You are on page 1of 489
GRAVITATION: an introduction to current research Edited by: Louis Witten RIAS, a division of the Martin Company John Wiley & Sons, Inc., New York + London Copyright © 1962 by John Wiley & Sons, Inc. All Rights Reserved. This book or any part thereof must not be reproduced in any form without the written permission of the publisher. Library of Congress Catalog Card Number: 62-15194 Printed in the United States of America Preface This book represents an attempt to describe recent progress in research associated with various features of gravitation as described by the general theory of relativity. The collection of chapters contained herein is aimed at permitting physicists, engineers, and mathematicians who have already acquired some knowledge of Einstein’s gravitational theory to become familiar with the current trends and status of the theory either as part of their scientific background or as a guide to serious study and active personal participation in the field. Honoring the central role played by experiments in physical theory, the first chapter describes the present experimental situation regarding gravi- tation. Following this is a group of chapters pertaining to solutions of the field equations of the general theory of relativity. Chapter 2 deals with exact solutions and with methods and tools concerned with their inter- pretation and the determination of their properties. Chapter 3 deals with approximate solutions and describes how the equations of motion of Matter are a consequence of the field equations. The demonstration that the gravitational field equations alone are sufficient to determine correctly the equations of motion of particles is one of the greatest theoretical advances made in the study of the general theory of relativity since the original statement of the theory. The fourth chapter discusses the Cauchy problem, how to give data on a space-like hypersurface at a given time in such a way that at later times solutions of the field equations can be found which are consistent with the given data. Next come a group of chapters dealing with some qualitative aspects of the theory. As we see in the first chapter, experiments in gravitation provide too meager a basis for deciding what phenomena predicted by the theory are worthy of special attention. However, we know from other studies in physics that conservation laws and radiation are very important concepts. Their properties in gravitational theory are examined in Chapters 5 and 6 respectively. The following chapter deals with the dynamics of general relativity and discusses how this theory can be cast into a Hamiltonian form. The theory is thereby reduced to a normal field v vi Preface theoretic form, including only independent canonical variables (which obey the standard Poisson bracket relations). The methods of usual field theory are then applied in discussing concepts such as energy and radia- tion; many of the usual properties can thus be carried over to the Einstein field. The two parts of Chapter 8 describe an approach to the quantization of the gravitational field based on a theory of measurement. Quantum phenomena and their description by the quantized theory of fields play a central role in modern attempts at describing nature in the microscopic domain. Since quantized fields presumably serve as sources for the gravitational field, it is widely believed that a consistent picture of nature demands the quantization of the gravitational field. Hence, much of today’s research is motivated directly or indirectly by problems relating to the quantization program. Indeed, the work described in Chapters 4 through 8 of this book falls largely in this category. A proper survey of the various attempts at quantizing the gravitational field would require a book of its own. Instead of presenting such a survey, Chapter 8 describes the approach of and progress made by its author in such quantization. Chapter 9 describes how the combined theory of gravitation and source- free electromagnetism can be given a completely geometric description. This is followed by a chapter which shows in outline how most of the concepts of physics can be given meaning without the explicit requirement of singularities in the physical fields. Indeed, they can apparently be given acompletely geometric description. The last chapter discusses the use of gravitational theory in cosmology. A book of this type must make some omissions. Perhaps one of these is Mach’s principle, which, though discussed at various places in the book, is nowhere given a full treatment. The reason for this is primarily that there is no unified approach or program regarding the meaning of this principle. Most physicists will say something to the effect that Mach’s principle means that the inertia of an object depends on the existence and distribution of other objects in the universe. Each individual physicist then takes off from this statement in a direction of his own—which may lead him to conclude that the universe must be spatially closed, closed in all four space-time directions, or flat at spatial infinity in a certain way; or he may say that Mach’s principle does not deal with these boundary problems at all but leads to local spatial anisotropies or lack thereof; and so forth. Another omission is that the many attempts at deriving a unified field theory are not discussed in the book. We have rather stayed with Einstein’s original formulation of the general theory of relativity and its con- sequences. However, this restriction does allow the unification of source- free electromagnetism and gravitation discussed in Chapters 9 and 10. Preface vii Instead of the usual type of index, the book contains at the back a complete outline of the contents of each chapter. This should enable the reader to pick out those parts of the book in which he is most interested or which he wishes to reserve for extra consideration. It also gives a ready view of the complete content and coverage of the book. Because of the diverse nature of the chapters, there should be very little difficulty in picking out the chapter of special interest; the accompanying outline should render an index unnecessary. I should like to express my gratitude to the Aeronautical Research Laboratory of the United States Air Force for its continued support during the preparation of this book. Louis WITTEN Baltimore, Maryland June, 1962 Contents 1 Experiments on Gravitation, 1 Bruno Bertotti, Dieter Brill, and Robert Krotkov 2 Exact Solutions of the Gravitational Field Equations, 49 Jiirgen Ehlers and Wolfgang Kundt 3. The Equations of Motion, 102 Joshua N. Goldberg 4 The Cauchy Problem, 130 Yvonne Bruhat 5 Conservation Laws in General Relativity, 169 Andrzej Trautman 6 Gravitational Radiation, 199 F. A. E, Pirani 7 The Dynamics of General Relativity, 227 R. Arnowitt, S. Deser, and C. W. Misner 8 The Quantization of Geometry, 266 Bryce S. DeWitt ix 10 Contents A Geometric Theory of the Electromagnetic and Gravitational Fields, 382 Louis Witten Geometrodynamics, 412 John G. Fletcher Relativistic Cosmology, 438 O. Heckmann and E. Schiicking chapter 1- Bruno Bertotti,* Dieter Brill,t and Robert Krotkovt Experiments on Gravitation This chapter is devoted to a review of the connections between theory and experiments in the physics of gravitation. In Section 1-1 we stress the theoretical importance of an invariant definition of any observable quantity and discuss the idea of inertial frame of reference. Section 1-2 contains an outline of the fundamental concepts and experiments which play a role in an understanding of gravitation. In Section 1-3 we sum- marize the existing evidence concerning the three specific tests of general relativity: the gravitational frequency shift, the deflection of light rays, and the anomalous advance of the perihelion of a planet. No mention is made of the connections between general relativity and cosmological theories, to which a special chapter of this book is devoted. We wish to express our gratitude to Professor R. H. Dicke for his help and his interest; we acknowledge also an enlightening correspondence with Professor G. M. Clemence in connection with astronomical observations. 1-1. The Theory of Measurements in General Relativity 1-1.1. Observables are invariant, A scientist who is familiar with other physical theories in which the symbols one deals with are endowed with a direct experimental meaning must provide himself with great care and patience in studying general relativity. Thermodynamics, for example, deals with quantities—temperature, volume, etc.—which are easily and precisely defined; thus the mathematical development has at any \ * Laboratorio Gas Ionizzati, Frascati, Rome, Italy. The chapter was written while at the Plasma Physics Laboratory, Princeton, New Jersey. + Palmer Physical Laboratory, Princeton University, Princeton, New Jersey. 1 2 Experiments on Gravitation step a direct physicalinterpretation. At the logical beginning of the theory of relativity we find instead only two elementary concepts: the idea of space-time coincidences (“events”) and the proper time: this is all there is in our equipment for the long journey to a complete understanding of gravitation. Every other physical quantity—distance, angle, energy, etc.—has only a secondary meaning and must be constructed, if possible, from the two fundamental concepts; hence, a full theory of measurements isneeded. True, every physical theory, being in itself only a mathematical model, should be given a sound operational basis; but this program meets with particular and basic difficulties in a generally covariant theory. We take the point of view that the coordinates are only an arbitrary way to label the events of space-time: any one has the right of doing it as he likes. The often heard phrase ‘‘a frame of reference attached to an observer” has only a restricted and local meaning which needs a careful explanation (see further). We are concerned with the theory of a Rie- mannian manifold, and only geometrically defined objects have a right to citizenship within its realm; hence, any observable quantity must be invariantly defined. \t may be asked then, what does one mean by measuring (say) the x component of a vector. The very fact that the description of the experiment involves the specification of what is the x direction shows that in this case the observable is the projection of the vector along a direction fixed by the instrument itself, which is an invariant. The measuring apparatus determines schematically some geometrical elements in space-time; for example, an ideal clock is nothing but a time- like world-line on which a set of equidistant pulses is marked; a rigid rod corresponds to two equidistant world-lines, and so forth. The interaction of the observed object with these elements leads then to well-defined invariants. One might point out that in practice the usual experiments about gravitation are performed in a naive way, without any sophisticated definition of the observed quantities. This, however, is in general allowed only as an approximate consequence of the smallness of the region of space and time involved, compared with the local radius of curvature of space- time. We shall discuss later the main experiments on gravitation and indicate that in fact a rigorous theory of measurement would add nothing essentially new to their physical content. The point we wish to make here is important only in principle. 