons we have already found. 1 dimensional space. We now illustrate this statement. 1 space dimensions.1. where the maps belonging to di erent classes could not be continuously deformed into each other. (. The domain walls were solitons in a eld theory in one space dimension and one time dimension.e. However. a gauge transformation were the gauge parameter ~(x) = 0 for jxj > R can never change this condition. solitons in d space dimensions play an important role in the quantum theory in d . i. and we can as well change notation and write q(t). (2): U (1) gauge-Higgs theory in one dimension.190) de nes a map x ! (x)= 0 which can be viewed as a map from S 1 (space) to S 1 (vacuum). Each from (3. If we identity points at 1. The action will just be Z 1 2 (3. 1 dimensional theory.1) = 2 n in order that we can identify 1. No physics dependent on these remote points in the Rd. could be classi ed in di erent homotopy classes.3. Vortices were solitons in d = 2. This means that we can identify all points at in nity in our d . . Any local gauge transformation. but of the vacuum itself: The classical vacuum con gurations fall into distinct homotopy classes in the d . (q2(t) . The quantum theory will just be ordinary quantum mechanics.1 into the vacuum manifold V . the space of vacuum con gurations. This means that the the eld is only a function of time: (t). The corresponding lower dimensional gauge-Higgs theory is formulated in one space and one time dimension. (1): Classical mechanics. Let us assume that physical. i. We have two classical vacua: q(t) = q0 .190) e where (1) .9 Instantons 125 The solitons described in the last section are "lumps" of energy. gauge invariant observables fall o su ciently fast at in nity.e. In all the soliton cases we have considered until now the non-trivial boundary conditions arouse precisely because the continuous mappings from S d. and compactify space to S d. From a mathematical point of view we have compacti ed space and topologically identi ed it with a (large) hyper-sphere S d. The vacua belonging to the di erent homotopy classes are (x) = 0 ei (x) A1(x) = 1 @1 (x) (3.1 . localized in space. space is identi ed with a (large) circle S 1. The corresponding lower dimensional theory is in zero space dimension and one time dimension. this topological classi cation is no longer one of boundary conditions.

If we compactify space to S 2 the general vacuum con guration can be written as ( denotes the three component Higgs eld and Via the vector eld): (x) = O(x) 0 O(x) 2 SO(3) . Let us use the solitons already found to illustrate the general statement. In the lower dimensional theory we can identify the compacti ed space with S 3 and a general vacuum con guration is just a pure gauge: Ai = .126 GAUGE THEORIES (3): SO(3) gauge-Higgs theory in two space dimensions. For the gauge theories we have considered above. Let us consider a vortex. where by local gauge invariance we mean gauge transformations. Note that the variable x in the lower dimensional (the zero-dimensional) theory has no interpretation as a space variable. 0 to = 0. They will not change the topology. but if we leave the vacuum con gurations we see that the static solitons in d +1 spacetime dimensions provide us with a continuous interpolation between classical vacua in the di erent topological sectors. In all the cases we have a degeneracy of the classical vacuum. In the case of the vortex we have to work a little to make the identi cation of an interpolating con guration clear. We found soliton solutions in the pure Yang-Mills theory when the dimension of space was four. but just acts a parameter which labels the continuous deformation from = .4 provides an interpolation between the vacuum .1(x) (Vi ) = i" a Vi(x) = . which are di erent from the identity transformation only in some compact region jxj < R of space. We can still make gauge transformations which can be continuously deformed to the identity. the classical vacuum structure is therefore as shown in g. This means that x ! U (x) U (x) 2 SU (2) de nes a map from S 3 (space) to S 3 (SU (2)) and the winding number of this map is invariant under local gauge transformations. (4): Yang-Mills theory in three space dimensions. 0 and the vacuum 0. On a large square outside the vortex Again the map . 3. In the case of a domain wall it is clear that the soliton (x) of g. and assume as usual that we have a vacuum for r > R.1 (x) there U (x) is an SU (2) matrix.9.i @i O(x)]O x ! (x) is from S 2 (space) to S 2 (vacuum manifold) and the winding number of this map is invariant under local gauge transformation. 3. The corresponding lower dimensional gauge-Higgs model is formulated in two space and one time dimension. The monopoles were de ned as static solutions in d = 3.i @i U (x)]U . which is larger than the trivial degeneracy present by local gauge invariance. It is impossible to make a continuous deformation of a classical vacuum con guration from one topological sector to another without interpolating through con gurations which are not vacuum con gurations: For a theory in d-dimensional spacetime the di erent topological vacuum sectors are separated by energy barriers.

Now a vortex solution (x y) will start out as a vacuum con guration ~(x) = (x 0) in the lower dimensional theory. In the semiclassical WKB approximation the tunneling transition amplitude between the two classical ground states is just (3. The same kind of arguments can be applied for the monopole (as we have attempted to illustrate in g.11. q0 )2 Further the energy of our domain wall above the vacuum is precisely Z q0 q 2V (q)dq Ewall = . and nally end up at a new vacuum con guration ~y.3. or whether quantum tunneling will result in a vacuum which is invariant under the socalled large gauge transformations which corresponds to a winding number di erent from zero.192) since we derived the equation of motion for the domain wall by the Bogomol'nyi trick: Z 1 Eclassical = dx 2 (@x )2 + V ( ) Z q q = dx 1 (@x .) and for the Yang-Mills soliton.3a.0 q d 2V ( ) for Ewall . interpolate through con gurations ~y (x) = (x y) which are di erent from vacuum con gurations of the lower dimensional theory since (x y) enters the vortex core where j (x y)j < 0. In case (1) above we know the answer: The potential is the double well potential shown in g. 3.q0 2V (q)dq where the potential V (q) is given by (3. The horizontal axis is the complete con guration space.10.max (x) = (x ymax ) which has a winding number. 3. .q0 (3.191) R q0 p T e.9 INSTANTONS 127 Figure 3.9: The periodic vacuum structure of some gauge theories. 2V ( ))2 + @x 2V ( ) 2 = Z 0 .189): 2 V (q) = (q2 . and is in nite dimensional core we can therefore gauge the eld to the orientation shown in g. since the total winding number of (x y) must be one. 3. In the quantum theory it is now important to know whether this separation will persist. Quantum tunneling will ensure that the ground state wave-function is symmetric.

