he Kerr Spacetime Rotating Black Holes in era UMN eieaen David L. Wiltshire Matt Visser and TEUTHE KERR SPACETIME Rotating Black Holes in General Relativity Edited by DAVID L. WILTSHIRE University of Canterbury, Christchurch MATT VISSER Victoria University of Wellington SUSAN M. SCOTT Australian National University, Canberra CAMBRIDGE UNIVERSITY PRESSCAMBRIDGE UNIVERSITY PRESS ‘Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, Sao Paulo, Delhi ‘Cambridge University Press ‘The Edinburgh Building, Cambridge CB2 8RU, UK Published in the United States of America by Cambridge University Press, New York www.cambridge.org Information on this title: www.cambridge.org/9780521885126 © Cambridge University Press 2009 This publication is in copyright. Subject to statutory exception and to the provisions of relevant collective licensing agreements, ro reproduction of any part may take place without the written permission of Cambridge University Press. First published 2009 Printed in the United Kingdom at the University Press, Cambridge A catalogue record for this publication is available from the British Library Library of Congress Cataloguing in Publication dara ‘The Kerr spacetime ; rotating black holes in general relativity / edited by David L, Wiltshire, Matt Visser, I ‘Susan Scott, pe om. 1 |ncludes index. ISBN 978-0-521-88512-6 |, Kerr black holes. 2. Black holes (Astronomy) ~ Mathematical models. 3. Space and ti 2 models. 4, Kerr, R. P. (Roy P) I. Wiltshire, David L., 1956— I. Visser, Matt S IV. Title. QB843.B5SK47_ 2009 3 523.8'875-de22, 2008045980 : ISBN 978-0-521-88512-6 hardback 4 5 6 Tl 7 8 9 Cambridge University Press has no responsibility forthe persistence or accuracy of URLs for extemal or third-party internet websites referred to in this publication, and does not guarantee that any content on such 10 ‘websites is, of will remain, accurate or appropriate. Contents Contributors Foreword Preface List of Illustrations General relativity: Classical studies of the Kerr geometry The Kerr spacetime — a brief introduction Matt Visser The Kerr and Kerr-Schild metrics Roy P. Kerr Roy Kerr and twistor theory Roger Penrose Global and local problems solved by the Kerr metric Brandon Carter Four decades of black hole uniqueness theorems David C. Robinson Ray-traced visualisations in asymptotically flat spacetimes: the KerrNewman black hole Benjamin R. Lewis and Susan M. Scott Astrophysics: The ongoing observational revolution in our understanding of rotating black holes The ergosphere and dyadosphere of the Kerr black hole Remo Ruffini Supermassive black holes Fulvio Melia The X-ray spectra of accreting Kerr black holes Andrew C. Fabian and Giovanni Miniutti Cosmological flashes from rotating black holes Maurice H. P.M. van Putten page vii ix xi xiv 95 115 144 161 213 236 281vi Ii 2 Iv 13 Contents Quantum gravity: Rotating black holes at the theoretical frontiers Horizon constraints and black hole entropy Steve Carlip Higher dimensional generalizations of the Kerr black hole Gary T. Horowitz Appendices Gravitational field of a spinning ma: special metrics Roy P. Kerr Gravitational collapse and rotation Roy P. Kerr Index s an example of algebraically 311 332 347 349 354 Contributors Steve Carlip Physics Department, University of California at Davis, CA 95616, USA Brandon Carter Observatoire de Paris-Meudon, Place Jules Janssen, F-92195 Meudon, France Andrew C. Fabian Institute of Astronomy, University of Cambridge, Madingley Road, Cambridge CB3 OHA, UK Gary T. Horowitz Physics Department, University of California at Santa Barbara, CA 93107, USA Roy P. Kerr Department of Physics and Astronomy and Department of Mathematics and Statistics, University of Canterbury, Private Bag 4180, Christchurch 8140, New Zealand Benjamin R. Lewis Centre for Gravitational Physics, Department of Physics, Australian National University, Canberra ACT 0200, Australia Fulvio Melia Department of Physics and Steward Observatory, University of Arizona, Tucson, Arizona 85721, USA Giovanni Miniutti Institute of Astronomy, University of Cambridge, Madingley Road, Cambridge CB3 OHA, UK Roger Penrose Mathematical Institute, Oxford University, 24-29 St Giles St, Oxford OX1 3LB, UKviii Contributors Maurice H. P. M. van Putten Kavli Institute for Astrophysics and Space Research, Massachusetts Institute of Technology 77 Massachusetts Avenue, 37-287, Cambridge, MA 02139, USA David C. Robinson Mathematics Department, King’s College London, Strand, London WC2R 2LS, UK Remo Ruffini Dipartimento di Fisica, Universita di Roma “La Sapienza” , Piazzale Aldo Moro 5, 1-00185 Roma, Italy; and ICRANet, Piazzale della Repubblica 10, 1-65122 Pescara, Italy Susan M. Scott Centre for Gravitational Physics, Department of Physics, Australian National University, Canberra ACT 0200, Australia Matt Visser School of Mathematics, Statistics, and Computer Science, Victoria University of Wellington, PO Box 600, Wellington, New Zealand Foreword Ttis an amazing fact that there are roughly 10°° rotating black holes in the observable universe, and that the spacetime near each of them is, to a very good approximation, given by an exact solution of Einstein’s vacuum field equations discovered in 1963 by Roy Kerr. The Kerr spacetime is a defining feature of modem astrophysics. It is becoming increasingly important as the basis for understanding astrophysical Processes from core collapse supernovae which produce gamma-ray bursts while forming black holes to the supermassive black hole engines that power quasars. These processes are the most violent and energetic phenomena in the universe since the Big Bang, and the key to understanding them is the Kerr geometry, ‘Yet the man behind this remarkable solution has remained somewhat enigmatic. When one of the contributors to this book heard that a conference was to be held to celebrate Roy’s 70th birthday in 2004, he was suprised to learn that Roy was still alive! The idea of such a conference was suggested to me by another of my former teachers, Brian Wybourne, shortly before Brian passed away late in 2003. This book, which contains the invited