1-1.2. Local measurements. It is important to recall that every experimental method whose theory is based on special relativity can be still validly carried out in a small region of a curved space-time. Thus the measurement of a proper time interval can in principle be done locally by The Theory of Measurements in General Relativity 3 a geometrical construction based on geodesic lines, as explained by J. Fletcher in his chapter “Geometrodynamics.” The clock one normally uses, however, is provided either by atomic energy levels or by astronomi- cal observations. The latter method leads to the “ephemeris time,” defined as the variable with respect to which the planetary motion best approaches the motion calculated theoretically (see, for example, Cle- mence, 1956 and 1957). It must be stressed that in principle there is no reason to expect that these two methods agree with each other or with a geodesic clock; we shall mention later the deep implications of possible disagreements.* Fig. 1-1. Measuring a space-like interval: OP = —cVini, t= |B t=0 t= a4 (Small) spatial distances can be measured in the following way. An observer, provided with an ideal clock, sends at the time — ¢, a light pulse tothe mirror at P and receives it back at the time ¢ (see Fig. 1-1). The distance between P and the origin O is then OP = —eWtyfz. The maxi- mum value of OP, for fixed P and /, corresponds to the choice of O for which OPis orthogonal to / (Synge, 1956, p. 24). A standard for angles is then provided by the application of elementary geometry. * On the connection between time measurements and Mach’s principle, see also Bertotti, 1962. 4 Experiments on Gravitation 1-1.3. Inertial frames. We would like to discuss now the notion of “inertial” frame of reference in general relativity; and whether the concept of “‘absolute rotation” has any meaning at all. In special relativity one can define an inertial’ frame in three different ways: (a) geometrically: the coordinate lines form an orthogonal mesh with parallel sides; (6) dynamically: the acceleration of a free observer vanishes; (c) astronomically: distant stars are, in the average, at rest. L Fig. 1-2. The locally inertial frame of reference, A® da A@® The complete generalization of (a) to general relativity is impossible: a Riemannian manifold where a Cartesian coordinate system can be set up is, ipso facto, flat. One can, however, restrict oneself to the neighborhood of an observer and define a Jocally inertial frame of reference in the following way. We choose the observer’s world-line / as a time coordinate line; while the spatial coordinates issuing from any point of / shall be orthogonal to / and to each other. The orientation of the spatial axes as one proceeds along / must then be determined by introducing the notion of a non-rotating local frame of reference. The most obvious choice is to take the three unit vectors Aj) (i = 1, 2,3) in such a way that their covariant change along / occurs entirely in the direction of / itself; that is to say, 5A) BNy* _ c= wr, =e (1-1.1) * Greek indices range from 0 to 3, 8/8s indicates the covariant derivative along /; the metric is of the type (+ — — —). A comma and a bar denote ordinary and covariant derivatives respectively. The Theory of Measurements in General Relativity 5 This mode of propagation was called Fermi propagation by Walker (1935); when / is a geodesic, it reduces to ordinary parallel propagation. Form- ally, in this frame of reference, the metric at / satisfies (war = (1-1-1, -1)3 Bu pdarr = 0 (1-1.2) The reasonableness of this definition, of course, must be checked from its physical consequences. To discuss the dynamical definition (6) we must introduce the idea of relative motion. Let »(s) be the orthogonal distance between two neigh- boring geodesics /and/’. If A(s) is the unit vector along /, it can be proved (see, e.g., Synge and Schild, 1949, §3.3): 82qe 85? + Rigg? = 0 (1-1.3) (equation of geodesic deviation). The relative velocity and acceleration with respect to our inertial frame are obtained by (proper) time differentia- tion of the invariants nw = Mote (1-14) From (1-1.1) and (1-1.2), (1-1.3) reduces to nee Go + RuypoMyWPX = 0 (1-1.5) Fig. 1-3. Measuring the curvature of space-time with two freely falling mirrors. 6 Experiments on Gravitation which has a clear physical interpretation: the relative acceleration between two neighboring geodesics vanishes if space-time is flat and is otherwise proportional to a linear combination of the second derivatives of the gravitational potentials g,,; in other words, there are no inertial forces (Pirani, 1956). The argument can obviously be reversed: if we know that the local frame we are using does not rotate, the observation of the relative acceleration of a nearby free particle determines, according to (1-1.5), three components of the full Riemann tensor. We mention in passing an interesting experiment of this type, which consists in sending a light pulse back and forth between two freely falling mirrors; the measure- ment of the three time intervals 25,, 252, and 2s3 determined by the reflections on / leads, then, to the determination of the second derivative of the orthogonal distance (s) between the two mirrors: ay 1 1, $381 (2), = a5, Se soe (1-1.6) This formula can be easily obtained by expanding the function In 7(s) around s = 52 up to terms of the second order and assuming that the angle between the two lines is an infinitesimal of the same order as 52. Denoting by 1 x 7” the unit vector in the direction of y, one finds from (1-1.3) and (1-1.6), with an easy calculation: ae Se + am SE = Ryan NAPA" (-1.7) The expression in the right side i is nothing but the Gaussian curvature of element of geodesic surface determined by / and /’; the first term on the left is essentially the relative angular velocity.* We would like to point out, however, that an experiment of this kind could hardly be interpreted as a proof that space-time is curved. From the dynamical point of view, (1-1.3) is nothing but a sophisticated formulation of the law of relative gravitational motion, which must hold in any case; since an observable gravitational field is, by definition, inhomogeneous, it is no wonder that the distance between the two mirrors does not change linearly. To come back to our main point, the absence of rotation in a Fermi propagated frame can also be tested by considering a spinning test particle: one would expect no change in the orientation of its angular momentum vector relative to the basis Aj). A study of the equations of motion for such a particle confirms that this is actually the case (Pirani, 1956). * This effect was pointed out for the two-dimensional case by Wigner (1957 and 1960). The Theory of Measurements in General Relativity 7 We now turn to (c) and find that, in general, the frame of reference previously defined will not show the distant stars on the average at rest. Notice, in fact, that our definition was precisely a local one, while the astronomical criterion involves the behavior of the metric tensor all the way between the stars and the observer; it can be expected that a gravita- tional field which is not static for the observer we consider may twist the null rays so as to produce a net overall rotation. A good example of this is given by a circular time-like geodesic (r = const) in a Schwarzschild field. The equation for parallel propagation (1-1.1) shows after some t Co Fig. 1-4, The twisting of a Fermi-propagated frame in a Schwarzschild field. calculation that the projection on a hyperplane ¢ = const of a vector attached to the geodesic will make after one revolution an angle of the order of Gm/cr with itself. The local definition of absence of rotation, therefore, does not agree with the other one provided by the time lines of the static field; even worse, since the angle depends on the distance r, two different “locally inertial” frames may disagree with each other. It is clear therefore that this form of Mach’s principle, concerned only with the actual frame of reference to be adopted in the laboratory, is not well founded; which, of course, does not exclude a deeper form of interaction between the local metric properties and the universe as a whole. 1-1.4, Astronomicel measurements. To conclude this section we briefly mention the problem of the theoretical definition of astronomical quantities. Mast and Strathdee (1959) discussed the connection between the actual motion of the star and its observation by optical means. The distance of a star has been defined in an invariant way by Whittaker (1931) (see also Kermack et al., 1932, and Robertson, 1933) in the following 8 Experiments on Gravitation way: an infinitesimal pencil of light emitted by a point source inter- sects an area S on a two-dimensional surface orthogonal to the ray itself; the distance d of the source is then taken to be proportional to VS. One arrives at the same result if one considers d to be proportional to the linear separation between two given light rays (Newman and Goldberg, 1959). Both these methods, however, can not be actually used in practice if there is not a well-defined pencil of light. The method of parallax, with which the fundamental scale of astronomical distances is established, has not, as far as we know, so far been investigated from the point of general relativity. The definition of apparent luminosity in terms of flux of the Poynting vector has been given by Joseph (1958). 1-2. Fundamental Ideas and Experiments 1-2.1. The weak and strong principles of equivalence. In order to assess clearly the theoretical implications of Eétvés’s and other fundamental experiments, it is necessary to distinguish between the “weak” and the “strong” form of the principle of equivalence. Eétvés’s experiment can verify the former (see, e.g., Dicke, 1959a), which states thatthe trajectory of atest particle, under the influence of gravitational fields only, depends only on its initial position and velocity, but not on its mass and its nature. As a matter of fact, for the sake of logical clarity, we may take the weak principle of equivalence as a definition of the gravitational forces; Edtvés’s experiment then tests simply the extent to which the interaction between neutral bodies is confined to gravitation. The strong form of the principle of equivalence involves two additional assumptions: (a) At every event of space-time there exists a class of local, “inertial” frames of reference (equivalent to within a Lorentz trans- formation), in which all the physical laws take on a standard, “flat-space”’ form—the same they would take in the absence of gravitation; in other words, all effects of gravity (except for “tidal” effects) disappear in a freely falling laboratory. (6) The flat-space form and the dimensionless constants it contains are the same throughout the universe. According to this view it is not possible, by means of /ocal experiments, to distinguish between different positions of the laboratory in space-time, nor to find out anything about the geometrical properties of the universe as a whole. The strong principle of equivalence has the advantage of simplicity; also, our understanding of astrophysics makes it implausible that physical laws in other parts of the universe are radically different from what they are here and now (Minkowski and Wilson, 1956). Only recently, how- ever, some thought has been given to the outstanding problem of assessing directly the extent of its validity (see below). Fundamental Ideas and Experiments 9 How could the strong principle be violated? The assumption (a), for example, would not hold for a physical law which contains explicitly the curvature tensor. A change in the physical laws from point to point, which would contradict (5), could be described in terms of changes of the “dimensionless coupling constants” which describe