Figure 3.11: The Higgs eld of a monopole as an interpolating eld between a trivial two-dimensional Higgs vacuum and a non-trivial two-dimensional Higgs vacuum. by identifying the points at the circumference.128 GAUGE THEORIES Figure 3. Note that the disks for y = 0 and y = ymax can both be identi ed with S 2. .10: The Higgs eld of a vortex as an interpolating eld between a trivial onedimensional Higgs vacuum (y = 0) and a non-trivial one-dimensional Higgs vacuum (y = ymax). as illustrated.

This theory has two classical vacua: (x) = 0.3. But is there a nite energy barrier between the two vacua? The answer is no! An interpolating con guration would be a soliton in the 2 + 1 dimensional scalar theory. 3. much like what happens in a crystal where the classical ground state is also in nitely degenerate.1 . For a eld theory in d dimensional minkowskian space-time the static part of the corresponding d + 1 dimensional theory can be identi ed with the euclidean version of the d dimensional theory (x0 ! ixd ) and instantons become solutions to the d-dimensional euclidean theory.193) Similar results are true in the higher dimensional cases as well: The quantum transition amplitude between topological distinct classical vacuum states in a theory in d dimensional space-time is exponentially suppressed by the energy of the static soliton in the corresponding d + 1 dimensional theory. It even has a soliton connecting the two vacua. i. on a large hyper-sphere S d.9 INSTANTONS We can summarize this by the formula 129 T (vacuum 1 ! vacuum 2) exp(.193) as T (vacuum 1 ! vacuum 2) exp(.e.SE (instanton)) (3. but from Derrick's theory we know that purely scalar theories can have solitons only in 1 + 1 dimensional theories. When the solitons appear in this context we call them instantons. It might be enlightening to compare the situation described above with the situation in the scalar theory in two-dimensional minkowskian space-time considered earlier. Unfortunately we have no space to discuss some of the interesting implications of this vacuum structure.194) The true ground ground state must be a coherent superposition states associated with the in nite number of classical vacuum states shown in g. .9. Since the euclidean action SE ( ) in d dimensions S can be identi ed with the static energy of the d + 1 dimensional theory we can nally write (3. which satisfy non-trivial vacuum boundary conditions at in nity in Rd . The energy barrier must be in nite! This is a genuine eld theoretical result and could not happen in a dynamical system with only a nite number of degrees of freedom.E (soliton)) (3.

130 GAUGE THEORIES .

Chapter 4 Applications The standard big bang model face (at least) four major problems when extrapolated backwards to time much earlier than one second: 1. In the standard model there is no explanation for these perturbations and one has to put them in by hand as part of the initial conditions. It is hard to understand how two regions more than 50 horizon distances apart came to be at the same temperature at the same time. although 131 4. At least on smaller scale the universe is not homogeneous. and it is very di cult to give an explanation of the observed isotropy of radiation. Thus. and at the time of the so-called \grand uni cation" (see later) must be one and equal crit with fty decimals precision. voids etc. One sees galaxies. The horizon problem 2. The atness problem 3. these innocent looking order of magnitude observations became truly embarrassing when extrapolated backwards in time. The monopole problem The horizon problem was discussed in sec. Now it turns out that mass density uctuations are gravitationally unstable . This means that the microwave photons which we observe have been emitted from causally separated regions if only their directions di er by a few degrees. where it was shown that at early times t 1 the causal distance was much smaller than the scale factor. The atness problem was also mentioned in sec. The density uctuation problem 4.1. To account for this structure one has to assume an initial spectrum of mass density uctuations = . 10 of the cosmology notes.1 Motivation for in ation 4. 10 of the cosmology notes and originated from the observation that 1( crit ). if one extrapolates to very early times one must assume that these perturbations. The problem of density uctuations is of a similar nature. In the standard model this large-scale homogeneity is simply assumed as an initial condition.1 In ationary Cosmology . clusters. However.a region with a mass excess will produce an attractive gravitational eld which will try to further increase the mass excess.