lectures of the 2004 Kerr Fest, is the lasting result. As someone who took courses from Roy at the University of Canterbury, one personal anecdote is in order. Roy amused his students with his laid back style and laconic humour, while always impressing upon us how to see quickly to the point in complex calculations, to back-track and fix our mistakes. He was also disdainful of needless bureaucracy, and took it as a badge of honour that he was the last person to deliver his exam scripts to the university Registry for printing. “I can’t see why the Registry always want the exams so early,” he complained. “In Texas we just used to write the exams on the board. Look, don’t worry about the exam. I just rehash the old questions from past years.”x Foreword In one lecture that year, after filling a blackboard with calculations that weren't getting closer to the expected solution, Roy wrote “= mess “Look, at this point I think I’d go off to the pub”, he sighed. Standing back to look at the board for a minute, he enlightened us further. “Ah, I screwed up a factor of ct in the second line. ..I can’t be bothered doing all that again. Just take down a note: The stupid bastard screwed up the ct.” When we came to sit the exam for that course, some last minute rushing in its preparation became evident. The questions may just have been “rehashed”, but when one put the values in, the algebra rapidly became very formidable, for question after question. One student left halfway through in an obvious state of distress. While wrestling with one particular question, which two pages of algebra made me suspect was actually not analytically soluble, the value of Roy’s teaching became clear in a flash of inspiration. “.-. = mess”, I wrote, “I can’t be bothered solving this equation, but if I was to proceed here is the method I would use:. ..” This anecdote not only illustrates the fun we had as Roy’s students, but serves as auseful analogy for the state of understanding in 1963 about the problem of solving Einstein’s equations for the exterior field of a rotating mass. After some decades of work the consensus of the experts was that the problem equalled a “mess”, which most general relativists had given up on. The status of “black holes” as possible physical objects in the universe was also disputed. The term “black hole” itself had yet to be coined by Wheeler in 1967. Despite the work of Oppenheimer and Snyder on gravitational collapse in 1939, and the eventual understanding of the properties of the horizon in Schwarzschild’s 1916 solution, everyone knew that realistic bodies rotated. Thus back in 1963 it was still a possibility that the Schwarzschild horizon was a mathematical curiosity, which might not survive the perturbation of adding angular momentum. Roy’s solution changed all that. He established black holes as Possible entities in the physical universe, and recent decades of many astrophysical observations have confirmed that nature has made good use of them. Roy’s brilliance in achieving what nobody had for decades, by pen and paper before the days of algebraic computing, was really a consequence of his insights into dealing with “messes”. He understood deeply the importance of simplifying symmetry principles, of differential and integral constraints, and how to quickly Sort out the relevant degrees of freedom from the irrelevant ones. The Kerr solution is a monumental legacy to Roy’s incisive insight. David Wiltshire Preface The book is organized into three parts. * General relativity. Wherein the contributors discuss the classical physics of rotating black holes. There is an orientational overview of the Kerr spacetime, followed by Roy Kerr’s own account of its discovery, twistorial applications, constants of the motion, the uniqueness theorems and visualization of the Kerr geometry. * Astrophysics. Observational evidence for rotating black holes, supermassive black holes, merger and collapse. * Quantum gravity. Theoretical frontiers, black hole entropy and quantum physics, higher dimensional black strings and black rings. The contributions encompass by and large the invited lectures of the Kerr Fest: a Symposium on Black Holes in Astrophysics, General Relativity and Quantum Gravity, held at Roy Kerr’s home institution, the University of Canterbury, in Christchurch, New Zealand, in August 2004 to celebrate Roy’s 70th birthday a few months beforehand on 16 May. As one of the original speakers, Zoltan Perjés, sadly passed away a few months after the Symposium, one of Roy’s former colleagt who was unable to attend the Symposium, Roger Penrose, kindly agreed to submit an article in his place. As appendices, we also reprint Roy Kerr’s renowned Physical Review Letter of 1963, at less than two pages, stunning in its impact and brevity, along with his con- ference proceeding article presented at the First Texas Symposium on Relativistic Astrophysics in December 1963, in which the black hole property of the solution was first described. There should be something in this book for almost everyone — there are relatively few books one can tum to to get significant coverage of the Kerr spacetime, and most of the relevant material is scattered around the scientific literature in small bits and pieces. We hope that this book will serve as a coherent starting point for xii Preface those interested in more technical details, and as a broad overview for those who are interested in the current state of play. David Wiltshire Christchurch Matt Visser Wellington Susan Scott Canberra January 2008 Roy Patrick Kerr, 2007.epg TAL 7A2 TAZ TA4 ge) 7.16 Hee: 718 Illustrations Rotating Kerr black hole. Rotating Kerr black hole: Outer ergosurface and outer horizon. Polar slice through a Kerr black hole. Ivor Robinson and Andrzej Trautman. Ezra T. Newman. Conformal projection of {r, r} section through a Kerr black hole. Sketch of polar {r, 8) section through a Kerr black hole. Multiple coordinate charts on the Schwarzschild spacetime. Null geodesics in Kerr spacetime as visualized by @Rworkbench. Estimating the origin of a photon. First-order effect of frame