the strength of different interactions (e.g., the ratio of gravitational to electrostatic force between an electron and a proton: Gm,m,/e? ~ 10-4°), A variation in a dimensionless coupling constant has an invariant meaning; but it can be interpreted as a variation in G, e, m,, etc., depending on the units used to measure these quantities. It has been argued that if the Eétvds experiment were to give an exactly null result, the dimensionless coupling constants could not vary with position (Dicke, 1957; Schiff, 1959, 1960). The argument is essentially the following. A varying coupling constant would lead to a binding energy dependent on position. Hence the work required to lift a body in a gravitational field would not be proportional to its energy (i.e., inertial mass), and this anomalous gravitational mass would result in a net, additional acceleration for a freely falling body which in general need not be the same for all bodies. Actually the result of the Eétvds experiment is known only with a finite accuracy, which is great enough to make it unlikely that the strong coupling constants are position dependent; but it allows no conclusion to be drawn about the weak and gravitational interactions, The argument given above can be criticized on the following ground: if nature can really be described in terms of fields, a variation in a coupling constant is presumably due to an interaction with some presently unknown “cosmological” field. Although the argument stated above depends on energy conservation it does not take into account energy exchange with this cosmological field. In other words, if the coupling constant were to be thought of as depending on a given external field, the energy of a nucleus would not be conserved. We must also remark that equality between inertial mass and total rest energy has not been tested experi- mentally with the same accuracy as the null outcome of the Eétvds experiment (Dicke, 19605). 1-2.2. Dirac’s hypothesis, About twenty-four years ago (1938) Dirac pointed out a possible connection between the dimensionless coupling constants of nature and was able to make some reasonable conjectures about how they might vary, if they vary at all. The relevant constants are listed in Table 1-1. Dirac made the suggestion that in the light of an as yet unknown fundamental theory the numbers of Table 1-1 may be connected in such a way that their relative orders of magnitude 10 Experiments on Gravitation remain unchanged in time. Since 7, the age of the universe, is about 10!0 years, all dimensionless constants of order of magnitude 10-40 would be expected to decrease by a part in 10! per year. Hence, in atomic units, G would decrease yearly by this amount. Arguments similar to Dirac’s would lead us to expect that the weak coupling constant (also listed in Table 1-1) might decrease inversely as the square root of the age T TABLE 1-1. Dimensionless Constants in Nature Coupling constant for strong interactions = 10 10 Fine structure constant = a = 1/137 262 10-20 | Weak coupling constant = 3 x 10-14 = (&) Ratio of gravitational to electrical force in hydrogen atom = Gmnetty = 5x 10-4 e 10-40 Ratio of an atomic period to age of universe — elmec® _ 40 =a = 0.2 x 10 = 10-80 | (Number of particles in universe)-! ~ Mr = 3x 10-79 $rR3p {Notation: G = Newtonian gravitational constant = 6.67 x 10-® cgs units m, = mass of electron, m, = mass of proton, m, = mass of pion R = Hubble radius of universe = cT = 1028 cm p = mnass density of universe ~ 10-30 gm/cm} g = beta decay coupling constant = 1.4 x 10-49 erg cm3] of the universe. In addition, a variation of G with T may imply a weak time-variation of the fine structure constant «. Landau (1955) has suggested that if gravitation provides the cutoff needed to give finite results in quantum electrodynamics, « would be related to InG, and would therefore be expected to vary as In (1/T) within a Dirac type cosmology. P. Jordan (1952) has developed a detailed theory (an extension of general relativity) which includes a space-time dependent gravitational constant. This theory has been discussed by Fierz (1956); similar ideas have been developed by Brans (1961) and Dicke (19615). ©. Klein (1958) has sug- gested an explanation of some of the numbers in Table 1-1, sometimes called Eddington numbers, in terms of general relativity, without violating the strong equivalence principle. Fundamental Ideas and Experiments 11 1-2.3. Mach’s principle. Mach’s principle was originally conceived as an explanation of the origin of inertia; it is suggestive of a number of different conjectures and considerations which connect the local physical laws as observed in a terrestrial laboratory with the properties of the universe as a whole. The simplest question one can ask along this line is, why are the local values of the metric field so close to Euclidean? One notices that in a closed, spherical world the total matter content M of the universe deter- mines through the field equations its radius of curvature R, according to the formula GM 1s OR (1-2.1) (Einstein, 1955).* Equation (1-2.1) may be taken to suggest that the actual amount of matter present in the universe determines the order of magnitude (i.e., 1) of the components of the metric tensor. This beautiful example of a “‘ Machian” property can also be interpreted by imagining an integral form of Einstein’s equation. Following the electrostatic analog, a typical value of a metric coefficient would then be given by an expression of the type “gr x [P aqvon Sh” = (1-2.2) and would arise mainly from the universe as a whole; nearby matter would be responsible for the small deviations from flatness. A model vector theory of inertia based on considerations such as these has been given by Sciama (1953). By analogy one might expect that variation of dimensionless constants is a similar cosmological effect, one of the consequences of a not yet under- stood long-range field. The local value of a coupling constant would then be mainly determined by the bulk of the matter in the universe, with local matter density resulting in small variations. This principle can then be used to find reasonable values not only for the time variation of dimensionless coupling constants, as does the Dirac hypothesis, but also for their variation in space. For example, a “Machian” integral for the gravitational coupling constant, ef" d (vol) £ ra would show a contribution of order of magnitude M/Rc? due to the * The connection between the relation (1-2.1) and Mach’s principle has also been discussed by Dicke (1958). 12 Experiments on Gravitation universe as a whole, and an amount ~ m/rc? due to the sum, of mass m and ata distance r; this represents a change in G of about a part in 108 due to the presence of the sun, so that the yearly modulation of ~1% in the radius of the earth’s orbit would give rise to a yearly change in G at the earth of about one part in 1010. Along the same lines, one might speculate (Dicke, 1960a) that the active gravitational mass of a body and hence the locally measured value of G would be affected by the mass distribution of the universe in such a way as to depend on the velocity of the laboratory relative to distant matter. (Such an effect might be pictured as due to a Lorentz contraction of distant matter, a spherical mass distribution becoming an elliptical one.) Tf such an effect existed, it might be expected to be of order (v/c)?, where v is the velocity of the laboratory relative to distant matter. Supposing the sun to move at a velocity vg relative to distant matter, the velocity of the earth relative to distant matter would change with a period of one year, with the result that the locally measured value of G would also suffer a yearly change of amplitude 8G/G ~ vgv,g/c? (where vg = 30 km/sec is the orbital velocity of the earth about the sun). If vs were also ~ 30 km/sec, the fractional change in G would be ~ 10-8. It is unlikely that vs > 300 km/sec (Dicke, 1960). Speculations have also been made (Cocconi and Salpeter, 1958) on possible effects of asymmetric matter distributions in the universe on local phenomena via Mach’s principle. If masses contribute to local inertia according to a I/r law, then the mass M in our galaxy would contribute a fraction A mim ~ (M[x)(R/m) ~ 10-3 to the inertial mass of bodies. Here r is the average distance of the masses in the galaxy from the earth, and R and M are radius and mass of the universe respectively. An anisotropy in local inertial mass, due to the anisotropy in the mass distribution of the galaxy would be some fraction of this value. Such an anisotropy of mass would shift energy levels of oriented atoms and nuclei, depending on the direction of orientation with respect to the major axis of the galaxy; it would also broaden lines in the spectrum of non-oriented nuclei. Hughes et al. (1960) were able to place an upper limit of 10-20 on dm/m, by looking for line broadening in a nuclear magnetic resonance experiment on Li? in its ground state; for other less restrictive estimates, see Cocconi and Salpeter (1960) and Sher- win et al. (1960). On the other hand, Dicke (1961a) has argued on the basis of Mach’s principle that no mass anisotropy effect should be observable. It has been pointed out by Dicke (1960a) that if the fine structure Fundamental Ideas and Experiments 13 constant « is indeed logarithmically related to G, as suggested by Landau (1955), the changes in G mentioned above would be accompanied by changes in « about two orders of magnitude smaller: 5a/a ~ 10-2 5G/G. On this basis, a secular decrease in G of a part in 10!° per year would be accompanied by a corresponding drift in « of ~ a part in 10!? per year, while a yearly modulation of G ~ a part in 108 per year would imply a periodic variation in « with amplitude ~ a part in 10!°, The existence of such effects is highly speculative, and the effects are small, but they are susceptible of measurement; such experiments will be described in the next section. 