2 Tr V( )=. hermitian 5 5 matrix of elds (with independent 24 components) which transforms under the gauge group as 0 (x) = U . it is fair to include this problem on equal footing with the rst three problems.2) where the prefactor is chosen such that Tr V ( ) = .3) (4. is speci c to the so-called grand uni ed theories (GUT). much smaller than one would \naturally" expect. This model contains scalar elds. represents a traceless.1) where U (x) denotes an SU (5) matrix. In the so-called minimal SU (5) theory2 the full gauge symmetry is spontaneously broken to U (1) SU (2) SU (3) by a set of Higgs elds which transform according to the adjoint representation of SU (5). The symmetry breaking is accomplished by the elds acquiring a vacuum expectation value of the form 2 61 s 6 2 60 = 15 6 0 6 60 4 0 1 0 0 0 0 2 0 0 0 0 0 0 1 0 0 0 .1 (x) (x)U (x) (4. The SU (3) group describes the symmetries of the strong interactions. were incredibly small.4) (4. only mention one of the simplest examples of such a symmetry group.132 APPLICATIONS they had to be present. Let us rst start with a few words about grand uni ed models. The gauge group is U (1) SU (2) SU (3).2) we get: 2 2+ 3+ 4 = . the monopole problem. We will not enter into a discussion here. It should be mentioned here that the minimal SU (5) theory is no longer in agreement with experiment.5) A typical form of the Higgs potential will be as follows: 2 + a (Tr 2 )2 + c Tr 3 + b Tr 4 and for the choice of eld (4. namely SU (5). However. but since these to some extend also contain the seed to the solution of the four problems and since it was in fact the monopole problem which initiated the development of in ationary cosmology. At present all experimental high energy results are compatible with the so-called standard model. 1 2 (4. . gauge elds and fermion elds. (It should also be mentioned that there is a similar problem for domain walls and other topological objects which can be present in GUT's).3=2 0 0 0 .1 There are many reasons to believe that the above symmetry is only the shadow of a larger symmetry which truly unites the three forces and which is spontaneously broken to U (1) SU (2) SU (3). slightly modi ed SU (5) models are still viable candidates for a grand uni ed theory. The fourth problem.3=2 = 2 3 7 7 7 7 7 7 5 (4. That is. It predicts a too rapid decay of the proton. 1=30 q = a + 7b=30 For details we refer to the course in Elementary Particle Physics II. while U (1) SU (2) describes the symmetries of the weak interaction and electromagnetism.

These topological objects are all heavy. . and the symmetry would not be spontaneously broken. In order not to be in disagreement with experiments one has to choose the coupling constants of the models such that Mx is large Mmonopole Mx 1014GeV (4. This gives a monopole density: 1 nmonopole 3 1 (4. We can now discuss the monopole problem. While this is clearly a crude estimate we expect nevertheless that it is correct up to some orders of magnitude and this is con rmed by more detailed calculations.7) This scale is referred to as the grand uni ed scale GUT .114)) that dH (t) = 2t (4. . This explicit discussion of SU (5) is only included to illustrate that theory is indeed of the kind discussed in second part of these lecture notes. and gx denotes a typical gauge coupling constant.6) 2 gx where Mx is a typical mass of one of the massive vector particles of the grand uni ed theories after the symmetry breaking.5). and therefore to the presence of one monopole con guration. \Heavy" means that they typically will have masses (or mass densities in case of domain walls and strings) Mx (4. Higgs 2t. and we may estimate that each such cube with 3 volume Higgs corresponds to a topological knot in the Higgs eld expectation value. vortices (cosmic strings).4. the Higgs eld on one face of the cube is essentially independent of the Higgs eld on the opposite face. and monopoles. Then for each cubic region of space with edge length Higgs.8) However.5). At this state there would be no alignment of the Higgs eld of the kind shown in (4. it took time for the correlations like (4.9) This means that the Higgs eld cannot be correlated at distances larger than 2t. At very early times t in the history of universe we have seen (see Cosmology (2. If c = 0 it has the discrete symmetry ! . Thermal uctuation would be able to kick the Higgs eld above the barrier created by the Higgs potential (4.2) to be established over space distances and the orientation of the Higgs eld cannot be correlated over distances longer than the horizon distance dH .1 INFLATIONARY COSMOLOGY 133 We see that the potential has the form needed for a spontaneous breaking of the gauge symmetry. In the very early universe the temperature T was above GUT . Let us call the actual correlation length Higgs. For a general grand uni ed theory (we do still not have observations which allow us to single out a speci c model. as discussed. if any at all) we will therefore have all the solitons mentioned: domain walls.10) 8t3 Higgs c where tc is the time at which the temperature drops to Tc.2). A phase transition took place: SU (5) ! SU (3) SU (2) U (1) at Tc 1014GeV (4. When the universe expanded it cooled and when T dropped to Tc GUT the Higgs eld could settle in the minimum of (4.

13) (4.17) which is totally incompatible with the observable limits of the present value of as given by Cosmology (2.15) eff 8t2 45 which allow us to calculate nmonopole at Tc ' 104GeV . In sec.113)) a2 (t) ' 2t 83 G D D rad (t) = 4 a (t) Combining (4.90).71) and (2. In the very early universe.18g=cm3 (4. the universe was dominated by radiation. where T0 refers to the present microwave temperature.13)-(4. In particular an energy density as high as given by (4. 2:75K 0.80).13) and (4. before any symmetry breaking. the radius of the universe at time t: a3 (t (t0 ) = Mmonopole nmonopole a3(tc) M 0) (4.134 APPLICATIONS The present monopole mass density can now be calculated. In addition we found in the early universe where a(t) is small (see Cosmology (2.000 years ! .84): a(t0)=a(tc) = Tc=T0 .e.16) would imply that the expansion rate of the universe has slowed down to its present value in only 30. 2.74) we see that monopole monopole = c 1011 (4. The ratio a(t0 )=a(tc) can be calculated by using Cosmology (2. Let t0 denote present times. a(t).8 of the cosmology notes we calculated the energy density for one photon (Cosmology (2.12) We expect Neff to be of 102 in models we consider. If we insert these numbers we nd monopole 5 10.14) 1 = 2 3 G N (kT )4 (4. with h = 1) MW = 15 (kT )4 2 The photon has two degrees of freedom (the two directions of polarizations) and the general formula for Nb massless bosonic degrees of freedom and Nf massless fermionic degrees of freedom is rad Neff = Neff 30 (kT )4 = NB + 7 Nf 8 2 (4.16) It may not seem so. i.14) leads to: s (4. If we compare it with the critical density as given by Cosmology (2.11) Let us rst calculate nmonopole from (4. but monopole is an outrageously high density.10).