dragging. ‘Asymmetry of geodesic deflection in the Kerr spacetime. Schwarzschild ray-tracing. Kerr-Newman ray-tracing. Cloud of particles corotating about an extremal spinning black hole. Albert Einstein, Hideki Yukawa and John Archibald Wheeler. Brandon Carter. —Potential in the equatorial plane of an extreme Kerr black hole. Nuclear versus gravitational binding energy. Extraterrestrial civilization as idealized by Roger Penrose. ‘The ergosphere as sketched by Ruffini & Wheeler. Particle decay process in the field of a black hole. Annotated first page of a preprint by Floyd and Penrose. D. Christodoulou’s Ph.D. defense. Introducing the black hole. Riccardo Giacconi and Luigi Broglio; the Uhuru satellite. Cygnus-X1. Magnetic field lines in the equatorial plane of a Kerr black hole. Lines of currents in the magnetosphere. Grayscale representation of the image of Cygnus A at 5 GHz. Location of the black hole Cygnus-X1. A model of how the black hole created the bubble. page 23 33 41 42 108 109 145 146 148 ces 154 155 156 162 164 166 167 168 170 ler 174 175 176 177 181 183 184 184 185 186, 186 719 = 723 7.22 7.23 7.24 1.25 7.26 727 7.28 gee) 7.30 731 732 8.1 8.2 83 84 85 8.6 87 8.8 8.9 8.10 8.11 8.12 Loe 8.14 8.15 91 | 7 94 95 9.6 ce 98 99 9.10 ol 912 9.13 9.14 9.15 9.16 9.17 List of illustrations Zel'dovich after donating his book to Pope John Paul II. Vela 5A and SB satellites; a typical event. ‘The CGRO satellite; observed GRBs in galactic coordinates. Some GRB light curves observed by BATSE. Energy fluence-averaged hardness ratio for short and long GRBs. The Italian-Dutch satellite BeppoSAX. “Dyado-torus” for an extreme Kerr-Newman black hole. Expansion of the PEMB pulse. The GRB afterglow phase. Comparison between approximate and exact EQTSs. GRB 991216 within our theoretical framework. Energy radiated in the P-GRB and in the afterglow. TEST (Traction of Events in Space and Time) by Attilio Pierelli. The ICRA and ICRANet logos. ‘The quasar 3C 273, HST image of the host galaxy containing the quasar QSO 1229+204. Collision between two galaxies ~ unraveling of the spiral disks. The Chandra Deep Field North X-ray image. Composite image of Centaurus A. VLA image of the powerful central engine in the nucleus of Cygnus A. Black-hole-powered jet of plasma streaming from the center of M87. Stars surrounding the supermassive black hole at the galactic center. The butterfly-shaped galaxy NGC 6240. The spiral galaxy NGC 2613. Masses of supermassive black holes versus stellar velocity dispersion. Black hole accretion rate versus redshift. History of the accretion rate density. Spectra for flux density and polarization from Sagittarius A*. Polarization maps of Sagittarius A*. X-ray and reflection spectra of an accreting black hole. Photoionized gas near an AGN; typical spectrum of a Seyfert 2 galaxy. Emission line modified by Doppler/gravitational shifts. Inclination and emissivity effects on line profiles. X-ray reflection spectrum, Innermost stable circular orbit. Fe line in the SO spectrum of MCG-6-30-15. Broadband light curve and broad Fe line. HEG spectrum and Fe line. Flux—fiux plots, Fe line flux and reffection fraction. Typical emissivity and line profiles, PLC photon index and RDC normalization. Variations of the line profile; observed and predicted. Light curves during a flare. Spectra compared to power laws. Spectral decomposition: variable power-law plus reflection component. xv 187 188, 190 ey 192 193 196 199 200, 202 205 206 207 207 214 215 216 218 220 221 222 224 227 228 229 eal 231 232 234 238 240 242 245 247 248 250 oye = 257 257 258 259 260 262List of illustrations Flux-flux plots. Prominent relativistic lines in GBHCs. Line profile of XTE J1650-500; Fe line flux. Model based on X-ray data of Seyfert | and NLS1 galaxies. Constellation-X simulated spectrum. ‘Time-averaged line profile of NGC 3516. Red feature folded light curve; on- and off-phases. Excess emission image; observational and theoretical. Redshift-corrected distributions of 27 long gamma ray bursts. Black holes with small kick velocities. Uniformly magnetized torus around a black hole. Stability diagram: neutral stability curves. GRB-supemnovae from rotating black holes. The Carter-Penrose diagram for a non-extremal BTZ black hole. A spacelike “stretched horizon” A. 263 265 266, 269 270 273 274 7 283 289 290 295 301 317 322 Part I General relativity: Classical studies of the Kerr geometry1 The Kerr spacetime — a brief introduction Matt Visser 1.1 Background cetime has now been with us for some 45 years (Kerr 1963, 1965). scovered in 1963 through an intellectual tour de force, and continues to provide highly non-trivial and challenging mathematical and physical problems to this day. This chapter provides a brief introduction to the Kerr spacetime and rotating black holes, touching on the most common coordinate representations of the spacetime metric and the key features of the geometry — the presence of horizons and ergospheres. The coverage is by no means complete, and serves chiefly to orient oneself when reading subsequent chapters. The final form of Albert Einstein’s general theory of relativity was developed in November 1915 (Einstein 1915, Hilbert 1915), and within two months Karl Schwarzschild (working with one of the slightly earlier versions of the theory) had already solved the field equations that determine the exact spacetime geom- etry of a non-rotating “point particle” (Schwarzschild 1916a). It was relatively quickly realized, via Birkhoff’s uniqueness theorem (Birkhoff 1923, Jebsen 1921, Deser and Franklin 2005, Johansen and Ravndal 2006), that the spacetime geom- etry in the vacuum region outside any localized spherically symmetric source is equivatent, up to a possible coordinate ransformation, to a portion of the Schwarzschild geometry — and so of direct physical interest to modelling the spacetime geometry surrounding and exterior to idealized non-rotating spherical stars and planets. (In counterpoint, for modelling the interior of a finite-size spher- ically symmetric source, Schwarzschild’s “constant density star” is a useful