Vous êtes sur la page 1sur 13

1

1.1

Theoretical foundations of general relativity


The geometrical nature of physical laws

The laws of physics are very often presented as relations between quantities (location in space and time, momenta of particles etc.) that can be expressed in a certain coordinate system or reference frame. For instance, one can think of the vectorial quantities usual to Newtonian Physics (momenta, electric force etc.) as being merely triplet of numbers representing the vectors components on the axes of some spatial coordinate system. It is important to realize that it is not the case. Physics is not the only human activity which establishes relations between quantities which are themselves expressed as collections of numbers. However, what makes physical laws exceptional, as compared to e.g. economical or jurisprudential laws, is that they can be expressed in a geometric form which is independent of any coordinate system. In other words, physical laws are geometric, coordinate-free relationships between geometric, coordinate-free quantities. It is a fundamental principle of physics. Geometry is the constraint on the nature of physical laws. The reason is the following: by principle, physical laws are geometric, coordinate free relationship between geometric, coordinate free quantities. It means that if a law is not such a geometric relation, then by denition this law is not a physical law. Let us give some examples. Newtons law, dp = F, dt is a physical law. Also, is Newtons gravitational law 2 = 4G, (2) (1)

a law of physics ( is the gravitational potential, is the mass density, G is the Newtons gravitational constant and 4 is a geometrical constant), as well as a similar law that relates the electrostatic potential to the charge density. We already know that the gravitation potential (l) at a distance l from a particle of mass M is (l) = GM/l : this is also a physical law. The relation (l) = GM/l and relations (1) and (2) are physical laws in the sense that none of the quantities involved in them need a reference frame for their denition. Of course, it is sometimes useful to express physical 1

quantities in a particular coordinate system. For instance, one can express p, F and 2 in a Cartesian coordinate system, which for the latter reads : 2 = 2 2 2 + + 2 x2 y 2 z (3)

1.2

Newtons gravitational eld law

The form of Newtons gravitational law (2), a eld equation, is new to you. But it does not matter since we will need only a vague idea of what it is. Because you have heard of partial dierential equations, you recognize that (2) is a partial dierential equation. Law (2) relates the gravitational potential eld , a scalar eld, to the distribution of mass density , also a scalar eld (mass per unit volume at each point in space). Dimensionally, (2) reads /L2 = GM/L3 , hence, = GM/L. A kind of equation like (2) determines the spatial distribution of the gravitational potential created by a mass that is distributed in a certain volume of space. For a particle of mass M we already know that (l) = GM/l. From the scalar eld , one obtains the vector eld g = (coordinate free relation), g is the gravity eld. In a Cartesian coordinate system = ( /x)ex + ( /y )ey + ( /z )ez (ex , ey , ez are the unit vectors forming a basis of the Cartesian coordinate system, i.e. ex .ex = ey .ey = ez ez = 1, ex ey = ex .ez = ey .ez = 0. The force exerted by the gravity eld g on a particle of mass MP is F = MP g (coordinate-free relation).

1.3

Geometry and Metric

A fundamental notion of Euclidian geometry is the elementary interval of length ds between two points in ordinary space, P and Q, whose separation vector is dr. In a specic (Cartesian) coordinate system, dr = dxex + dy ey + dz ez , and this dierential distance ds is dened as : (ds)2 dr.dr = dx2 + dy 2 + dz 2 =

i,j

ij dxi dxj

(4)

Here ij is the Kroenecker delta, (unity if i = j ; zero otherwise), i or j are indices = 1, 2, 3 and (x1 , x2 , x3 ) = (x, y, z ). So, in a cartesian 3D Eu2 2 clidian space the dierential distance is given by ds2 = dx2 1 + dx2 + dx3 . In cartesian 4D Euclidian space the dierential distance is given by ds2 = 2

2 2 2 dx2 0 + dx1 + dx2 + dx3 . We just added a dimension that we labeled by the indices 0. This looks very much like 3D and 4D versions of Pythagoras formula. Pythagoras formula gives the length of a line in 2 spatial dimensions. In special relativity we encountered an other fundamental notion, the one of the interval (ds)2 between two events, P and Q residing in space-time. In a specic but arbitrary inertial reference frame it is given by:

(ds)2 c2 dt2 + dx2 + dy 2 + dz 2 = c2 dt2 +

i,j

ij dxi dxj

(5)

This is a generalization of (4) to 4 dimensions, involving both space and time. The dierence between (5) and the cartesian 4D Euclidian dierential distance is in the sign in front of the time variable. It is not too hard to verify that (5) can be written in a manner similar to (4), as (ds)2 =

i,j

gij dxi dxj

(6)

with gij (c2 if i = j = 0; unity if i = j = (1, 2, 3); zero otherwise) and (x0 , x1 , x2 , x3 ) = (t, x, y, z ). The quantity g , a number with two indices, is called a metric. And if the indices involved in the metric run from 0 to 3, this means that the metric is a 4 dimensional metric. The metric describes the geometry in arbitrary dimension. It describes the geometry because it denes the length or the interval in arbitrary dimension. The metric ij is the Euclidian metric of 3D space. The metric gij is the Minkovski metric of 4D space-time. Just as the geometry of Euclidian space implies the Euclidian metric, the geometry of special relativity implies a non-Euclidian metric : the Minkowski metric. The Minkovski metric is non-Euclidian because of the negative sign in front of the time variable: there, it involves the number 1 instead of the number +1. The geometry of space-time described by special relativity theory is non-Euclidian. The square of the length between two points P and Q in space-time (events) is an interval (s)2 and special relativity can be stated in terms of the invariance of space-time interval between two events as seen from any inertial reference frame. Examples of Euclidian and non-Euclidian metrics: We already know that dr = dxex + dy ey + dz ez in Cartesian coordinates. In spherical polar coordinates, dr = drer + rde + r sin de with er , e , e are being the unit vectors of the spherical coordinate system. We also know that ds2 = dr.dr. 3

So let us express ds2 in Cartesian and in spherical coordinates. The metrics ds2 = dx2 + dy 2 + dz 2 and ds2 = dr2 + r2 d2 + r2 sin2 d2 are both Euclidian metrics of the (at) 3D space. They are equivalent since we can go from Cartesian to spherical polar coordinates by : x = r cos sin , y = r sin sin , z = r cos . The metric of a at 2D surface or a plane, ds2 = dx2 + dy 2 is Euclidian. The metric of a 2D spherical surface with radius (of curvature) r = R, ds2 = R2 d2 + R2 sin2 d2 is non-Euclidian. A plane is a good approximation to a spherical surface only locally. Take a point on the surface of a sphere and think of the plane which is tangent to the sphere at that point.

1.4

What is the problem with gravity ? historical perspectives

Let us come back to Newtons gravitational eld law (2). In the decade following Einsteins formulation of special relativity (1905-1915) many theoretical physicists were trying to adapt Newtons gravitational eld law (2) to the framework of special relativity. The motivation for adapting the eld equation (2) to the framework of special relativity is the following : Newtons gravitational law (2) seems incompatible with special relativity since it implies instantaneous action at distance. Why (2) implies action at distance ? Because in law (2), time is ignorable. But wait a minute. By reasoning this way, arent we thinking of the two geometrical quantities on the left and right hand sides of (2) as being Euclidian geometrical quantities? And we know that they cant be so in special relativity because the geometrical framework of special relativity is not the 3D Euclidian geometry but is instead the 4D Minkosvskian geometry. Notice that in a specic, but arbitrary inertial reference frame of the 4-dimensional Minkovski space-time, the left hand-side of (2) takes the form : 2 = (1/c2 ) 2 /t2 + 2 /x2 + 2 /y 2 + 2 /z 2 [compare this with (3) and compare (4) with (5)]. Equating this to zero gives a so called wave equation, this is partial dierential equation which when it is solved (we dont know yet how to solve this equation but it doesnt matter) tells us that the quantity propagates through space at a nite speed c, just like electromagnetic waves. For Einstein, it was clear as early as 1907 that the correct description of gravity would not involve an adaptation of Newtonss eld equation (2) to the framework of special relativity. Contrary to many theoretical physicists,