19) The Einstein equation with a cosmological constant is R . 4.18) This result comes about as follows: In the at space-time we showed in the notes on gauge theories that the energy momentum tensor T of a single scalar eld is given by: 4. The value of V ( true) represents the energy density of the vacuum and is related to the cosmological constant by = 8 G V ( true) (4.21) true ) If we count all energies relative to V ( and = 0. In the case where is constant we get even in the co-moving frame of the RW metric: T ( ) = .20) 2 Since we know that the e ective cosmological constant in the present stage of the history of universe is zero with large precision and = true we conclude that 0' eff = .4.1. 1 g R + g = .2 The in ationary scenario T ( ) = . = true . Consider a potential V ( ) as shown in g.1. While we show the gure as if is one-dimensional it is a multicomponent eld. At zero temperature the global minimum of the potential is called the true vacuum. In grand uni ed theories this is not an unnatural situation.1: A typical Higgs potential in the grand uni ed theories.g V ( ) (4. a task which we leave to the reader. We imagine that the scalar eld today is undergoing small uctuations around this vacuum. 11 of the cosmology notes. we can therefore assume that V ( =0 .(@ )(@ ) .8 V( true ) true ) (4.8 G T ( ) (4. 1 f 2 @ @ + V ( )g This result may be generalized to multicomponent scalar elds.1 INFLATIONARY COSMOLOGY 135 Figure 4. a negative pressure in the early universe would lead to an exponentially fast expansion of the universe. As brie y mentioned in sec.

at least some places. The space described by a at RW -metric with an exponentially growing scale factor is called de Sitter space. In fact we have already argued that at very high temperature the Higgs eld should not be aligned in the vacuum.11: as long as the universe is close to the false vacuum it will expand exponentially: a(t) / eHinflation t where (4.g false (4. As discussed the critical temperature will be 1014GeV .23) false V ( false ) GUT 10 g=cm It is so large that a star would have to be compressed to the size of a proton to get this density. such that V ( ) has these properties.14) we see that it corresponds to a perfect uid with negative pressure P P = . From (4. We can now describe the chronology of the (original) un ationary model: we require the universe is hot (T > Tc). but almost similar results can be derived for k = 1.24) and by comparing with cosmology (2. only that some of the hot regions had false . it can be shown to be true under much more general conditions.136 APPLICATIONS We also require the potential to have a secondary local minimum. In the context of SU (5) grand uni ed theories one could imagine that the false vacuum either correspond to false 0 and unbroken SU (5) symmetry. (4. However.25) We are therefore precisely in the situation described in Cosmology.19) we have T ( false ) = . false = . If the Higgs eld were in thermal equilibrium within such a hot region then false . This means that one can imagine the \e ective" potential change with temperature as shown in Fig. Such a region will now expand exponentially. Due to temperature uctuations the \e ective" potential at a high temperature might not be identical to V ( ).22) At Tc the eld will therefore jump discontinuous in magnitude and we have the typical characteristic of a rst order phase transition.p (4. In eq. The energy density of the false vacuum is typical of the order of the fourth power of the characteristic mass scale of the theory which is 14 GUT 10 GeV : 4 73 3 (4. This is clear if it was homogeneous.25) and Cosmology 11. It is easy to arrange for the parameters of minimal SU (5) Higgs potential. or to a Higgs eld con guration which breaks the gauge symmetry to SU (4) U (1) rather than SU (3) SU (2) U (1). In addition any initial particle density is diluted to negligibility and s . due to the thermal uctuations. 1: For T > Tc the e ective potential will have a lower value near the false vacuum than near the true vacuum and at Tc Tc Tc Veff Veff ( false) (4. and the value of at this minimum will be called false .25) we assume strictly speaking that k = 0. isotropic and at (k = 0) as shown in (4.27) Hinflation = 83 G f 1010GeV 1034 s.26) (4. but we don't have to assume thermalization.1 This exponential expansion is the hallmark of the in ationary model.