first approximation (Schwarzschild 1916b). This is often referred to as Schwarzschild’s “interior” solution, which is potentially confusing as it is an utterly distinct physical ‘The Kerr Spacetime: Rotating Black Holes in General Relativity. Ed, David L, Wiltshire, Matt Visser and Susan M. Scott. Published by Cambridge University Press, © Cambridge University Press 2009,a 4 Matt Visser spacetime solving the Einstein equations in the presence of a specified distribution of matter.) Considerably more slowly, only after intense debate was it realized that the “inward” analytic extension of Schwarzschild’s “exterior” solution represents a non-rotating black hole, the endpoint of stellar collapse (Oppenheimer and Snyder 1939). In the most common form (Schwarzschild coordinates, also known as. cur- vature coordinates), which is not always the most useful form for understanding the physics, the Schwarzschild geometry is described by the line element 2 ae=—[1 -=| a +7°(d8? +sin26 dg?), (1.1) r dr? 1—2m/r where the parameter m is the physical mass of the central object. But astrophysically, we know that stars (and for that matter planets) rotate, and from the weak-field approximation to the Einstein equations we even know the approximate form of the metric at large distances from a stationary isolated body of mass m and angular momentum J (Thirring and Lense 1918; Misner, Thorne, and Wheeler 1973; Adler, Bazin, and Schiffer 1975; D’Inverno 1992; Hartle 2003; Carroll 2004; Pfister 2005). In suitable coordinates: in? wa—[1-F+0(5) or [288s o( LJ r r r r + : + 7 +0 (=)] [dr? + r?(d6? + sin? dg]. (1.2) This approximate metric is perfectly adequate for almost all solar system tests of general relativity, but there certainly are well-known astrophysical situations (e.g. neutron stars) for which this approximation is inadequate ~ and so a “strong field” Solution is physically called for. Furthermore, if a rotating star were to undergo gravitational collapse, then the resulting black hole would be expected to retain at __ least some fraction of its initial angular momentum = thus s iggesting on physical Brounds that somehow there should be an extension of the Schwarzschild geometry to the situation where the central body carries angular momentum. Physicists and mathematicians looked for such a solution for many years, and had almost given up hope, until the Kerr solution was discovered in 1963 (Kerr 1963) — Some 48 years after the Einstein field equations were first developed. From the weak-field asymptotic result we can already see that angular momentum destroys Spherical symmetry, and this lack of spherical symmetry makes the calculations much more difficult. It is not that the basic principles are all that different, but simply that the algebraic complexity of the computations is so high that relatively few Physicists or mathematicians have the fortitude to carry them through to completion, The Kerr spacetime ~ a brief introduction 5 Indeed it is easy to both derive and check the Schwarzschild solution by hand, but for the Kerr spacetime the situation is rather different. For instance in Chan. drasekhar's magnum opus on black holes (Chandrasekhar 1998), only part of which is devoted to the Kerr spacetime, he is moved to comment: ‘The treatment of the perturbations of the Kerr space-time in this chapter has been prolixius in its complexity. Perhaps, at a later time, the complexity will be unravelled by deeper insights. But meantime, the analysis has led us into a realm of the rococo: splendorous, joyful, and immensely ornate More generally, Chandrasekhar also comments: The nature of developments simply does not allow a presentation that can be followed in detail with modest effort: the reductions that are required to go from one step to another are often very elaborate and, on occasion, may require as many as ten, twenty, or even fifty pages. Of course the Kerr spacetime was not constructed ex nihilo. Some of Roy Kerr's early thoughts on this and related matters can be found in (Kerr 1959; Kerr and Goldberg 1961), and over the years he has periodically revisited this theme (Goldberg and Kerr 1964; Kerr and Schild 1965; Debney, Kerr, and Schild 1969; Kerr and Debney 1970; Kerr and Wilson 1977; Weir and Kerr 1977; Fackerell and Kerr 1991; Burinskii and Kerr 1995). For practical and efficient computation in the Kerr spacetime many researchers will prefer to use general symbolic manipulation packages such as Maple, Mathe- matica, or more specialized packages such as GR-tensor. When used with care and discretion, symbolic manipulation tools can greatly aid physical understanding and insight, Because of the complexity of calculations involving the Kerr spacetime there is relatively little textbook coverage dedicated to this topic. An early discussion can be found in the textbook by Adler, Bazin, and Schiffer (1975 second edition). The ly dedicated single-topic are also comparatively brief discussions in the research monograph by Hawking and Ellis (1975), and the standard textbooks by Misner, Thorne, and Wheeler (1973), D'Inverno (1992), Hartle (2003), and Carroll (2004). One should particu- larly note the 60-page chapter appearing in the very recent textbook by Plebatiski and Krasiriski (2006). An extensive and highly technical discussion of Kerr black holes is given in Chandrasekhar (1998), while an exhaustive discussion of the class of spacetimes described by Kerr-Schild metrics is presented in the book “Exact Solutions to Einstein's Field Equations” (Stephani et al., 2002). * For instance, the standard distribution of Maple makes some unusual choices forits sign conventions. ‘The signs of the Einstein tensor, Ricci tensor, and Ricci scalar (though nor the Riemann tensor and Weyl tensor) ate posite to what most physicists and mathematicians would expect. of is that by O*Neill (1995), There6 Matt Visser To orient the reader I will now provide some general discussion, and explicitly present the line element for the Kerr spacetime in its most commonly used coor- dinate systems. (Of course the physics cannot depend on the coordinate system, but specific computations can sometimes be simplified by choosing an appropriate coordinate chart.) 