he was thinking that instead, a description of gravity would involve a generalization of special relativity. While many theoretical physicist were trying to t gravity in the framework of special relativity, Einstein became aware very soon that gravity was the syndrome of the failure of special relativity to pretend to be the general geometrical framework for physics. Gravity does not t into special relativity although special relativity works pretty well without gravity. The starting point of Einsteins elaboration of general relativity is the notion of a local inertial frame of reference. Since an observer in a local freefalling reference frame near the Earth does not feel any gravitational force at all, Einstein was sure that local free falling frames (local inertial frames), would be in some sense the domain of special relativity, and gravity should somehow be described by the relative acceleration of such frames. It required 8 years for Einstein to formulate the theory of general relativity. It took him 4 years to conquer his aversion to Minkovskis (1908) geometric formulation of special relativity and to realize that the geometrical framework of general relativity would involve a curvature of the Minkovskis 4-dimensional spacetime. It took an additional 4 years for Einstein to master dierential geometry and to make the theory of gravity logically self-consistent.

1.5

Local inertial frames; strong equivalence principle

In special relativity we encountered the notion of an inertial (or Lorentz) frame of reference. It is a latticework of rods and clocks. A local inertial frame is the interior of a cage with the latticework inside the cage and the cage is small enough to neglect the eect of the inhomogeneity of the gravity. This is the reason why it is called a local inertial frame. The space-time coordinates, say t, x, y, z , that the latticework provides are called local inertial coordinates. For instance a free-falling lift, provided it is not too big, is a local inertial frame. Also, a near Earth space-shuttle, in which free-oating astronauts are carrying experiments in zero-gravity, is a local inertial frame. A modern version of the strong equivalence principle says the specic forms that the (non-gravitational) laws of physics take in a local inertial frame are the same as they take in the (global) inertial frames of special relativity and all the local laws of physics are the same in every local inertial frames. So, Einsteins strong equivalence principle states that the laws of physics within all local inertial frames are equivalent and can be described using the Minkovski metric, which gives the invariant interval between neighboring events. 5

1.6

Gravity as a curvature of space-time

Let us consider a family of adjacent local inertial frames (lifts with oating scientists inside) scattered around the Earth. We imagine them initially to be at rest with respect to each other. Under the action of gravity, they fall, and therefore, they acquire a relative velocity. Inhomogeneity of the Earths gravitational eld is the cause of this relative velocity. Because of their relative velocity, these local inertial frames cannot be meshed to form a global inertial frame. There is here a powerful analogy : local Euclidian coordinates on the Earths surface cannot be meshed to form a global coordinate system and the reason is the curvature of the Earth surface. So, since this is gravity that prevents the meshing of local inertial frames in space-time and since this is also the curvature of space that prevents the meshing of local Euclidian coordinates (see below), it must be that gravity is a manifestation of spacetime curvature. Exercise : consider the 2D metric of the Earths surface in a spherical polar coordinate system, ds2 = R2 d2 + R2 sin2 d2 (7)

Here R is the radius of the Earth or the radius of curvature of the Earths surface. Recall that the vector between (r, , ) and (r + dr, + d, + d) in spherical coordinates is dr = drer + rde + r sin de . So we obtain (7) from ds2 = dr.dr with r = R being a constant. After a change of variables, x = R, ), 2 it is not hard to show that this line element takes the form: y = R( ds2 = cos2 (y/R)dx2 + dy 2 . (8) (9)

(10)