It should also be mentioned that the in ationary model is much less sensitive to the initial conditions as discussed in the last section. Such a ne tuning is not natural. After the transition the energy density f will be released.3 Successes and problems of the in ationary model As already mentioned in sec.11 the horizon.4. Typically one nds a reheating temperature 1 (4. We will not enter into a discussion here. As this part of the universe expanded exponentially the temperature would drop. Supercooling is a well known phenomena in rst order transitions.1. However if we suppose that in ation continued for a time t the original hot region expanded by a factor Z eHinflation t (4. has been suggested. One has to remember that the bound (4. However. but it can be shown to be the case. In addition one has to show that not too many will appear at the transition to the true vacuum. and a closer investigation of the density uctuation problem lead to discovery of a major defect in the scenario outlined in the last section. Any original number of monopole in the false vacuum will be diluted. and chaotic in ation. extended in ation. After this transition the in ation will stop.10) is no longer valid. In a rst order transition one expects the transition to take place by bubble formation. since it was based on Higgs dH (tc) = 2tc. One would therefore expect that the hot phase where = false would supercool far below Tc before the transition. The temperature to which this gas reheat can be calculated for any particular GUT theory by using the conservation of energy. Also the monopole problem is solved. One of the four problems mentioned in the beginning.29) Treheat 3 Tc 4. the universe after the phase transition became extremely inhomogeneous.1 INFLATIONARY COSMOLOGY 137 any initial distortion of the metric is redshifted until it is no longer detectable during the expansion. was not solved by the in ationary model mentioned here. if the decay rate of the false vacuum was small. and in a considerable part of the physical volume of the universe the phase transition to true was never completed. C.and atness problems are solved by the in ationary models. 1 the transition will actually be a rst order transition. However the idea of in ation has remained a very simple and attractive way to solve the problems mentioned in the introduction. thermalize and produce a hot gas. A number of re ned models: the new in ation. and when T < Tc the system will undergo a phase transition from false to true. We see that the scenario mentioned in the table in Cosmology 11 emerges in a natural way in grand uni ed theories. Bubbles of the true vacuum are formed everywhere with certain probabilities. we still have to demand V ( true) = 0 with great precision to be in accordance with observations. exactly as one was postulating in the standard cosmological model.28) which can be tremendous. On the other hand. then the phase transition occurred very rapidly and in ation hardly took place. They cure some . A detailed analysis showed that if the rate of bubble formation was bigger than the speed of the universe expansion. the density uctuation problem. For completeness let us mention that with the potential shown in Fig. Only an unrealistic netuning of parameters in the GUT theory could avoid one of these situations.

However we have no time to cover these exciting topics here.138 APPLICATIONS or even all of the problems in the old in ationary model. and have to refer the interested reader to the literature. .

Show that then A is a vector. Show that an arbitrary tensor with two indices can be written as the sum of a symmetric and an antisymmetric tensor.A . V .61). Problem 2: Let A be a quantity consisting of four components. Generalize this result to arbitrary tensors. (1. A = . on quantum mechanics). The resulting theorem is called the \quotient{theorem". A0 A1 A2 A3. Do the same for spherical coordinates: x = r cos sin y = r sin sin z = r cos : Check the correctness of the resulting Laplace operator by comparing your result with some book (e.Chapter 5 Problems and constants Problem 1: Consider cylindrical coordinates: x = r cos y = r sin z=z Find the metric expressed in terms of these coordinates. Assume that A B = invariant for an arbitrary vector B . 139 . Problem 5: Assume that V = @V =@x +. Problem 4: Consider a two{dimensional surface of a sphere. Then show that V = g V satis es eq. Find the Laplace operator in these coordinates. Show that S A is identically zero.g. de ned by x2 + y 2 + z 2 = R 2 : What is the metric expressed in terms of x and y ? Find the metric in spherical coordinates. Problem 3: Let S = S be a symmetric tensor and A an antisymmetric tensor.

(1. nd the line element d 2 in a system where the x{y{plane rotates around the z-axis by using that the angle in the rotating system is ' = ' .136) reduce the number of independent components in the Riemann{tensor R to just one. Consider the surface of a sphere with metric ds2 = r2d 2 + r2 sin2 d'2 Show that the Gaussian curvature is given by a constant. Show that one obtains the centrifugal force as well as the Coriolis force.140 Problems Problem 6: Rotating coordinates. and compare this result to the time{dilatation in special relativity. 1 @ 2 b + 1 @b 25 K = ab 2 (@x1 )2 4b @x1 2 !3 3. K = 1=r2 : . Consider a surface described by the coordinates x = 1 2: Show that the symmetry properties (1.132) . R1212 is called the Gaussian curvature: g 2. ~ Show that this implies that watches in the rotating system are di erent from those in the \ xed" system. dx dx = 0 d2 d d First. Problem 7: Curvature in two dimensions: 1. K . and derivation of the usual \ ctitious" forces from d2x + . Investigate what happens to lengths. Show that for a metric of the form (a is a constant) g = a b(x10 x2) 0 K is given by ! 1 4. . !t (! = the angular velocity). Compute the forces by means of .

A circle on a curved surface can be de ned as the set of points P which have the same distance from the \center" Q. Consider a torus. (The geodesic is the shortest path between P and Q. Introduce coordinates and ' as shown in the gure below: a) Find the metric. measured along the geodesic from P to Q. We assume that for P su ciently close to Q there is only one geodesic). c) Compute the integrated curvature Z p d2x gK : 5.141 4. Introduce a \polar" coordinate system around Q. b) Compute the Gaussian curvature as a function of and '. de ned by: r measured x1 (p) = the distance from Q toPPand Q : along the geodesic through .

C 3 Here C is the perimeter of the \circle" with radius r. d) Show that unless the curvature in Q is singular then lim K = 3 r!0 2 rr. 0 = m0 =V0.e.where arbitrary constant . b) Show that the requirement that the curve (r(s) (s)) = (s 0 ) . c) Show that K diverges in Q unless (@ 2 a=@r2 )r=0 = 0. ! 1 0 g ' 0 r2 for r 1 imply that the metric can be written in the form ! 1 0 g = 0 a2 (r ) with a(r ) = r + 0(r2). Consider a particle at rest with mass m0 and volume V0. i. dx dx = 0 ds2 ds ds together with the requirement that the metric is \locally at". Introduce systems of inertia: . Let the particle move with constant velocity v along a straight line. Show that the energy{momentum tensor is 0 0 0 01 B 00 0 0 0 C T =B 0 0 0 0 C B C @ A 0 0 0 0 2.satis es the geodesic equation d2x + . The mass density is thus constant.142 Problems the angle in the point Qbetween the geodesic through P and Q x2 (p) = and a reference geodesic through Q and a xed point P0 : a) Show that g11 = 1. 0 is an Problem 8: The energy{momentum tensor for a particle in special relativity: 1.