1.2 No Birkhoff theorem Physically, it must be emphasized that there is no Birkhoff theorem for rotating spacetimes — it is nor true that the spacetime geometry in the vacuum region outside a generic rotating star (or planet) is a part of the Kerr geometry, The best result one can obtain is the much milder statement that outside a rotating star (or planet) the geometry asymptotically approaches Kerr geometry. The basic problem is that in the Kerr geometry all the multipole moments are very closely related to each other ~ whereas in real physical stars (or planets) the mass quadrupole, octopole, and higher moments of the mass distribution can in principle be independently specified. Of course from electromagnetism you will remember that higher n-pole fields fall off as 1/r2*", so that far away from the object the lowest multipoles dominate — it is in this asymptotic sense that the Kerr geometry is relevant for rotating stars or planets. On the other hand, if the star (or planet) gravitationally collapses — then clas- sically a black hole can be formed. In this case there are a number of. powerful uniqueness theorems which guarantee the direct physical relevance of the Kerr Spacetime, but as the unique exact solution corresponding to stationary rotating black holes, (as opposed to merely being an asymptotic solution to the far field of rotating stars or planets). ——— 4.3 Kerr’s original coordinates ‘The very first version of the Kerr spacetime geometry to be explicitly written down in the literature was the line element (Kerr 1963) 2mr ds? = —]1- [ 1? +a? cos?6 + 2(du + asin? @ dd) (dr +a sin? @ do) + (r? + a? cos” @) (dd? + sin? 6 d@?). (1.3) ] (du +a sin? 6 dp? The key features of this spacetime geometry are: The Kerr spacetime ~ a brief introduction 7 Using symbolic manipulation software it is easy to verify that this manifold is Ricci flat, Rap = 0, and so satisfies the vacuum Einstein field equations. Verifying this by hand is at best tedious. ‘There are three off-diagonal terms in the metric — which is one of the features that makes computations tedious. By considering (for instance) the g, component of the metric, it is clear that for m £0 there is (at the very least) a coordinate singularity located at r? + a? cos? @ = 0, that is: 6=n/2. (1.4) We shall soon sce that this is actually a curvature singularity. In these particular coordi- nates there are no other obvious coordinate singularities. Since the line element is independent of both u and @ we immediately deduce the existence of two Killing vectors. Ordering the coordinates as (u, r, 6, 4) the two Killing vectors are U* = (1,0, 0, 0); R? = (0,0, 0, 1). (1.5) Any constant-coefficient linear combination of these Killing vectors will again be a Killing vector. Setting a 0 the line element reduces to ds? + — [! - =| du? +2 du dr + r? (d6? + sin? @ dp), (1.6) r which is the Schwarzschild geometry in the so-called “advanced Eddington-Finkelstein Coordinates”, Based on this, by analogy the line element (1.3) is ofien called the advanced Eddington—Finkelstein form of the Kert spacetime. Furthermore since we know that r = 0 is a curvature singularity in the Schwarzschild geometry, this strongly suggests (but does not yet prove) that the singularity in the Kerr spacetime at (r = 0,0 = 77/2) is a curvature singularity. Setting m — 0 the line element reduces to ds? + dsj = — (du +a sin? 6 doy’ + 2(du +a sin? O dé) (dr + asin? @ dp) +(°? +a? cos” 6) (d6" + sin? 4 46), ay which is actually (but certainly not obviously) flat Minkowski spacetime in disguise. This is most easily seen by using symbolic manipulation software to verify that for this simplified line element the Riemann tensor is identically zero: Renea —> 0. For the general situation, m # 0 # a, all the non-zero components of the Riemann tensor Contain at least one factor of m.8 Matt Visser + Indeed, in a suitably chosen orthonormal basis the result can be shown to be even stronger: Alll the non-zero components of the Riemann tensor are then proportional to m: Rigbea 1.8) (This point will be discussed more fully below, in the section on the rational polynomial form of the Kerr metric. See also the discussion in Plebariski and Krasiriski (2006).) + Furthermore, the only non-trivial quadratic curvature invariant is Ravea RO = Cabca C4 (1.9) 48m? (r? — a? ci (1.10) is actually a curvature singularity. (We would have strongly suspected this by considering the a + O case above.) * Interms of the m = 0 line element we can put the line element into manifestly Kerr-Schild form by writing 2mr in? 0 dy? Tate (Wu basin? 9 doy, (11) ds? = dso + or the equivalent 2mr Lr 1.12 Pyare“? 2 Bab = (Bo)av + where we define -1 1 0 oO 0 asin?@ 0 (? +4?) sin? @ and lq = (1,0, 0, asin’ 8). (1.14) * It is then easy to check that ¢, is a null vector, with respect to both gas and (go)ao, and that 2mr____ go @, (1.15) ab = (gy — B® = (go) P+acoed The Kerr spacetime — a brief introduction 2 where eae _ 0 -a r oe r . : a qu —a -a 0. (sin? @)~! and &* =(0,1,0,0). (1.17) + The determinant of the metric takes on a remarkably simple form det(gan) = —(r? +a? cos” 6) sin? @ = det((golon). (1.18) where the m dependence has cancelled, (This is a side effect of the fact that the metric is of the Kerr-Schild form.) At the curvature singularity we have as3| = du? $a? a singularity — + (1.19) showing that, in terms of the “background” geometry specified by the disguised Minkowski spacetime with metric (go)zo, the curvature singularity is a “ring”. Of course in terms of the “full” geometry, specified by the physical metric gq, the intuinsic geometry of the curvature singularity is, unavoidably and by definition, singular. The null vector field ¢* is an affinely parameterized null geodesic: Va? =0. (1.20) * More generally O00 @ | , ed a ee 0 0 0 rsinto ab (And it is easy to see that this automatically implies the null vector field €* is an affinely parameterized null geodesic.) * The divergence of the null vector field é° is also particularly simple 2r viet +a? cos (1.22) * Furthermore, with the results we already have it is easy to calculate la la 2r? a9)" Cea = Can lgol” cago” = ——F——,_ (1.23) Lamy te? = 7 ' a) PF ae coe OF10 Matt Visser whence Leas) C2 (Wal 2 a (1.24) This invariant condition implies that the null vector field £4 is “shear free”. + Define a one-form £ by f= ly dx" = du + asin’ 6 dg, (1.25) then dé = asin26 d6 A dd #0, (1.26) but implying de Ade =0. (1.27) + Similarly Ade =asin26 du Add A dp £0, (1.28) but in terms of the Hodge-star we have the invariant relation #(£ Ade) = (1.29) or in component notation tly bea) = ee es (1.30) This allows one to pick off the so-called “twist” as oe eee (1.31) r? + a? cos?” This list of properties is a quick, but certainly not exhaustive, survey of the key features of the spacetime that can be established by direct computation in this Particular coordinate system. 