This is the same equation as (7) using dierent variables, (x, y ) instead of (, ). To obtain it we have used that = x/R, = /2 y/R, that d = dx/R, d = dy/R since R is a constant, hence d2 = dx2 /R2 and d2 = dy 2 /R2 , we have also used the trigonometric relation sin(/2 ) = cos . Now, Taylor expands the cosine function in (10) around y = 0 : cos(y/R) = 1 + O(y 2 /R2 ), 6 (11)

where O(y 2 /R2 ) means terms of the order of (y 2 /R2 ) or smaller. Therefore the expression for the metric of the surface of the earth takes the local form (close to y = 0) ds2 = dx2 + dy 2 + O(y 2 /R2 )dx2 . (12) By doing so, we have approximated the metric of the Earths surface by the Euclidian metric close to the line y = 0 (the equator). However, as one moves from the equator by an amount y , the two metric coecients start to dier by an amount O(y 2 /R2 ). So, local Euclidian coordinates can be meshed along the equator but not across it. What prevents this meshing is the curvature of the Earths surface. The metric coecients of a curved spherical surface diers from the 2D Euclidian metric coecients ij by an amount gij = O[(distance from great circle)2 /(radius of curvature)2 ]. Remark on notations. In the denition of a metric we used notations of this type: dx2 which really means (dx)2 and we dropped parenthesis when it was obvious; however it is a bit ambiguous because dx2 could also mean d(x2 ) and in this case d(x2 ) = 2xdx. So recall that, here, dx2 , d2 etc. really mean (dx)2 , (d)2 etc.

Gravitational time dilatation


ds2 = c2 dt2 [dx2 + dy 2 + dz 2 ], (13)

The Minkovski metric of at spacetime is given by

in a Cartesian coordinate system (change of convention : when ds2 < 0, c < dx/dt for dy = dz = 0). Gravity produces curvature of spacetime, i.e. it aects spacetime geometry. The Schwarzschild metric describes the curvature of spacetime near a stationary isolated object of mass M : ds2 = (1 2GM 2 2 dr2 2 2 2 2 2 ) c dt [ GM + r d + r sin d ] c2 r (1 2c ) 2r (14)

Questions: A) What is the name of this metric when the eect of gravitation is neglected, i.e. for M = 0? B) Show that, corresponding to a coordinate time interval t, the proper time interval , measured by a test particle in a circular equatorial orbit

(i.e. = /2) of radius R and angular velocity , is given by = [1 ( 2GM 2 R2 1/2 + )] t. c2 R c2 (15)

Hint : Recall that ds2 = c2 d 2 C) Neglecting the gravitational eect and introducing the velocity v = R, express (15) in term of the Lorentz factor : by analogy, what is the eect of the gravitation? D) A communication satellite is in circular, geostationary orbit at a height of 35900km above the equator. An atomic clock onboard the satellite (clock S) is initially synchronized with another atomic clock which is in a laboratory at sea level (clock E). After one year what will be the time dierence between the two clocks. 2 2 + 2cR in the parenthesis in (15) Hint : Notice that because the term GM 2 c2 R 2 2 is much smaller than one, we can write = [1 ( GM + 2cR 2 )]t. Use (15) c2 R to rst calculate S E , with S and E the proper time interval associated with clock S and E respectively. Answers: A) The Minkovski metric B) d 2 = (1 2GM 2 1 dr2 2 2 2 2 2 ) dt [ GM + r d + r sin d ] c2 r c2 (1 2c ) 2r d = (1 (16)

r = R, sin = 1, dr = 0, d = 0, d = dt, hence which means that 2GM R2 2 1/2 ) dt, (17) c2 R c2 C) Neglecting the gravitational eect and with v = R : = 1 t which corresponds to time dilatation. We conclude that clocks run slower in a stronger gravitational eld. (comment : this gravitational time dilatation eect is weaker at higher R ( distance from the mass) due to weaker spacetime curvature, i.e. weaker gravitation). D) 1 2 (R2 R2 ) GME 1 )]t (18) S E = [ E E 2 S + 2 ( 2c c RE RS RE = 6.378 106 m; E = 7.273 105 rad.s1 ; RS = 3.95 107 m : S E = 5.41010 t=0.017 s/year. Conclusion: We understand now that time does not ow at the same rate for an object close or far from a mass. 8