1-2.4. Experimental tests of “‘Machian”’ properties. A possible time variation of G has many geophysical and astronomical consequences which have been discussed by Teller (1948), Jordan (1959), and Dicke (1957); unfortunately, not enough is known about the earth’s and the sun’s remote history to separate the consequences of a varying G from all the other factors. Among the effects which have been considered we may mention the higher temperature of the earth in the past, due to a hotter sun (its luminosity is proportional to G7); the secular acceleration of the moon and planetary motions in general; the heat flow from the earth. Stellar evolution is also affected by a time dependent G, but this problem has not yet been investigated thoroughly. The most hopeful approach to determine whether G varies appears to lie in earth satellite experiments. A program is now under way to com- pare astronomical time, as determined by the motion of the moon, with atomic time (Markowitz et al., 1958), but will not yield answers for some time. Optical observations of an artificial earth satellite at a height of 20,000 to -50,000 km could detect a decrease in G of a part in 10! per year (Hoffmann, Krotkov, and Dicke, 1960); such a satellite could of course also detect a periodic yearly change in G of a part in 108. Another approach to detect a yearly variation in G consists of precision measurements of g, the gravitational acceleration at the surface of the earth. Gravimeters sensitive enough to see a change of one part in 10° are commercially available, but are not sufficiently stable over a year. Two experiments are now (1960) being carried out by members of Dicke’s group at Princeton in an attempt to monitor g over a year to high precision. In particular, W. Hoffmann is using a gravitational pendulum, consisting of a quartz fiber supporting a quartz bob which oscillates about its center of mass at ~22 c/sec. The frequency is monitored electronically and the system should eventually be capable of following changes in g amounting to one part in 108 per year. In another experiment, J. Faller uses a falling plate which is part of an interferometer and designed so as 14 Experiments on Gravitation to make the fringe system insensitive to the inevitable small rotations suffered by the dropping plate. This experiment will give an absolute measure of g, distances being measured in wavelengths, and times by a crystal oscillator. Experiments to test the atomic effects of a varying fine structure con- stant are at the limit of present-day experimental technique. One way of observing the possible changes mentioned above is to compare two atomic clocks whose rates depend on different powers of «. Such comparisons have been made by Bonanomi in Switzerland (unpublished), but his accuracy was not quite high enough to say anything definite about variations in «. The two transitions compared were in cesium and ammonia. The experiments are being continued. Cedarholm et al. (1958) compared two beam-type masers with the beams in opposite directions, and over a year found no anomaly greater than about 10-3(v4vg/c?) where vy is the velocity in the beam and vg is the velocity of the earth. This is just on the edge of being able to rule out the anomalous velocity dependence behavior of « which has been considered. A variation of the fine structure constant with time could have another observable consequence: this is in the results of radioactive dating of rocks and meteorites. If « were to vary with time, there would be small shifts in nuclear energy levels (because of the Coulomb energy) and this in turn would lead to changes in decay rates. [If the strength of the weak inter- action were changing relative to the strong, this would also lead to dis- crepancies among rock ages, determined by different radioactive decays, but this is a small effect (Dicke, 19596).] The rates of radioactive decays involving small energy changes are very sensitive to small shifts in the energy levels, and a decrease in the fine structure constant of the amount mentioned above could lead to large (~30%%) discrepancies between the ages of old (several billion years) rocks as determined by various decays. The effect is rather difficult to evaluate because the Coulomb energies of the interesting nuclei are not sufficiently well known, but preliminary investigations of the rock ages obtained from the decay of Re!87 to Os!87 seem to indicate that these do not differ significantly (Peebles, 1961) from ages obtained by other methods. The parent and daughter states in this transition differ by <5 kev. If further work were to confirm this result, it would be unlikely that the fine structure constant could be decreasing. 1-2.5. Tests of the weak principle; the Eétvés experiment. The “weak” principle of equivalence has been tested quite accurately experi- mentally (to about 3 parts in 10!°). The first tests (apart from Galileo’s apocryphal experiments with balls dropped from the leaning tower of Pisa) used pendulums, and showed that the difference between the inertial Fundamental Ideas and Experiments 15 and gravitational masses of the materials tested was less than one part in 105 (Potter, 1923). One difficulty with such pendulum experiments is that the effect sought is a small difference between two relatively large numbers. Another method which has been proposed is to use objects floating on or in liquids, an anomalous acceleration of the floating object then leading directly to an observable displacement (Rosza and Selenyi, 1931). Baron Edtvés’s method, using a torsion balance, is a null experi- ment, and therefore has a chance of being much more precise (Edtvés, 1922). In Eétvés’s method the two masses to be compared are hung from a thin fiber as shown in Fig. 1-5. (The reason for the L-shaped configuration, Torsion constant 0.5 dyne-cm /radian Length ~ 55 cm Telescope 620m = (1250 scale divisions) Mirror Materials to be compared ~ 25 gm 21cm Fig. 1-5. Diagram showing typical dimensions of Eétvis torsion balance. with one mass below the other, is irrelevant to this experiment.) In the plane perpendicular to the fiber, two kinds of forces act: gravitational and inertial, as shown in Fig. 1-6. Neglecting for the time being the motion of the earth around the sun, the gravitational force is one component of the earth’s pull, while the inertial force is due to the rotation of the earth on its axis. (These forces, in the plane perpendicular to the torsion fiber, correspond to an acceleration of about 1.7 cm/sec? at latitude 45°.) If the gravitational mass of each bob were exactly equal to its inertial mass, the forces would exactly balance and no net torque would be exerted on the fiber. However, if the ratios of the gravitational to inertial mass were different for the two materials being tested, one of them would be acceler- ated by the gravitational field slightly more than the other, so that there would be a net twist in the fiber. Since the anomalous torque would depend on the orientation of the pendulum relative to the earth, this torque could be detected by looking for changes in the wire twist as the whole apparatus (including the observing telescope) was rotated about the torsion fiber. The acceleration of the earth toward the sun is of the same order of 16 Experiments on Gravitation magnitude* as the acceleration associated with its daily rotation. There- fore it is also feasible to look for a diurnal change in the position of the torsion pendulum as a measure of anomalous accelerations. z Gravitational forces point N Inertial forces point S Fig. 1-6. Forces acting on the torsion balance in the plane perpendicular to the torsion fiber (the fiber intersecting the plane of the paper at the center of the above diagram, point F). If gravitational mass = inertial mass for both masses, inertial forces just balance the gravitational ones. Only the inertial forces connected with rotation of the earth on its axis are shown. Such methods can detect very small differences between gravitational and inertial masses. For example, with the typical dimensions shown in Fig. 1-5, one small scale division seen in the telescope corresponds to an angle ~4 x 10-4 radians, or a torque of ~2 x 10-4 dyne-cm. In the plane perpendicular to the fiber the force exerted by the earth on one of the masses is about 43 dynes, which amounts to a torque of 850 dyne-cm. Thus a positive effect equal to one small division would correspond to a ratio of gravitational to inertial mass equal to 1 + 7, where In| x 2 x 10-7, One of the main difficulties encountered is that the pendulum is very sensitive to inhomogeneities in the gravitational field. This last is particularly true of the design shown in Fig. 1-5, which indeed was originally intended for use as a geophysical instrument to detect such inhomogeneities. A near-by human being could give rise to a deflection greater in magnitude than that due to any possible value of ». Other difficulties are slow drifts in the torsion constant, sensitivity to tempera- ture, torques exerted by the earth’s magnetic field on magnetic contamin- ants in the pendulum, and electrostatic forces. * 0.6 cm/sec?. Fundamental Ideas and Experiments 17 Eétvés, Pekar, and Fekete (1922) observed the deflection of the pendu- lum when the whole apparatus was successively rotated in 90° steps. In general, large deflections were observed (~6 scale divisions, equivalent to an 7 of 1.2 x 10-5), These were due to inhomogeneities in the gravi- tational field, and were corrected for by control experiments using two weights of the same material. The statistical error in the means of a large number of observed deflections typically amounted to about 0.01 divisions, or an 7 of about 2 x 10-%. This represents the limit of accuracy of this experiment, assuming that there are no undetected systematic errors. The observations can be combined in such a way that the computed value of 7 is insensitive to small changes in the torsion constant. The apparatus can be further improved by using a torsion balance with more symmetry—the so-called double balance, which consists of two pendu- lums, each like that shown in Fig. 1-5, hanging from two fibers and arranged so that the lower weights are not next to each other. Two of the masses are of the same material, and the other two are the ones to be compared. With this design the results of the experiment can be made insensitive to slow changes in the inhomogeneities in the gravitational field with time. Although torsion pendulums are very sensitive to temperature gradients, Eétvis, Pekar, and Fekete state that changes in ambient temperature, or gradients, could not have significantly affected their results. The torsion pendulum was enclosed in several metal cans for insulation. They do not quote any quantitative estimates of the size of possible gradients, or of possible deflections produced by such gradients. The earth’s magnetic field could also exert appreciable torques unless the materials tested were thoroughly freed of magnetic contaminants; the authors do not quote an upper limit on how big such effects could be. Electrostatic effects can probably be neglected at this accuracy. The results obtained by Eétvés, Pekar, and Fekete are shown in Table 1-2. Table 1-2 also shows the results obtained by Renner (1935) who used a similar torsion balance and method, but somewhat more refined observational technique.* The sensitivity of Renner’s apparatus to changes in ambient temperature was such that a temperature change of 0.1°C (the largest change to be expected) was equivalent to an 4 of 0.6 x 10-9. He made magnetic corrections in comparing brass with bismuth, and found that these were appreciable (leading to a difference of ~1.3 x 10-9 in the observed value of 7). It seems clear from the results in Table 1-2 that if there is an effect for the materials tested, it is less than a part in 10°. * Experiments with a torsion balance have also been carried out by Potter (1927), but with somewhat less accuracy. 18 Experiments on Gravitation In view of the fundamental importance of this experiment it is presently (1960) being repeated by Dicke and his co-workers (Dicke, 1959), using more sophisticated techniques. It is hoped to reach an accuracy of a part in 101, or possibly better. The overall sensitivity to small torques has been increased by using a thinner fiber, of torsion constant ~ 10-2 dyne-cm/ radian. The strength of the fiber goes as the square of the radius, while the torsion constant goes as the fourth power, so it is advantageous to use smaller radii. A large mirror and an electronic detection system make the system sensitive to pendulum rotations of ~ 10-8 radians. The deflection signal is fed back onto electrodes which apply electrostatic forces to keep the pendulum locked into the axis of the telescope to within 10-8 radians. Thermal noise (Brownian motion) can be markedly reduced by evacuating the pendulum containers; other noise, due to disturbances localized in time, can be minimized by electrically controlling the damping. Two of the main difficulties experienced by Eétvés (i.e., gravitational inhomo- geneities and slow changes in the torsion constant) are practically elimin- ated by using a triangular pendulum (see Fig. 1-7) and strong feedback. 