1.4 Kerr-Schild “Cartesian” coordinates The second version of the Kerr line element presented in the original article (Kerr 1963), also discussed in the early follow-up conference contribution (Kerr 1965), The Kerr spacetime ~ a brief introduction a was defined in terms of “Cartesian” coordinates (1, x, y, z): ds? = —dt? + dx? + dy? + dz? (1.32) 3 yd y dx — xd : mr’ arp Et ly) | ay e WD za . r4 + az? a+r? err 7 subject to r(x, y, z), which is now a dependent function, not a coordinate, being implicitly determined by: Par eear+e[ (1.33) + The coordinate transformation used in going from (1.3) to (1.32) is t=u-r; xtiy=( ia) el sind; rosé. (1.34) Sometimes it is more convenient to explicitly write x = (rcos# +a sing) sind = vr? +a? sin@ cosfg —arctan(a/r)|; (1.35) y =(rsing —acos@)sind = Vr? +a? sind sin[p — arctan(a/r)]; (1.36) and so deduce (137) or the equivalent (1.38) * The m — 0 limit is now manifestly Minkowski space : ds? dsf=—d? bd tdye tae, 39) Of course the coordinate transformation (1.34) used in going from (1.3) to (1.32) is also responsible for taking (1.7) to (1.39). * The a > 0 limit is ds? > —dr? + dx? + dy? + dz? + y dy +2 de! Beg Str gr ) r r | » (1.40) now with at12 Matt Visser After a change of coordinates this can also be written as - 2 2 2692 : 2: 2m ds? > a1? + dr? + 2(46? + sin? 9 ag?) + — [ar + ar, (1.42) which is perhaps more readily recognized as the Schwarzschild spacetime. In fact if we set w= 1 +r then we regain the Schwarzschild line element in advanced Eddington— Finkelstein coordinates; compare with Equation (1.6). + The full m 4 0 # a metric is now manifestly of the Kerr-Schild form a Bab = Nab + aes ba bos (1.43) where now rx-fay ry—ax z fa = (1, 5-3, 5-5. -)- : ( P+a?? +a? ;) _ Here €, is a null vector with respect to both guy and nap and (1.45) with (1.46) a where the positive root has been chosen so that r > R at large distances. Because of this relatively complicated expression for r(x, y, 2), direct evaluation of the Ricci tensor via symbolic manipulation packages is prohibitively expensive in terms of computer Tesources — it is better to check the Ry = 0 in some other coordinate system, and then appeal to the known trivial transformation law for the zero tensor. * Consider the quadratic curvature invariant Rabea R° 4 = Capea C4, which we have already seen exhibits a curvature singularity at (r = 0, @ = 2/2). In this new coordinate system the curvature singularity is located at wy’ z=0. (1.48) Again we recognize this as occurring on a “rin background space. ", now in the (x,y,z) “Cartesian” The Kerr spacetime — a brief introduction 13 « In these coordinates the existence of a time-translation Killing vector, with components K* =(1,0,0,0) (1.49) is obvious. Less obvious is what happens to the azimuthal (rotational) Killing vector, which now takes the form = 0,0, y,—x). (1.50) * Note that in these coordinates i. 2mr3 OST ae oe and consequently (1.52) Thus V+ is certainly a timelike vector in the region r > 0, implying that this portion of the manifold is “stably causal”, and that if one restricts attention to the region r > 0 there is no possibility of forming closed timelike curves. However, if one chooses to work the maximal analytic extension of the Kerr spacetime, then the region r < 0 does make sense (at least mathematically), and certainly does contain closed timelike curves. (See for instance the discussion in Hawking and Ellis (1975).) Many (most?) relativists would argue that this» < 0 portion of the maximally extended Kerr spacetime is purely of mathematical interest and not physically relevant to astrophysical black holes. So there is a quite definite trade-off in going to Cartesian coordinates — some parts of the geometry are easier to understand, others are more obscure. 1.5 Boyer-Lindquist coordinates Boyer-Lindquist coordinates are best motivated in two stages: First, consider a __ slightly different but completely equival Kerr's original “advanced Eddington—Finkelstein” form (1.3) via the coordinate substitution u=t+r, (1.53) in which case ds? = —di? 4dr? + 2a sin? 9 dr dp + (r? + a? cos? 6) do? + (r? + a?) sin? @ dg? Imray 4 dr +a sin? day. (1.54) r? +a? cos’ t form of the metric which follows from ___14 Mau Visser Here the second line is again simply flat 3-space in disguise. An advantage of this coordinate system is that ¢ can naturally be thought of as a time coordinate — at least at large distances near spatial infinity. There are, however, still three off-diagonal terms in the metric so this is not yet any great advance on the original form (1.3). One can easily consider the limits m —+ 0, a > 0, and the decomposition of this metric into Kerr-Schild form, but there are no real surprises. Second, it is now extremely useful to perform a further m-dependent coordinate transformation, which will put the line clement into Boyer-Lindquist form: t=to+2m [ ¢ ~tan—a f . —2mr +a?’ —2mr +a? (1.55) r=rei 9=6p,. (1.56) Making the transformation, and dropping the BL subscript, the Kerr line-element now takes the form: 2mr 4mra sin? 6 [l= ase fd? - ara Ls [ areas | Parco VAP (157) 4 [2 bereos?e dr? (2 + aco? 8) 662 rE a r? = Imr + a? nF cos 2mra? sin? | +[P+ Me Prooaey|sie dg’. + a? cos * These Boyer-Lindquist coordinates are particularly useful in that they minimize the number of off-diagonal components of the metric ~ there is now only one off-diagonal component. We shall subsequently see that this helps particularly in analysing the asymp- totic behaviour, and in trying to understand the key difference between an “event horizon” and an “ergosphere”, — : Another particularly useful feature is that the asymptotic (r —> oo) behaviour in Boyer Lindquist coordinates is in? as? -[-™ )] ar [A + 0 do dt r ~ - 2m 1 2 4 2 qQ2 2 2 - =... . (1.58) From this we conclude that m is indeed the mass and J = ma is indeed the angular momentum, * Ifa — 0 the Boyer-Lindquist line element reproduces the Schwarzschild line element in standard Schwarzschild curvature coordinates. The Kerr spacetime ~ a brief introduction 15 + If m — 0 the Boyer-Lindquist line element reduces to 2, +a? cos6? ds? + -dr? + ae dr? (1.59) +07? +a? cos6”) dé? + (r? +a”) sin? @ dp, This is flat Minkowski space in so-called “oblate spheroidal” coordinates, and you can relate them to the ustal Cartesian coordinates of Euclidean 3-space by defining x=svre+a? sin@ cos¢; (1.60) Vr? +a? sin sing: (61) r cosd. (1.62) * One can re-write the Boyer-Lindquist line element as 2 603 62 le ie = ar (1.63) rte +0? +a? cos6?) db? + (7? +47) sin? dg? 2m [ [dt —asin?6 dg? | (1 +a? cos? @/r?) dr (1 +a? cos? 4/7?) ; mr +a? /r? dt? + In view of the previous comment, this makes it clear that the Kerr geometry is of the form (flat Minkowski space) + (distortion). Note, however, that this is not a Kerr-Schild decomposition for the Boyer-Lindquist form of the Kerr line element. * Of course there will still be a Ker—Schild decomposition of the metric 2mr (1.64) (1.65) or the equivalent (1.66) Pethaps more surprisingly, at least at first glance, because of the m-dependent coordinate transformation used in going to Boyer-Lindquist coordinates the “background” metric 8. In these coordinates we16 Matt Visser (8o)ab takes on the form: -1 ~ateet 0 c Imr__ (Prva? cos? 9M? Amro") 0 2mar sin? Panta mr eay aan 0 0 P+ a*cos?6 0 0 0 (r? +a") sin? This is (arguably) the least transparent form of representing flat Minkowski space that one is likely to encounter in any reasonably natural setting. (That this is again flat Minkowski space in disguise can be verified by using symbolic manipulation packages to compute the Riemann tensor and checking that itis identically zero.) In short, while Boyer-Lindquist coordinates are well adapted to some purposes, they are very ill-adapted to probing the Kerr-Schild decomposition. + In these coordinates the time translation Killing vector is K* =(1,0,0,0), (1.68) while the azimuthal Killing vector is R’ =(0,0,0, 1), (1.69) * In Boyer-Lindquist coordinates, (because the r and @ coordinates have not been modified, and because the Jacobian determinants arising from the coordinate transformations (UF e)) and (1.55)-(1.56) are both unity), the determinant of the metric again takes the relatively simple m-independent form det(gas) = — sin’ @ (r? + a? cos” 9)? = det([golen). (1.70) ____..*_In Boyer-Lindquist coordinates, (because the rand-@- coordinates have-not-been-mod= ified), the invariant quantity Ryyea R% looks identical to that calculated for the line element (1.3), Namely: 48m? (r? — a? cos? 0) [(7? + a? cos? 4)? — 1672a? cos? 6)] Raped RT = ved (P? +a? cos? 06 (71) * On the axis of rotation (@ = 0,0 = ls =—[ ), the Boyer-Lindquist line element reduces to: The Kerr spacetime — a brief introduction 7 This observation is useful in that it suggests that the on-axis geometry (and in particular the on-axis causal structure) is qualitatively similar to that of the Reissner-Nordstrém geometry. © On the equator (9 = 77/2) one has Ole =-[i-28 ar? - 2a : ed = | dg. + tmp eye +[F Fa |e 2 Oe ard i? — drag (1.73) Alternatively > a? Aare ce aye eee ce + 2m (dt +a dg)’. (1.74) r This geometry is still rather complicated and, in contrast to the on-axis geometry, has no “simple” analogue. (See, for example, Visser and Weinfurtner (2005).) 1.6 “Rational polynomial” coordinates Starting from Boyer-Lindquist coordinates, we can define a new coordinate by X = cos@ so that x € [-1, 1]. Then sind = /T— x2 and dx dy=singd0 > d0= (1.75) In terms of these (t, r, x, ) coordinates the Boyer-Lindquist version of the Kerr spacetime becomes 2mr 4amr(1 — x?) 2 oe i ds? ft 7 a) ae Tea wet fama oe) 2 Xe so ar + Foam tO +e) Te 2ma?r(1 — x?) ae 2 +a-x{r +a+ alge dg*. (1.76) This version of the Kerr metric has the virtue that all the metric components are how simple rational polynomials of the coordinates — the elimination of trigono- metric functions makes computer-based symbolic computations much less resource intensive — some calculations speed up by factors of 100 or more.18 Matt Visser * For instance, the quadratic curvature invariant A8(r? — a? x )[(r? + a? x?)? — 16r2a? x2] RS Rares = ca (+ a2x23° (1.77) can (on a “modem” laptop, c. 2008) now be extracted in a fraction of a second — as opposed to a minute or more when trigonometric functions are involved. * More generally, consider the (inverse) tetrad e 4 é (1.78) & specified by (BHF 0 0 (1.79) which at worst contains square roots of rational polynomials. Then it is easy to check AB sar enep =n®; nap ed ef = gap. (1.80) * Furthermore, the associated tetrad is a (1.81) specified by (1.82) ? Other computer-based symbolic manipulation approaches to calculating the curvature invariants have also been explicitly discussed in Lake (2003, 2004) The Kerr spacetime — a brief introduction 19 + In this orthonormal basis the non-zero components of the Riemann tensor take on only two distinct (and rather simple) values: Row: = —2 Ror = —2Ro3o3 = 2Ry212 = 2Ri3s3 = —Ro323 2mr(r? — 3a?x?) Pea 2 and 2max(3r? — a?x) (+ax2yP Rois = Rois = —Rosi2 (1.84) (See the related discussion in Plebaiiski and Krasiriski 2006.) « As promised these orthonormal components of the Riemann tensor are now linear in m, Furthermore, note that the only place where any of the orthonormal components becomes infinite is where r = 0 and x = 0 — thus verifying that the ring