Gravitational bending of light

Let us consider a free-falling lift (a local inertial frame). Light is emitted from one side of the lift and received by a detector at the other side of the lift. An observer (A) in the lift see the light travel in a straight line. However, an observer (B) outside the lift sees the lift accelerated downward and, therefore, he sees the photon travel in a curved path between the source and the detector. Observer (B) concludes that light follows a curved path near a gravitational body. (B) decides to give (A) a phone call to share his discovery. Call L the distance between the source and the detector. It takes a time t = L/c for the light to hit the detector. During this time the detector has fallen a distance z = gt2 /2. The angle of deection from the straight line is thus given by tan = z/L. It is gL 2c 2 (19)

For a lift of size 3 m in the Earth gravity, 9.81 3/(2 9 1016 ) = 1.6 1016 rad.

Gravitational Doppler shift of light

On a rainbow, red is on one side and blue on the other side. Red corresponds to large wavelength and blue to small wavelength. We know that wavelength and frequency are inversely proportional to each other: = c. (20)

And according to the Planck-Einstein relation, E = h , i.e. energy is proportional to frequency, therefore, red corresponds to the low energy and blue to the high energy part of the visible spectrum. The wavelength (or frequency) of light can change while light travels, i.e. from the region it is emitted to the region it is received. There are many reasons for this, but one well-known reason is the Doppler shift. For instance, we know that light emitted from a body that moves away from an observer, has wavelength which is larger in the frame of the observer than in the reference frame of the emitting body. Such Doppler shift toward long wavelength is a redshift. When this Doppler shift is due to the eect of gravity, this is called a gravitational Doppler shift. 9

As we shall see, and should remember, there is gravitational redshift of light when it moves away from a massive body, i.e. as light propagates upwards against a gravitational eld. In other words, an electromagnetic wave emitted from a region of strong gravity will be found to be of larger wavelength when received by an observer in a region of weaker gravity. Light emitted from the Sun suers a gravitational redshift as it moves toward the Earth. As light climbs uphill out of a gravity well (produced by the sun), it looses energy, has smaller frequency, increases wavelength. On the contrary, light that goes in a region of stronger gravity, gains energy and is said to be blueshifted. The shift in wavelength is dened as r e , e (21)

with r means received and e emitted. Since = c, it means that d + d = 0, hence, the shift in frequency is related to the shift in wavelength by: = (22) We will demonstrate below that, when light travels vertically away from a Star (or any massive object), from r to r + dr, it changes its frequency according to d g (r) = 2 dr (23) c with g (r) being the intensity of the gravity eld at r. And we know that at a distance r from an object of mass M, g (r) = GM/r2 , this means that d/ = (GM/c2 r2 )dr Consider a lift (or a lif, local inertial frame) of height h. A photon is emitted at the oor of the lift when the lift begins to fall under the action of gravity. For an observer A inside the lift the frequency of the photon is the same when it strikes the ceiling as when it was emitted. However, a terrestrial observer B which sees the lift moving away from him, will measure a redshift of the light. Thus B can conclude that it is gravity that redshifts light. Observer B makes the following observation: the photon takes a time t = h/c to reach the ceiling. At time t, the lift has vertical velocity v = gt = gh/c. Classically, the shift is given by v / , c 10 (24)

assuming v/c small. Hence, the gravitational redshift is given by gh 2, c (25)

which is similar to (23). In 1960, at Harvard University, a test of General relativity was carried out in which a gamma ray from an unstable iron isotope was emitted at the bottom of a 22.6 m tower. The fractional change of frequency detected at the top is / = gh/c2 = 9.98 22.6/9 1016 = 2.5 1015 . The frequency received was smaller than the emitted frequency, i.e larger wavelength at the top: redshift. A star of mass M and radius R emits photons. What is the gravitational redshift of these photons at distance innity from the star? To answer this question we need to integrate (23) :

d GM dr = . c2 r 2 R

(26)