0.0005-in, tungsten Magnetic ‘damping cup Bottom plate Fig. 1-7. Triangular pendulum used by Dicke and collaborators to test the equivalence of gravitational and inertial mass (Dicke, 1959). One of the chief remaining troubles is that at the accuracy sought the torsion pendulum is sensitive to temperature changes. To minimize these, the pendulum has been hung in a vacuum and the whole apparatus has been placed underground where the temperature is uniform. Never- theless, the temperature will have to be monitored and corrections made Fundamental Ideas and Experiments 19 for its effects. Care has been taken to avoid all magnetic materials. However, variations in the earth’s magnetic field (amounting to about 10-4 gauss) are still expected to produce torques equivalent to an 7 of 10-19; the earth’s magnetic field will also have to be monitored and corrections made. A preliminary run (Dicke, 1959) has yielded the result |y| < 3 x 10-19 for the difference between the ratios of gravita- tional to inertial mass for gold and aluminum. A number of inferences can be drawn directly from the Edétvés experi- ment. If it is observed that two nuclei (e.g., gold and aluminum) have the same acceleration in a gravitational field to some accuracy (for gold and aluminum, 3 parts in 10!°), only somewhat weaker conclusions can be drawn about the ratios of gravitational to inertial mass for neutrons, TABLE 1-2. Observed Ratios (1 + ) of Gravitational to Inertial Mass Edtvés, Pekar, and Fekete (1922) Material (Compared to Platinum) 7 Gin units of 10-9) Magnalium (90% Al, 10% Mg) +441 Snake wood -1+2 Copper +442 Water -64+3 Copper sulfate crystal -1+3 Copper sulfate solution -3+3 Asbestos +143 Tallow -243 Renner (1935) Materials Compared. 7 Gn units of 10-9) Platinum, brass 0.45 + 0.65 Batavian glass, brass —0.01 + 0.67 Crushed glass, brass +0.21 + 0.65 Bismuth, brass —0.14 + 0.74 Paraffin, brass 0.41 + 0.44 Ammonium fluoride, copper 0.13 + 0.57 Alloy (30% Mg, 70% Cu), copper +012 + 0.22 protons, electrons, and the nuclear binding energy (Wapstra and Nijgh, 1955; Dicke, 1957). For example, roughly half the mass of aluminum is in the form of neutrons while for gold this figure is 60% ; this difference in proportion.is 0.1, so from the observation that gold and aluminum have the same acceleration in a gravitational field to 3 parts in 101° we can conclude that the neutron and proton have the same acceleration to 3 parts 20 Experiments on Gravitation in 10%. Similarly, if the acceleration of an electron were anomalous, this anomaly would be less than 1 part in 10°; for the strong binding energy this figure is 3 parts in 106. The Eétvés experiment should clearly be done with nuclei which differ as much as possible in their binding energy per particle, and the relative numbers of their constituent neutrons and protons. This argument has been extended by Schiff (1959), who pointed out that, since a nucleon is surrounded by virtual particle-antiparticle pairs, an anomalous ratio of gravitational to inertial mass for a positron would imply a positive result for the Eétvés experiment. He estimates that if a positron had negative gravitational mass, the ratios of gravi- tational to inertial mass for aluminum and platinum would differ from one by about 3 x 10-7. Such a possibility clearly seems to be ruled out experimentally. 1-2.6. Gravitational waves. The overall cosmological picture of the universe presents us with a spherical space expanding at a rate deter- mined by the total amount of matter and its average pressure and on which local deviations due to galaxies and stars are superimposed. There is no reason to exclude, however, the presence of collective oscillations of the universe as a whole, which might show up locally as gravitational waves, carrying energy and momentum. To get an idea of their maximum allowable density, one would assume that their energy density contributes to the curvature of the universe, like ordinary matter density; since the latter by itself is not quite sufficient to account for the known value of Hubble’s constant, the energy density of the gravitational waves could be even higher, up to 10-29 g/cm3 (Wheeler, 1958). The main difficulties in detecting such waves lies of course in their presumably very long wave length (possibly of the order of the intergalactic distances). An attempt to detect, if it exists, the short wave length band of the gravitational waves is being carried out by J. Weber and his collaborators at the University of Maryland (Weber, 1960, 1962). The essential principle on which the experi- ment is based is that any material system held together by elastic forces will oscillate in resonance with the wave; the main problem is then the detection of such oscillation and the elimination of the noise. Weber showed thata piezoelectric crystal connected to a low noise amplifier is an excellent receiver for wavelengths smaller than its side. At frequencies of about 103 sec~! it is hoped to detect a power flux ~ 10-4 ergs/cm2sec. If the wave amplitude is peaked around a given propagation vector, the rotation of the earth will modulate the output according to the sidereal day. The theoretical interpretation of these experiments is still open, until more is known about the mechanism of production and the properties (e.g., isotropy) of the gravitational waves; for example, does radiation of such The Classical Tests of General Relativity 21 high frequencies arise from non-linear breaking of “cosmological” oscillations or are they produced by celestial matter? It would be very interesting to investigate them also at astronomical wavelengths; a detailed critical survey of the observed planetary motions from this point of view has not been made to the authors’ knowledge. If gravitational waves could be detected, intensity measurements could be used to check whether their velocity of propagation is indeed the same as that of light (Braginskii et al., 1960. This paper also gives references to other work on the subject). 1-3. The Classical Tests of General Relativity This section contains a summary of the present status of the three classical tests of general relativity: the gravitational red shift, the deflection of light by the sun, and the precession of the perihelion of mercury. Only the precession of the perihelion is clearly a test of the full formalism of general relativity (Schiff, 1960; Bondi, 1960). There are some other effects, associated with rotating bodies (Lense and Thirring, 1918), which could in principle be used to test general relativity, but all such effects are too small to be observed with present-day techniques; for example, the rotation of the earth induces a gyroscope precession relative to the distant stars of ~ 10-9 radians per day. The astronomical effects of rotating bodies are also too small to observe (De Sitter, 1916). The present state of the three classical tests has also been surveyed recently by Trumpler (1956), Freundlich (1955), Ginsburg (1957), and Levy (1959). 1-3.1. The frequency shift THe THEORETICAL PREDICTION. A general formula for the frequency shift as predicted by general relativity has been given by Schrédinger (1956, p. 49). In Fig. 1-8 a source S with four-velocity v, emits light of frequency v,, as measured in the rest frame of the source; the observer O with four-velocity v, receives this light and measures the frequency v, in his rest system. The ratio of the frequencies is then given by the invariant Yo _ Swto'P'le (13.1) Ys Buibs!P'ls where p is the null vector transferred parallelly along the null ray r. In this general formula no distinction is made between the ordinary Doppler effect and the gravitational frequency shift; according to the principle of equivalence they are two different names for the same phenomenon. In fact, it has been pointed out (Einstein, 1911; Schiff, 1960) that the 22 Experiments on Gravitation prediction of the shift does not require the full machinery of general relativity, but follows from the principle of equivalence. Experiments aimed at testing (1-3.1) are therefore interesting mainly as a check on the principle of equivalence. Fig. 1-8. The general formula for the frequency shift: vo _ Suwo'D" lo vs Busts |s In a static gravitational field, with respect to which source and observer are at rest, the frequency shift is most conveniently expressed in a stationary coordinate system, defined by the condition that the g,, be x°-independent, and gio = 0(@@ = 1,2,3). In such a system (1-3.1) takes the simple form ¥ _ |Bo0o(S) _ 7S gon(0) 032) It is customary to state any frequency shift dv = v, — », in the spectral lines of stars in terms of the “equivalent velocity” v = c(dv/v,) which the source would have if the entire shift were a Doppler shift. In a weak gravitational field, with no relative motion between source and observer, one obtains from (1-3.2) 1 v=2b.- $0 Here ¢ is the Newtonian gravitational potential. The smallness of the effect makes it necessary to use atomic or nuclear frequency standards to observe it. The Classical Tests of General Relativity 23 We shall now briefly review the existing experimental evidence con- cerning the red shift in the spectra of the sun and of white dwarfs, and finally the terrestrial confirmation by means of the Méssbauer effect. Tue ReD SHIFT IN THE SPECTRUM OF THE SUN. The “‘equivalent velocity” for the sun (observed from an inertial frame) is 0.636 km/sec. The correction for the earth’s potential amounts to 0.029% of this value and is negligible for the accuracy of present-day data. The Doppler shifts due to orbital motion of the earth and sun are of comparable magnitude and are subtracted out in the data given below. The predicted value 54/A = 2.12 x 10-6 can be compared with typical values of the “line strength” W/A to obtain a comparison between the difference in wavelengths 8A that has to be measured and the width W of the Fraunhofer lines which are used in the comparison. M. G. Adam (1958) lists values of W/A ranging from 5 x 10-6 to 300 x 10-6; thus the red shift ranges typically from 1/10 line width for medium strong lines to 1/100 line width for strong lines. A shift of this magnitude can easily be measured if one assumes symmetry of the line, or if one assumes that the narrow “cores” of many very strong and wide solar absorption lines occur at the line center. Since these methods are used in actual observations, it is essential that only symmetric lines be selected. In actual spectra, tails of neighboring lines often “blend” with the line under observation, causing asymmetries, In the sun’s spectrum, in particular, there are other factors causing asymmetry, to be discussed below. They are associated with various processes in the sun’s atmosphere and are often difficult to correct. Experimental information on the detailed shape of the lines suitable for red shift measurements would help us understand these processes, but are lacking to date.* The main sources of observational material on the solar red shift are summarized in Table 1-3. It is found that the wavelength shift 5A varies not only with wavelength A, but also with the strength W/A of the line, and with the position of the source on the solar disk. The largest number of measurements have been made on medium strong lines (see Fig. 1-9). The different observations agree qualitatively in showing a slight red shift m the center region, comprising as much as 90% of the distance to the limb; the red shift then suddenly increases as the limb is approached, reaching the relativistic value, or higher, at the limb, The center red shift increases with increasing line strength; very strong lines show a center shift exceeding the relativistic value, and this shift apparently decreases, but remains above the relativistic value, at the limb (Evershed, 1931). It is clear that any comparison between the observational material and the * An experiment of this type is now in progress in Professor Dicke’s group at Princeton. 