vt) Show that the energy{momentum tensor in the system S is given by 0 2 1 2 0 0 v 0 0 B 2 C 2 2 T = B 00 v 0 0 v 0 0 C B @ A 0 0C 0 0 0 0 Give a physical interpretation of this result ( = 0 2).) 2 where k k = 0 and @ A = e k = 0 (why?).r .143 t0 = (t .. @ A B=r A @t can be written in covariant form ~ F =@ A . The energy density in = p11 v2 . Show that if the particle has an electric charge e.@ A A = ( A) : 4. 3. ~ ~ ~ ~ ~ ~ E = . . vx) x0 = (x . Find T . Show that the relation between elds and potentials. g = an arbitrary system of inertia is given by 1 ~ ~ T 00 = 2 (E 2 + B 2) Show that this expression is reproduced if T =F F . then the charge density transform like = 0 . Consider electromagnetism without gravity (i.1 F F 4 ). : 2. Show that the current of energy T 0i is given by the Poynting vector ~ ~ T 0i = (E B )i : 3. Problem 9: The energy{momentum tensor for an electromagnetic eld: 1. Consider a plane wave A = 1 e (eik x + complex conj.e.

Next. b) The quantity a D=D (D x =D ) can be interpreted as the four-acceleration of a particle moving along the geodesic x ( ) + x ( ). c) Consider a Schwarzschild black hole and an observer who is in a free. (1. v2. following a geodesic x0 = t = t( ) and x1 = r = r( ): At time = 0 she passes the point P0 = (t0 r0). 2MG=r) 2MG=r3: If you do not bother to do this rather tedious calculation. . 2MG=r0 @r0 0 for some v. radial fall towards the center. Explain why. satis es the di erential equation D D x = R (x( )) dx dx x D D d d where D=D denotes the covariant derivative along the curve x( ). Show that her freely falling coordinates (x00 x01 ) = (t0 r0) and the coordinates (t r) are related by @t (P ) = q @t0 0 1 . 2MG=r0 @r (P ) = vq1 .144 Problems Problem 10: Tidal forces. with 0 jvj < 1. Here = 1= 1 . the diligent student could show that 1 R01 010 (P0) = 1 . as seen by a xed observer at r = r0. 2MG=r R1010 (P0) 0 R1010 (P0) = (1 . 2MG=r0 @t (P ) = q v 0 0 @r 1 . p d) Show that regardless of the value of v. Hint: Use freely falling coordinates at the point x( ). as seen by an observer moving along the geodesic x ( ). just assume the validity of this equation.67). 2MG=r 0 @t0 0 q @r (P ) = 1 . see eq. The quantity v can be interpreted as the velocity of the freely falling observer. a) Show that the separation between two nearby geodesics x ( ) and x ( ) + x ( ).

because we use the letter E to denote the electric eld. so the functions E ! A and F ! B are the unknown quantities. (1 .190) and (1.1dr2 . Then insert your result in the energy-momentum tensor. F 01 = . r2 (d 2 + sin2 d 2): Here q is the charge. as she passes the point P0. and determine the integration constant from this. 2GM=r + Gq2=4 r2). assuming that she falls with the feets rst towards the black hole. Please notice that the unknown functions A B cannot be determined as in the Schwarzschild case. At this stage. relative to her head in the freely falling coordinate system. and solve the Einstein equation. and discuss what happens. For r ! 1. We can therefore take g and R from the main text. you obtain the solution for E (r) as a function of r and the (so far) unknown functions A(r) and B (r). as well as an integration constant.216). 2GM=r + Gq2 =4 r2)dt2 . and that the trace of the energy-momentum tensor vanishes (T = 0). (1. Problem 12: The energy of a test particle in the gravitational eld g is de ned m0 = the rest mass: E . you should have the equations of motion (away from the point r = 0) R = .145 e) The observer's height is called h. Calculate the acceleration of the observers feet f) Estimate the size of this acceleration for an observer.8 GT and F = 0: By solving the second of these.m0 g0 dx d a) Show that E is a constant of motion if @g = @ g = 0 : 0 @t as b) Consider the Schwarzschild metric and a test particle which moves in the plane = 2 . Show that the Schwarzschild radius becomes modi ed: Consider q2 to be smaller/larger than M 2 . . Express the energy in terms of dr=dt and d'=dt. require at space. Show that the electromagnetic eld strenght has the following non-vanishing components: Problem 11: The Reissner-Nordstrom solution: Consider a point mass which is charged. These functions should be renamed. The result is d 2 = (1 .F 10 = E (r) F = 0 for all other with x1 = r and E (r)=the electric eld. since the energy-momentum tensor is now given by the general relativistic generalization of the result found in problem 9. Argue that the solution must be spherically symmetric. Show that T = 0 for 6= . so that the metric tensor and curvature tensor are of the same form as in eqs. Then compute the components of T . as she crosses the Schwarzschild radius of a black hole with mass=the mass of the sun.