GM )= 2 , (27) 0 cR with 0 and being the frequency of the photons on the stars surface and at innity. Dening / = ( 0 )/0 , we see that ln( /0 ) = ln(1+/ ). When / is much smaller than one, ln(1 + / ) / . Hence ln( GM 2 cR (28)

Hence we nd that :

More on curvature: geodesics and intrinsic curvature

A curve is the simplest curved geometrical object (it has no intrinsic curvature). First, let us see how to calculate the length of a curve. The curve is dened by y = f (x). The element of length of this curve is ds with : ds2 = dx2 + dy 2 , dx2 + dy 2 , hence,

(29)

or ds =

ds = dx 1 + ( 11

dy 2 ). dx

(30)

The length of the curve y = f (x) between the points x = a, y = f (a) and b x = b, y = f (b) is L = a ds with dy/dx f (x). For a curve parametrized by t, i.e. the locus of points x = x(t), y = y (t) with x(t) and y (t) being given functions, it is easy to verify that

ds = dt (

dx 2 dy ) + ( )2 dt dt

(31)

This generalizes to more dimensions: if the curve lives in 3D or more. The locus of points x(t) = R cos t, y (t) = R sin t is a circle of radius R and it can be veried that the circumference of this circle is 2R. A straight line x(t) in the (x, t) plane has constant slope, i.e. dx/dt = C , it also means that d2 x/dt2 = 0. By denition, geodesics are lines of minimal length between two points (it is true on a curved surface and in curved space of arbitrary dimension). Hence, geodesics in at space are straight lines, but on a sphere, the geodesic are great circles. To determine a geodesic on a surface one can take a thin strip of paper and stick it on this surface without folding the paper (in fact, keeping the strip tangent to the surface). To determine the intrinsic curvature (or Gaussian curvature) at one point on a surface : take a geodesic circle of radius r centered at that point (a geodesic circle is the locus of points which can be joined by geodesics of length r to a common origin), then determine the length (circumference) C (r) of the geodesic circle. The intrinsic curvature K is given by : K= 3 2r C (r) limr0 r3 (32)

It has the dimension of the inverse of the square of a length :1/L2 , it is local quantity. The Gaussian curvature of a sphere of radius R is 1/R2 . The Gaussian curvature of a plan is zero. A piece of paper cannot be bent onto a sphere without crumpling, they dont have the same curvature. However, a thin strip of paper can be stick on a sphere along any great circle (or along any geodesic on a curved surface). Recall that the meshing of local Cartesian coordinate systems on a spherical surface is possible on any great circle. Two other examples: a cylinder has zero curvature, a saddle has negative curvature (in fact K = K1 K2 with K1 , K2 the principle curvatures).

12

Black holes formation

Recall that the Schwarzschild metric, which describes the curvature of spacetime near a isolated body of mass M is: ds2 = (1 2GM 2 2 dr2 2 2 2 2 2 ) c dt [ 2GM + r d + r sin d ] c2 r (1 c2 r ) (33)

The Schwarzschild radius is dened as: 2GM (34) c2 Usually, Rs R, where R is the radius of the massive object. Rs is 3 km for the Sun and 10 mm for the Earth. The Schwarzschild metric only applies in space outside the gravitating body. Inside the body a dierent metric applies. Star more massive than the Sun, with mass larger than 3 times the mass of the Sun, will ultimately collapse. Eventually this gravitational collapse cause the stars radius to be less than Rs . An object which is smaller than its Schwarzschild radius is called a black hole. Any particle or photon inside Rs cannot escape. It must fall into the centre. This collapse continues until the object becomes a point singularity at r = 0 (where the theory breaks down). Proof: rst recall that two neighbouring events cannot lie on the world line of a particle or light ray if ds2 < 0 (in which case the proper time d 2 = ds2 /c2 is imaginary). Suppose a particle at rest inside Rs : hence GM 2 dr = d = d = 0, this means that ds2 = (1 2c < 0 when 2 r ) and ds r < Rs . Which is not possible. Therefore, inside a black hole there are no statics, only dynamics. Moreover, it can be shown that all particles move inwards. Rs =

13

Vous aimerez peut-être aussi