24 Experiments on Gravitation km/sec 01; ° 4 oL—i__t el {| Jj | {ij 10 09 08 07 06 05 04 03 02 01 0 Center Limb —~cos 8-> ‘822: Range of typical observations by Evershed (1931) + Adam (1948) © Freundich, Brunn, Brick (1930) © Adam (1989) Fig. 1-9. The main experimental material on the center-limb shift for the sun. Obser- vations of four observers are plotted vs. cos 0, following Adam (1959), in order to spread out the interesting region near the limb. Absolute measurements (Evershed, 1948, and Adam, 1948) should, according to the radial current theory, reach the telativistic value at or near the limb. The difference between Evershed’s and Adam’s data is probably due to the relatively small size of the solar image used by Evershed, preventing him from looking at a very narrow range of 6 near the limb. Relative measurements (Freundlich et al., 1959, and Adam, 1959) are plotted with the center value normalized to zero. They should therefore all differ by the amount of the center shift from the corresponding absolute measurement. Under this assumption the radial current theory would predict a limb value somewhere in the hatched region. Adam’s very detailed 1959 measurements seem to exceed this value at the limb. The considerable scatter in her points is apparently due to local velocity fields. Only the data for the east limb are plotted because the west limb data were affected considerably by light scattering. The Classical Tests of General Relativity 25 red shift predicted by theory presupposes an understanding of this “limb effect.” TABLE 1-3. Main Sources of Observational Material on the Solar Red Shift Reference Observations Burns, Meggers, Kiess Absolute measurements of 158 lines of wide range of (1927) strengths, for the integrated disk. St. John (1928) Absolute measurement of 1957 lines at the solar center, 133 lines at the limb, of wide range of strengths; only ~4°% of these were free from blends. Freundlich et al. (1930) | Measurements, relative to center, of 9 medium strong lines at 6 points along each of 12 radii. Evershed (1931, 1936) Absolute measurement at center and limb of num- (summary of observa- erous weak, strong, and very strong lines; also very tions 1912, 1936) strong Ca lines in prominences. Adam (1948, 1955) Absolute measurements of 14 medium strong lines at 13 points across the polar diameter. Adam (1958) Absolute measurement of 34 lines of wide range of strengths at the center only. Adam (1959) Detailed relative measurements of 3 medium strong lines at 25 points near the equatorial east and west limb. Herzberg (1957, 1960) Measurements, relative to center, of 6 infrared lines at 16 points on the limb. Schroter (1959) @ Absolute measurements of 36 Fel lines as a function of line strength, (b) Limb effect on 30 medium strong lines. (c) Limb effect on 3 strong lines. No limb effect found in NaD. The mechanism considered today most likely to account for the limb effect was originated by St. John (1928) and elaborated by M. G. Adam (1948) and Schréter (1957). These authors point out that radial currents in the solar atmosphete would give rise to a Doppler shift of Fraunhofer lines which would depend on both line strength and disk position. Exist- ence of such radial currents in the solar atmosphere has been made extremely plausible by observations on solar granulation (Schwarzschild, 1959) (Fig. 1-10). The granules are interpreted as masses of hot gas, from 300 to 1,800 km in diameter, which move upward, cool, and return down- ward in the darker “intergranular” spaces. An observed spectral line is a combination of a Doppler blue-shifted line, formed in the brighter granules, 26 Experiments on Gravitation Fig. 1-10. Photograph showing the solar granulation and a sun spot (dark area). This picture was taken from a balloon flown at high altitude as part of Project Strato- scope of Princeton University, sponsored by the Office of Naval Research and the National Science Foundation. and a Doppler red-shifted line formed in the darker intergranules. A detailed model of the two streams was developed by Schréter (1957). He finds that the streaming of the gases shifts the wings, but not the center of the line contour. The result is an asymmetric line, whose average position is shifted. The model can account qualitatively for the limb effect; at the limb, where the radial motion produces no Doppler The Classical Tests of General Relativity 27 shift, it predicts the relativity value for the red shift. Since the strength of a Fraunhofer line is a measure of the average depth at which it is formed—the weaker lines being formed in lower layers—the model also predicts the dependence of wavelength shift on line strength. However, in its present form this model does not lead to quantitative agreement with recent measurements. Schwarzschild (1959) has investi- gated solar granulation with improved resolution by using high-altitude balloon observation, and found a much smaller contrast between granula and intergranula (about 1:1.05) than the value 1:1.45 used by Schréter. In addition, recent careful work on the center-limb shift concentrating on the region near the limb (Adam, 1959; see Fig. 1-9) makes it plausible that the equivalent velocity near the limb actually exceeds the value predicted by relativity. This excess seems quite inexplicable on the radial current theory. Both the relativistic line-shift and that due to any Doppler shift depend linearly on A. This linearity could in principle be tested even if the limb effect is not satisfactorily understood, given only that it is produced by Doppler shifts. M. G. Adam (1958) has examined the very extensive AO-NBS observations for this effect; however, the points scatter far too much to allow any reliable conclusion concerning the linearity. In summary, to date we are not in a position to explain the details of the deviations of the solar red shift from the theoretically predicted value, nor can we determine whether the functional form of the relationship (8A proportional to A) is correct. At best, the relativistic effect makes the solar red shift somewhat less puzzling than it would be if we had no theoretical prediction of this kind.* THE RED SHIFT IN THE SPECTRUM OF WHITE Dwarrs. The red shift in the spectrum of a star is proportional to the gravitational potential at its surface, i.e., to the ratio M/R. Therefore white dwarfs with their small radii (1/10 to 1/100 solar radius) and appreciable masses (about solar mass) are the most suitable stars for observations of the relativistic red shift. White dwarfs are recognized by their low luminosity and high surface temperature. About 400 such objects are known today. Only very few of these are suitable for a red shift observation, because the parameters * No satisfactory alternative to the radial current theory has been proposed to explain the center-limb shift. For an attempt to explain the effect as a pressure shift, see Spitzer (1950); for objections to this, see Schréter (1956). Freundlich (1954) has proposed a mechanism explaining the red shift without the principle of equivalence, as aan interaction between photons and radiation fields. Schrédinger (1955), has shown that such a theory would predict a considerable broadening of the spectral lines from shstant galaxies, if it accounted correctly for the cosmological red shift. Freundlich’s \beory also cannot explain the dependence of the solar red shift on line strength (see Schréter, 1956) and gives an incorrect value for the limb effect (Adam, 1959). 28 Experiments on Gravitation Mand R have to be known with reasonable accuracy. Themass M can be determined if the white dwarf forms part of a suitable double or multiple star system. One measures the distance from the earth (by parallax), the angular separation of the components, and the period of revolution, T. The law of gravitation then gives the sum of the masses M; + M, = 4x2 (relative distance)}/(GT?). The separate masses are determined either by using more detailed information about the motion of the components (as for 40 Eridani B), or by computing the mass of the normal component theoretically (as for Sirius A). To determine the radius of a star one first finds the surface temperature from the spectral distribution, and the absolute luminosity. The surface area (and hence the radius) is then given by the ratio of luminosity to radiation emitted per unit area at the surface temperature. These measurements are somewhat problematical for white dwarfs, much of whose radiation is given off in the ultraviolet and absorbed by the earth’s atmosphere, The spectra of white dwarfs usually show mainly strong hydrogen lines; unfortunately these lines are very broad and diffuse, due to rotational Doppler shifts and interatomic Stark effect; again one must rely on the assumed symmetry of the lines, or on narrow cores, to make measurements. Meaningful measurements of the red shift have been possible on only the two brightest white dwarfs, the companion of Sirius (Sirius B), and one of the components of a star system in the constellation Eridanus (40 Eridani B). The angular separation of Sirius B from Sirius A has been too small during the last 30 years to make any observations. Even in the 1920's, when measurements could be made, scattering of light from Sirius A into the spectrum of the much fainter companion could not be avoided. This effect tends to falsify both the computed radius and the line shift. The theory of degenerate matter in white dwarfs (Chandrasekhar, 1939; Kuiper, 1941; Gamow, 1949) gives a unique value for the radius of a white dwarf once its mass is known. Since the mass of Sirius B nearly equals that of the sun, its radius should be 1/100 solar radius, so that the pre- dicted red shift amounts to 60 km/sec. Observations were made by Adams (1925) and Moore (1928) on the shift of H, and H, in Sirius B under varying conditions of atmospheric scattering. The shift was measured relative to the same lines in Sirius A, so the Doppler shift correc- tions were necessary only for the accurately known relative motion. The results range from 9 km/sec to 31 km/sec and seem to be correlated with the intensity of the scattered light of Sirius A (Schréter, 1956), indicating that higher values might be obtained if the scattering had been reduced. Thus to date the results on Sirius B are not in disagreement with theory, but inconclusive. The Classical Tests of General Relativity 29 40 Eridani B is sufficiently separated from its companions that its spectrum is not falsified by scattered light. For this reason, however, the motion of the system is so slow that one does not know the proper orbit of the components. The values of the mass and the radius of the white dwarf are therefore rather uncertain. Since the hydrogen lines in 40 Eridani A suffer from blends, it is here not possible to measure the wave- length difference between the white dwarf and the normal star, which would avoid a Doppler shift correction. D, M. Popper (1954) made absolute measurements of the line shift and corrected for the velocity of the system by using values for the velocity of 40 Eridani A taken from the tables of standard velocity stars published by the Lick Observatory. He finds a net red shift on 21 + 3 km/sec, as compared with the predicted value of 17 + 3 km/sec, and estimates that the ratio of observed to pre- dicted value lies well within } to 2 if allowance is made for all uncertainties. This measurement