c) How much does V deviate from the Euclidean result V0 = 43 R3 (you are allowed to expand in b=a !). and compare with the analogous expression in special relativity. Problem 14: Consider the Robertson{Walker metric. At a given time t the spatial dr2 + r2(d 2 + sin2 d'2) 1 . This is described by the Saha{equation Nr+1 0 Ne = 2 gr+1 0 (2 mekT )3=2 e. Why? Problem 15: For a non{relativistic gas of atoms of type A at the temperature T it is possible to have equilibrium with the gas of free electrons: Ar ! Ar+1 + e.146 Problems Express the energy in terms of v. V0)=V0 for a b (b 1010 106 light years light year the distance to the nearest galaxies). e) What is the volume of the closed universe (k = +1). : =kT Here Ar is the r times ionized atom. Show that a photon (light particle) can move around in a circular orbit r = 3GM .1 k = 0. Hint: Only half the universe is spanned by 0 r 1. Nr 0 gr 0 h3 . kr2 Consider the sphere with surface r = b. c) What is the velocity v measured by a local observer at rest ? Problem 13: Consider the de ection of light around a black hole described by the Schwarzschild metric. d) How large is the deviation (in per cent) (V . b) Compute the volume V in the three cases k = 1 k = . Is this possible for other values of r. ds2 = a2 part is given by " # a) Compute the physical radius R of the sphere.

147 where

Nr 0 = # per m3 of the r times ionized atom with the bound electrons in the ground state: Ne = # of electrons per m3: gr 0 = statistical weight for the r times ionized atom in the ground state. = the ionization energy for the r times ionized atom: me = the electron mass (0:91 10;30 kg): k = Boltzmann's constant (1:38 10;23J=0K ): h = Planck's constant (6:6 10;34 Js)

For hydrogen we have

gH + = 1 gH = 2 mH = 1:67 10;27 kg:

= 13:6eV (1eV = 1:6 10;19J ) :

**a) Consider the ionization of hydrogen around the time of recombination. At this time
**

the universe is matter dominated. Show that the density of protons plus hydrogen atoms is given by

N

NH + + NH = BT 3

where B is a constant. Assume that the density of baryons today is baryon crit. 0 Assume for simplicity that all the baryons are hydrogen (in reality it is only 70 %).

**b) Find the approximate value of B by assuming that
**

following we assume that B = 0:3 (0 K );3m;3 .

crit

10;26kg m;3 . In the

NH c) Show that the fraction NH+ +NH of neutral hydrogen varies like

T 46000K : 3 % T 41000K : 48 % T 36000K : 94 %

Problem 16: Assume that we modify the Einstein eqs. by introduction of the cosmological constant ,

R ; 1 g R ; g = ;8 G T : 2 a) How does this manifest itself in the \Newton{limit" (v 1 g ' + h h time independent). b) Consider a Robinson{Walker universe. How are the Einstein eqs. (2.49) and (2.50) modi ed for 6= 0 ?

148

Problems

Hint: The term g just gives rise to a linear shift in pressure and density. c) Show that it is possible to obtain a static solution a(t) = constant by choosing a

suitable value for . Which one ? (Assume that p). This static universe was proposed by Einstein (1917), who assumed that k = +1. d) Express the volume V = 2 2a3 of this universe in terms of the density .

**Problem 17: a) Find the connection between the red shift z and the physical distance L0 to second
**

order in L0 . Express the result by means of the Hubble constant a(t ) _ H0 = a(t0) 0 and the deceleration parameter (t (t0 q0 ; a(_0()a))2 ) : a t0

1 q0 < 2 in an open universe q0 = 1 in a at universe 2 q0 > 1 in a closed universe 2 when we assume that the universe is matter dominated. Problem 18 Consider the canonical energy{momentum tensor (in Minkowski space) T = ;(@ r ) r + L : a) Show that if we add a term @f to T , where f vanishes at spatial in nity, and satis es f = ;f then the value of the total 4{momentum Z P = d3x T 0 is unchanged. Show that T + @ f satis es the continuity equation (3.37). b) Consider an electromagnetic eld with L = ;1F F 4 Show that the canonical energy{momentum tensor is neither symmetric nor gauge invariant.

v

b) Show that the deceleration parameter satis es

149

c) Show that T can be made symmetric and gauge invariant by adding a term of the

type discussed in a). Hint: Use the eq. of motion @a F a = 0)). Problem 19: Consider a variation g of the metric. a) Show that the corresponding variation of the Christo el symbols ; transform like a tensor. b) Show that the variation of the Ricci tensor is given by R =( ; ) ;( ; ) :

Hint: Use freely falling coordinates. 1 c) Show that pg = ; 2 pg g g . d) Show that the variation of the Einstein{Hilbert Lagrangian L = ; 1 p g R g = ; 1 pg R

EH

is given by

16 G

16 G

h i 1 ; LEH = 16 1G pg R ; 2 g R g ; 161 G @ (pg g ; ) ; @ (pg g ; )

SEH

e) Show that the requirement that the action

Z

d4x LEH

is stationary leads to Einstein's equations in the vacuum. Problem 20: Consider the action for the gravitational eld and matter, S = SEH + SM Z with SM = d4x LM : stationary under variations of g , reproduces the Einstein equations if we de ne the energy{momentum tensor for matter by 1 Z d4xpg T g SM = ; 2 : b) Use this result to nd T for a scalar eld with action Z S = ; 1 d4xpg g @ @ 2 and for an electromagnetic eld with action 1 Z S = ; 4 d4xpg F F g g :

a) Show that the variational principle, i.e. the requirement that the full action S is