represents, to date, the best astronomical con- firmation of the theoretical predictions for the gravitational red shift. TERRESTRIAL EXPERIMENTS. The potential differences which are available on the earth for red shift experiments are so small that only methods of extremely high accuracy will succeed in detecting the red shift. Atomic clocks can reach an accuracy of one part in 10!!, and have been proposed for red shift experiments (Moller, 1957). The possibility of using an atomic clock on a high mountain has been superseded by the experiments using nuclear transitions to be discussed below; however, satellite experiments, using atomic clocks or their nuclear analogues, may in the end give us the most accurate check on the predicted red shift (Singer, 1956). Nuclear gamma transitions represent a much more accurate frequency standard than atomic transitions, because the emitted frequency is much higher, and the natural line width can be appreciably less. In ordinary gamma-emission, the lines are considerably Doppler broadened, due to thermal motion; in addition, the recoil nucleus takes up some of the transition energy, so that the gamma ray has the wrong frequency to be resonantly absorbed by another nucleus of the same isotope, unless the recoil shift is compensated by an external Doppler shift obtained by moving the source. Recent techniques due to R. L, Méssbauer (1958) have eliminated both. the above difficulties by imbedding the emitting and absorbing nuclei in a crystal lattice. The lattice acts much like a buffer gas for the analo- gous atomic experiment: by restricting the motion of the emitter and absorber to regions small compared to a wavelength of the emitted radiation it prevents Doppler broadening, so that the emitted line width is close to the natural line width (Pound finds their ratio, for Fe57, to 30 Experiments on Gravitation be ~1.7). Also, the recoil momentum is taken up by a volume of crystal of linear extent ~(velocity of sound in crystal) x (lifetime of excited state), so that the energy loss is reduced by ~(number of atoms in this volume)~!, and resonance absorption becomes possible. In actual crystals some residual frequency shifts are observed (Pound and Rebka, 19605), which vary both from crystal to crystal and with temperature. Pound finds small shifts even when the source and absorber are imbedded in crystals of the same material. These effects are not yet understood in detail. In addition, a recoil-free gamma ray line has several satellites due to hyperfine structure interaction with internal effective magnetic fields. It is, however, not essential for red-shift experi- ments that these effects be understood in detail: any method of generating and observing a reproducible sharp frequency can be used. Pound and Rebka (1960a) and, less conclusively, Cranshaw et al. (1960), have been able to measure a recoilless gamma ray line in Fe57 with a fractional width of the order of 10-12, and used this in a gravitational red shift experiment. Over the 74 ft of height difference used by Pound, the theory predicts the value Av/v = 2.5 x 10-15 for the fractional red shift, To measure to this accuracy, the source was moved back and forth at low frequency, in order to modulate the source frequency by Doppler shifts; the gravitational red shift then results in an asymmetry in the absorbed frequency, which was detected synchronously, As mentioned above, other shifts appeared to be present than that due to temperature differences and differences in the gravitational potential. The extraneous shifts amounted to 18 x 10-15 as compared to 2.5 x 10715 for the fractional gravitational shift, The measured value for the latter was obtained from the difference between the frequency shifts observed with falling and rising gamma rays, and amounts to 1.05 + 0.10 of the theoretically predicted value. Thus the terrestrial experiments provide to date the most accurate verification of the gravitational shift. 1-3.2. The bending of light rays THE THEORETICAL PREDICTION. In the theory of general relativity, the bending of light rays by a static gravitational field can be thought of as due to an effective index of refraction (Pauli, 1958, p. 154).* In particular, in a spherically symmetrical field where the metric is dst = (FSFE 2 dee — 1 + 3U yt ae? (1-33) +40, [u- Be dee ac tartar paws ys al * The behavior of electromagnetic radiation in a gravitational field has been exhaust- ively treated by Plebanski (1960). The Classical Tests of General Relativity 31 light travels as if space had an index of refraction _ (+ 4Uy n=ag <1 tw (1-3.4) om. a Fig. 1-11, The gravitational bending of light rays. In formulas (1-3.3) and (1-3.4) Gis the gravitational constant (6.67 x 10-8 cgs units), c the velocity of light and M the mass of the gravitating body. Notice that the coordinate r has no direct operational meaning; but where U < l,can be considered approximately equal to the ordinary distance, measured according to Euclidean geometry. The bending of light (Fig. 1-11) can be easily calculated by classical optics using (1-3.4) and turns out to be ox ay = 20M re (1-3.5) Here r is the distance of closest approach to the origin. The deflection angle @ (for rays grazing the limb) has the value 1775 for the sun and and 0°02 for Jupiter. 8 can be defined in an invariant way as follows, Consider in a universe otherwise flat the two cases in which the sun is present or not. The light coming from a distant star determines at the world-line / of the observer two null vectors p and p’; if A is the unit vector tangent to /, we have @-p’) cos@=1 @nPD (1-3.6) Since space-time is practically flat for an observer far from the sun, the ungle @ = 4GM/rc? calculated by classical optics using (1-3.4) is just the angle given by the invariant (1-3.6). 1t has been claimed by Lenz (Sommerfeld, 1948, p. 314; Schiff, 1960) that the linearized Schwarzschild line element (1-3.3) and hence the deflection of light rays can be derived solely from the principle of equi- valence and Newtonian mechanics, without any knowledge of the field equations. However, there is disagreement on this point. In our 32 Experiments on Gravitation. opinion, Lenz’s argument is only an interpretation of the line element itself, and hinges upon a particular (and to a certain extent arbitrary) transformation to an accelerated frame. SURVEY OF THE EXPERIMENTAL Work. The experiment consists in photographing the star field around the eclipsed sun and comparing the photograph so obtained with one of the same star field taken about six months later, when the sun is no longer in it. Measurements of the light deflection have been made on many eclipse expeditions; among the more trouble-free measurements are those by Campbell and Trumpler from the Lick Observatory in 1922 (Campbell and Trumpler, 1923, 1928) and by Freundlich, Kliiber, and Brunn from Potsdam in 1929 (Freundlich et al., 1931a and 6). The fundamental difficulty in the experiment is that the scales of the two photographs to be compared are unavoidably different and it is very difficult to obtain an independent measure of the scale factor. Loti) Lalit 0 05 (ie 0 05 1° Scale of chart Scale of star displacements Fig. 1-12. Star displacements observed by Campbell and Trumpler in the eclipse of 1922, These data were taken with a camera of 15-ft local length and aperture 5 in. Plate scale factors were determined in such a way that the average radial displacement of the stars outside the dotted circle of radius 2° was zero. The inner dotted lines represent the sun’s corona. The figure is taken from Campbell and Trumpler (1923). The Classical Tests of General Relativity 33 To show some typical experimental data, we reproduce in Fig. 1-12 a plot given by Campbell and Trumpler which shows the shifts they observed in the 1922 eclipse. This eclipse was a particularly favorable one in that there were many fairly bright stars and these were uniformly distributed around the sun, Corrections have been made for proper motion, parallax, differential refraction, and aberration. In plotting this figure, Campbell and Trumpler fixed the plate scale factor so that the stars outside the dotted circle of radius 2° had no radial shift on the average. It appears that though there is considerable scatter, stars near the sun do tend to be shifted outwards. Each vector in Fig. 1-12 actually represents an average of several pairs of plates. Typical data showing the scatter to be expected in different measurements on a single star are exhibited in Fig. 1-13. These shifts fluctuate considerably. The scatter might be compared with the typical size of a star image: this is determined by the “seeing” (i.e., unsteadiness in the image due to turbulence in the earth’s atmosphere) and, though very variable, amounts to a few seconds of arc. If the observed scatter were due to random errors only, the radial shift obtained by averaging would be fairly reliable, even though small compared to the scatter. However, such astronomical measurements are prone to systematic errors and these are almost certainly present. Among these the following have been suggested: 1. Refraction effects in the earth’s atmosphere due to a temperature drop in the moon’s shadow. 2. Refraction of light in the sun’s corona (this would lead to a bending over and above any relativistic effect). 3. Displacement of star images due to a gradient in the background blackening of the negative. This would tend to move stars in toward the eclipsed sun. 4. Distortions introduced by the shrinking of the emulsion while it was drying. Those parts of the emulsion near the black image of the sun’s corona can, under suitable conditions, dry faster and shrink more than the rest of the emulsion (Ross, 1920; Silberstein, 1920); this gives an apparent radial displacement toward the sun, 5, Distortions in the optical system due to temperature changes occurring during the eclipse. Point 1 has been extensively discussed by English workers at about the time of the 1919 eclipse (Anderson, 1919; Eddington, 1919; Crommelin, 1919; Dines, 1919; Richardson, 1919; Schuster, 1919; Freundlich, 1920; Poor, 1923; Varnum, 1924); the effect seems to be negligible. Early estimates of point 2 were indecisive (Lindemann, 1918; Newall, “(ig6 1) wang pur ‘ioqnyy ‘Yyorpunsry Wo ose IEP ay] “sorenbs seo] Aq pourergo sem ejoqiedAy poop pur poysep oup oprym ‘AIAnETAL [VIOUS Jo UONIpord oy} SJUdsOIdOL SUT] POYsep YL “(AEIS YOR IO} PSAIOSqO S3j!Ys INOJ a1OM asoy} ‘9Sed SIU} U1) SaeId SNOLIEA UO seIS B JO} PIAJOsgo IJTYS [eIPeL oY} Ul JONES OUP sHgnyXe auNSy SIL “EI-T ‘By ped 4ej0g 8 SL L go 9 gs s st v Ge £ Se é ST 1 T T T 7 T T T T T T T T T Experiments on Gravitation aie Jo spuoges ul Bupueg 0% r POP 4d 0 094 Weld 7 }— 06 4 ald o Obs eld + + 1 ' ‘ ' 1 ' ' ' 1 ' ' ' ce The Classical Tests of General Relativity 35 1919, 1920; Crommelin, 1920; Anderson, 1923). Rough estimates using modern data on the electron density in the sun’s corona (van de Hulst, 1953) indicate that the index of refraction at optical frequencies is too small to be significant. It is interesting to note that this is no longer true at radio frequencies (Paris Symposium on Radio Astronomy, 1958), so that experiments on the gravitational bending of light using radio stars do not seem feasible. Point 3 is quite unimportant (Bottlinger, 1920; Wolf, 1920). Although distortions introduced during drying of emulsions (point 4) can be large under suitable conditions, these conditions do not seem to have held in the processing of plates taken during the 1922 and 1929 eclipses, Both eclipse and comparison (night) plates for the 1922 eclipse carried images of a check field of stars (about 90° from the sun) in addition to images of the eclipse field of stars. The stars of this check field did not show strong systematic radial shifts, although some weak trends may have been present. Distortion of lenses and mirrors due to temperature changes during the eclipse (point 5) has definitely been observed and has led to poor star images (Dyson, Eddington, and Davidson, 1920; Biesbroeck, 1950). It is necessary to shield optical and mechanical parts from direct sunlight. The main uncertainty in measurements of the bending of light remains the determination of the plate scale factors. The scales of the two photographs (eclipse and comparison plates) are different because of temperature changes and uncertainties in mechanical positioning of the plates relative to the objective. Campbell and Trumpler determined the change in radial scale factor by demanding that distant stars should show no radial shift on the average, and computed a value for the bending 0 at the limb of the sun by plotting radial shift against distance r from the sun and fitting this by a curve er + 6/r, « and 6 being two parameters to be determined by a least squares analysis. (The linear term ar is necessary because if there really is a gravitational shift the stars far from the sun should exhibit a small but non-zero radial displacement.) The open circles in Fig. 1-14 show essentially* their experimental plot of shift versus distance, each point on the curve representing the average of many measurement on many stars. They found @ = 175 + 0.09, which is to be compared with the predicted value 1°75. The error quoted represents the internal consistency of the data and is not a measure of the reliability. ‘The results from particular pairs of plates vary from 2738 to 1°01. As mentioned above, the residuals for stars of the check field (shown in * The open circles in Fig. 1-14 actually show the data of Campbell and Trumpler as slightly modified by Freundlich et al., but the qualitative features have been unchanged by this modification. Experiments on Gravitation “(16D wunsg pur “soqn[y “yoypunosy wWosy uoyxer st oindy sy, ‘eIep Weps}og oY} 01 14 sosenbs yseo] eB Aq pourejqo sem vfoqredky payop pue paysep oy} spyM “AWANIOL pegUaT jo uonopaid oti st efoqiedéy poysep oy, “wunig pue “SoqnTy ‘Yorpunary Aq payypow useq Susaey osdijos 776 3} 10J SyNsar AIOVAISAO ANT Ap “Sasdit9o (sof! PHOS) GZ6T Pue (Sap2119 usd) ZZE_ a BUMP partosqo sjuswoe|ds|p eis [IPEA “bI-T BA ped 4ej0§ 8 Z 9 TTT T T T T T TT T T T T T T T T T T T T vou = . ZOo- pr] 1 wepsjog « PTT TT TTT TTT TTT The Classical Tests of General Relativity 37 Fig. 1-15) were not quite randomly distributed, but showed some weak systematic trends. If this were taken to imply a small correction to the shifts of stars in the eclipse field, the value of @ would become 2.05 sec of arc. However, it is not clear that such a correction should be applied because the stars of the check field were exposed about six hours after those of the eclipse field, and the plate scale factors might well have been different. 2 +0°.3- Trumpler: .- . +0".2b . . 4 +071, . . . 070 : ie =O" ~072b . . : -0°3 pore tate ttt | TIT TTT TTT TTT r Campbell: . +074 . . +0".3- 4 +0"2+ x oo +071 oro -0"1 -0"2 -0"3 —0".4 Radial residual 0°.5 1°.0 15 2°.0 2°75 3°.0 Distance from plate center Fig. 1-15. The radial residuals of check stars observed on plates taken with the 15-ft camera during the 1922 eclipse. Asterisks are group means and the dotted lines show the corrections derived from these and applied to stars of the eclipse field. This figure is from Campbell and Trumpler (1923). In the Potsdam expedition of 1929 (Freundlich et al., 1931; Freundlich, 1930) an attempt was made to measure the plate scale factors directly. Immediately after the eclipse, the image of a grid was projected both onto the photographic plate recording the stars of the eclipse field, and on another plate in a second camera which was focussed on the stars of a check field far from the sun, both cameras collecting light from the same coelostat. (For a description of the apparatus, see Kliiber, 1932.) The same procedure was followed about six months later when the comparison plates were taken, so that even if the grid had changed its dimensions in six months’ time, the scale factors on the comparison plates taken at night could still be determined in terms of the scale factors for the eclipse plates. 38 Experiments on Gravitation Unfortunately, the image of the grid on the plates showing the stars of the check field at night was rather distorted, presumably due to strains in the objective of the second camera. In their analysis, Freundlich, Kliiber, and Brunn assumed that the grid had not changed its dimensions. The solid circles in Fig. 1-14 show the radial shifts they obtained, each circle representing an average of many measurements. A least squares fit of their data by a curve 6/r gives a value for @ of 272 + 0.10. There were fewer stars near the sun during the 1929 eclipse than during the eclipse of 1922, and they were not uniformly distributed around the sun. There has been some controversy about the best way to treat the data obtained during the eclipse expeditions of 1922 and 1929, Freundlich et al. have criticized the way Campbell and Trumpler have carried out their least squares analysis; if one were to accept the prediction of general relativity and ask merely what is the best value of the coefficient 6, then the constant « in Campbell and Trumpler’s analysis should not be treated as a free parameter, to be determined by least squares, but should be calculated on the assumption that the radial shift is 6/r. Reanalyzing Campbell and Trumpler’s data in this way, and applying a correction for the systematic shifts in the stars of the check field, Freund- lich et al. obtained values for 6 of 2°1 to 2"3 in good agreement with their own result. The open circles in Fig. 1-14 show the Lick Observatory data, as corrected by Freundlich et al. On the other hand, Trumpler has pointed out (Trumpler, 1932, 1956) that the distortion in the image of the grid in the Potsdam experiment throws doubt on the scale factors used in the analysis of Freundlich et al. He reanalyzed their data using the same method as was used in his analysis of the Lick Observatory data and obtained @ = 1.75 + 0.13 sec of arc. In addition Ludendorff (1932) has pointed out systematic trends in the tangential components of the displacements observed in the 1929 Potsdam results, and suggests that the value of @ should be 1.9-2.0 sec of arc. There have also been other analyses of the data obtained in these experiments (Hopmann, 1923; Jackson, 1931; Danjon, 1932; Mikhailov, 1956), all yielding various values for the constant 6 in the range 1.75 to 2.20 sec of arc. The general theory of the observational errors in an eclipse measurement of the light deflection has been discussed by Freund- lich and Ledermann (1944). (See also Freundlich, 1951, and Trumpler, 1956.) The statistical analysis made by these authors has emphasized how important it is to secure an independent measure of the scale factor, instead of extracting it from the observed deflections themselves. SUMMARY OF EXPERIMENTAL RESULTS. There have been a number of determinations of the bending of light, in addition to those described above. In many of these, practical difficulties of one sort or another The Classical Tests of General Relativity arose, often being connected with optical distortions introduced by tem- perature changes, or to poor weather. These have also been discussed in a review article by the results to date. Kliiber (1960). TABLE 1-4, Measurements of the Bending of Light by the Sun 39 The following table summarizes Focal No. of Result, 6, length stars (seconds of Different Eclipse Ref. (em) (approx.)_rmin* are) Analysis Ref. 1919 1 675 7 2 1.98 + 0.12 2.0 to 2,2 4 2164014 2 2.06 13 495 $0.09 14 1 5 2 1.61 + 0,30 1922 3 533 92 210 1724011 2.2 4 2.14 +018 2 2.12 8 2.00 13 1.83 £0.20 14 3a 178 145 21 1824015 2.1 4 2.07 13 1922 5 163 14 2 1,18 to 2,35 1922 6 391 18 2 1.42 to 2,16 1929 4 850 17 15 2.244010 1.98 + 0.20 8 1.75 +013 7 2.06 13 1.96 +011 14 1936 9 600 25 2 2.71 + 0,26 2.70+040 14 1936 10 500 8 4 1.28 to 2.13 1947 it 609 51 3.3 2.0140.27 2.01 +027 15 2.20 +035 14 1952 12 609 10 2.1 1.704010 1.43 +016 14 References 1. Dyson, Eddington, and Davidson, 1920 8. Jackson, 1931 2. Hopmann, 1923 9. Mikhailov, 1940 3. Campbell and Trumpler, 1923 10. Matukuma et al., 1940 3a. Campbell and Trumpler, 1928 11. Beisbroeck, 1950 4, Freundlich, Kliiber, and Brunn, 1931 12, Beisbroeck, 1953 5. Dodwell‘and Davidson, 1924 13. Danjon, 1932 6, Chant and Young, 1924 14. Mikhailov, 1956 7, Trumpler, 1932 15. Kliiber, 1960 *rmin = distance of closest star from sun’s center, in solar radii. 40 Experiments on Gravitation In all of these experiments, a radial shift outward of about the right order of magnitude was observed, and the shift was less for more distant stars. The decrease does not contradict the 1/r law predicted by general relativity, but neither does it give much support to such a dependence on distance. In Fig. 1-14 a straight line would fit the data about as well as a hyperbola (see Mikhailov, 1959). Since in all of the observations to date, practically all the stars were more than two solar radii from the center of the sun, the value of the bending at the limb represents quite an extrapolation. From this point of view it is not surprising that relatively minor changes in the observed deflections can lead to a much greater change in the com- puted bending at the limb. Assuming that there is a radial shift and it is of the form 6/r, values of @ range from 1.75 sec of arc (the predicted value) to about 2.2 sec of arc. None of the experiments support the Newtonian value 1.75/2 = 0.875 sec of arc. The subject is still a live one (Mattig, 1956; Kliiber, 1955) and more experimental data can be expected. Under the conditions of an eclipse expedition it appears to be very difficult to get good results using the photographic technique and it may be that photoelectric methods for observing stars near the sun in the daytime will be applied. The use of red-sensitive plates has also been advocated to this purpose by Lindemann (1916). There has recently been a proposal to do the experiment from a satellite (Lillestrand, 1960). 1-3.3. The motion of the perihelion of a planet THE THEORETICAL PREDICTION. The theory of general relativity, although conceptually so different from Newtonian dynamics, in practice does not affect in any relevant way the predictions of celestial mechanics, except in one exceedingly minute effect, the anomalous advance of the perihelion of a planet.* Only due to the centennial accumulation of painstaking astronomical observations was this discrepancy, which puzzled the astronomers for a long time before Einstein’s proposal, established with any certainty; which makes it truly astonishing the fact that a purely theoretical and independent speculation was sufficient to understand it. . The motion of the perihelion is the only observable consequence of the field equations which is connected with their non-linear terms. Notice, * For a full account of the problem of motion in gencral relativity, see L. Infeld and J. Plebanski (1960). If one allows for negative gravitational masses, which is theoretic- ally allowed and even reasonable, the sun’s gravitational field would polarize all the planets, which hence would possess an additional dipole field. The secular perturba- tions in the lunar motion, for example, would place an upper limit to the percentage of negative mass present in the earth.

You might also like