4 ( 2 )2 i Problem 25: The Lagrangian for three real elds is given by 2 2 i 2 < 0 i 3 Show p= 21 22 andand masslessand > 0.148) ]. Find F .ik0t + a+(k)ei~ ~ +ik0 t x q k0 = ~ 2 + m2 : k Compute the energy density.. Show L = . 1 2 2 1 .e. i(@ U )U .128) explicitly. Show that these can be transformed to eqs.87) . 1 (@ i)2 .137). for a vortex (3. Write these equations in . and the energy. that it describes a massive eld with mass .135) by solving (3. containing j j and the gradient of .134). where ! (x) is a set of functions.150 Problems c) Show that T found in a) is symmetric. (3.147). Problem 27: Consider the eqs.(3. two Goldstone Problem 26: For the U(1){case show that exp(i dxA(x)) discussed in (3.1 . Problem 22: Consider a free complex eld (t ~ ) (i.(k)ei~ ~ . Find a gauge transformation such that A0 = UA U . Z Hint: Use that A j is the interaction energy between a current j and the electromagnetic eld (show this by use of the eq. a) Consider a gauge eld A (x) = ! (x) . of motion is Z h i kx kx (t ~ ) = d3k a. of motion for the A { eld). Problem 24: Consider eq. Problem 21: Consider the SU(2) Yang-Mills theory. b) Assume that F = 0. V ( ) = 0).128) and show that it leads to the solution (3. Show that the eld = U L ! can be transformed to 0= 0 ! 0 L : that the energy is given by (3.99) can be considered as exp(iE ). Problem 23: Consider SU(2). Show that the x general solution of the eq. = 0 1 2 3 and is a constant 2 2 matrix which is hermitean and traceless. (3. Use this observation as a devise to nd a symmetric T even in at space (where g is replaced by ).1 = 0.bosons. (3. where E is an integrated energy (which one ?).

Problem 28: Show that j j = 0 is not a solution of the equations found in problem 27. then r2 log j j r2 ln r a two{dimensional {function for r ! 0 (why?). Find A1 and A2 expressed in terms of j j and . ij =0 Problem 30: Find a \gauge transformation" which \removes" the angle . Show that if F12 is nite. Show that after such a transformation e the energy density is in nite! (Thus. (3. and for i = j and 12 = . How can one understand this apparent singularity in the equation r2 log j j = : : : ? . Find the second order equation satis ed by j j (of the type r2 log j j = : : :). it was not an allowed gauge transformation . Problem 31: Consider the case = e2=2 and use eq. where c is a constant ( nd it!). What happens to the zero's of ? If j j r for r ! 0.151). Generalize this to the case where the phase of is given by eiN N = 2 3 : : :. 21 = +1. Minimize the energy by putting the squares equal to zero. Problem 29: Show that ij @i @j = c (x) (y).151 polar coordinates (r ). then A ! A0 = A + 1 @ (x). then j j must necessarily have a zero.why not?). Show that F12 can be expressed in terms of j j. Use that if ! 0 = ei (x) .

152 Problems .

13 cm 1 eV = 1:6021892 10.19 Coulomb hc = 197:32858(51) MeV fm = e2=hc = 1=137:03604(11) 1 fm = 10.10 esu = 1:6021892(46) 10.1 .22 MeV s e = 4:803242(14) 10.29 cm g.2 G=c2 = 7:425 10.8 dyn cm2 g.y. Planck's constant Electron charge Conversion constant (useful in converting to \natural units" h = c = 1) Fine structure constant 153 c = 2.2615 l.Some constants: General Speed of light (This value is \exact" in the sense that the meter is de ned .27 erg s h = h=2 = 1:054 5887(57) 10.y.12 erg General Relativity and Astronomy Light year Parsec 1 l. = 9:4605 1017 cm 1 pc = 3:0856(1) 1018 cm = 3.since 1983 .99 792 458 1010 cm s.as the length of the path traveled by light in vacuum during a time interval of 1=299 792 458 s). Solar mass M = 1:989(2) 1033 g Solar radius R = 6:9598(7) 105 km Earth mass M = 5:977(4) 1027 g Earth equatorial radius R = 6:37817(4) 103 km Mean earth{sun distance 1:495985(5) 1013 cm Gravitational constant G = 6:6732(31) 10.27 erg s = 6:582 173(17) 10.1 h = 6:626 176(36) 10.

154 Problems .

155 .

Sign up to vote on this title

UsefulNot useful- Througout the History of Mathematics
- The Secret of Light - Russell, Walter
- Rare Earth Magnetism
- Walter Russell - The Message of the Divine Iliad (Vol. 2)
- Magnetic Current by Ed Leedskalnin
- Basic Properties of Matter by Dewey B Larson
- Nothing but Motion by Dewey B Larson
- Compressed Marko Rodin Tesla Tech Article June 14th 2010
- Walter Russell - The Message of the Divine Iliad (Vol. 1)
- Tensor Gauge Condition and Tensor Field Decomposition
- 03-23
- Simplified Calculation of Component Number in the Curvature Tensor
- Nazrul Islam-Tensors and their applications-to New Age International Pvt Ltd Publishers (2006).pdf
- General Theory of Relativity by P. a. M. Dirac
- Lecture 0
- GRCalcs
- final_2012
- scalar2010_1022
- On The T Curvature Tensor of Generalized Sasakian Space Form
- A type of Semisymmetric metric conection ina LP-Ssasakian manifolds
- Quantum Theory of Gravitation
- Chapter 8, Astrophysics
- General Relativity Thooft
- Einstein Field Equations
- HilbertActionFromEinsteinsEquations-3696865
- PHY_F315_1494
- Introduction to General Relativity -'t Hooft
- 0812.1513v1
- Cosmology