singularity we identified by looking at the scalar Rases R*’™ is indeed the whole story ~ there are no other curvature singularities hiding elsewhere in the Kerr spacetime. * On the equator, x = 0, the only non-zero parts of the Riemann tensor are particularly simple Rows = —2 Ror = —2Ro303 = 2Riaiz = 2Risi3 = — Ross = 2m (1.85) r which are the same as would arise in the Schwarzschild geometry. * On the axis of rotation, x = +1, we have Row = —2Ro303 = 2Ri2i2 = 2Risis = — Rosa 3a’) 5 (1.86) P+ay and _ (QmaBr— at) Rows = Roos = — Rosy = + (1.87) 12s = Ronis 0312 aay 1.7 Doran coordinates The coordinates relatively recently introduced by Chris Doran, see (Doran 2000), give yet another view on the Kerr spacetime. This coordinate system was specifi- cally developed to be as close as possible to the Painlevé-Gullstrand form of the Schwarzschild line element (Painlevé 1921, Gullstrand 1922, Visser 1998), a form that has become more popular over the last decade with continued developments in the “analogue spacetime” programme.20 Matt Visser Specifically, one takes the original Eddington—Finkelstein form of the Kerr line element (1.3) and performs the m-dependent substitutions dr du = dt + —"___., (1.88) : 1+ J2mr/(r +a) a hoa (1.89) r2 + a2 + /2mr(r? + a’) After dropping the subscript “Doran”, in the new (t, r, 4, @) coordinates Doran’s version of the Kerr line element takes the form (Doran 2000): ds? = —dr? + (7? + a? cos? 8) d? + (r? + a”) sin? 8 dg” (1.90) . [ae] fers /2mr(r? + a) rP+a r? +a? cos? (dt — Key features of this line element are: * Asa — > Oone obtains 2 2 ds? > ar? + {a Fe ea 7 which is simply Schwarzschild’s geometry in Painlevé-Gullstrand form (Painlevé 1921, Gullstrand 1922, Visser 1998). * Asm — O one obtains +r°(d0 + sin? 6 d¢?), (1.91) 2 ds? > dsj = —dr? + [ (1.92) + (r? + a? cos” 0) d6? + (r? + a?) sin? 6 dg, ~which is flat Minkowski space irr oblate spheroidat coordinates: Verifying that the line element is Ricci flat is best done with a symbolic manipulation package. Again, since the r and 0 coordinates have not changed, the explicit formulae for det( gq») and R®¢ Ry. are unaltered from those obtained in the Eddington—Finkelstein or Boyer-Lindquist coordinates. A nice feature of the Doran form of the metric is that for the contravariant inverse metric gi=-l. (1.93) In the language of the ADM formalism, the Doran coordinates slice the Kerr spacetime in such a way that the “lapse” is everywhere unity. This can be phrased more invariantly The Kerr spacetime — a brief introduction 2 as the statement 8” Vat Vet = 1, (1.94) implying gt Vat VeVot =0, (1.95) whence (Vat) 8% Vo(Vet) = 0. (1.96) That is, the vector field specified by W? = Vt = 9" Vit = (1.97) is an affinely parameterized timelike geodesic. Integral curves of the vector W* = V"t (in the Doran coordinate system) are thus the closest analogue the Kerr spacetime has to the notion of a set of “inertial frames” that are initially at rest at infinity, and are then permitted to free-fall towards the singularity. + Because of this unit lapse feature, it might at first seem that in Doran coordinates the Kerr spacetime is stably causal and that causal pathologies are thereby forbidden. However, the coordinate transformation leading to the Doran form of the metric breaks down in the region r < 0 of the maximally extended Kerr spacetime. (The coordinate transformation becomes complex.) So there is in fact complete agreement between Doran and Kerr- Schild forms of the metric: The region r > 0 is stably causal, and timelike curves confined to lie in the region r > 0 cannot close back on themselves. There is also a “Cartesian” form of Doran’s line element in (t, x, y, z) coor- dinates. Slightly modifying the presentation of (Doran 2000), the line element is given by Gs? = de? + dx? + dy? + de? +(F? Va Vo + FLVaSp + Sa Vol) dx? de®, (1.98) where (1.99) and where r(x, y, z) is the same quantity as appeared in the Kerr—Schild coordi- Rates, of. Equation (1.33). Furthermore r2(r? + a?) ay -ax Va = aaa (|) as + 1.100) r? 4 arg? Pa? ra 0) ‘ :22 Matt Visser S= rr? + i) _ 4+ a?z? Was It is then easy to check that V and S are orthonormal with respect to the “back- ground” Minkowski metric nap: and 101) nf? VaVe=—1; nf VaSp=0; 1” SaSp = +1. (1.102) These vectors are not orthonormal with respect to the full metric g,5 but are nevertheless useful for investigating principal null congruences (Doran 2000). 1.8 Other coordinates? We have so far seen various common coordinate systems that have been used over the past 44 years to probe the Kerr geometry. As an open-ended exercise, one could find (either via Google, the arXiv, the library, or original research) as many different coordinate systems for Kerr as one could stomach. Which of these coordinate systems is the “nicest”? This is by no means obvious. It is conceivable (if unlikely) that other coordinate systems may be found in the future that might make computations in the Kerr spacetime significantly simpler. 1.9 Horizons To briefly survey the key properties of the horizons and ergospheres occurring in the Kerr spacetime (see Figure 1.1), let us concentrate on using the Boyer-Lindquist coordinates. First, consider the components of the metric in these coordinates. The metric components have singularities when either ea cine #0 tha _-and__@.= 1/2, (1.103) or that is, r=rs=m+tVm—-a@ The first of these possibilities corresponds to what we have already seen is a real physical curvature singularity, since Rapea R04 Ravea R24 remains finite at ry. In fact the second option above (r = rs.) corresponds to all orthonormal cur- vature components and all curvature invariants being finite — it is a coordinate singularity but not a curvature singularity. Furthermore as a > 0 we have the 1—2m/fr +a?/r? (1.104) — oo there. In contrast The Kerr spacetime — a brief introduction 23 z uter event horizon re amt vinta Inner event horizon Jt a Outer ergosurface hams vata cod Inner ergosurface m—Vmt =a OO Ergoregion 0 Ring singularity Pe yt aa? and Symmetry axis 0 = 0,7 Fig, 1.1. Schematic location of the horizons, ergosurfaces, and curvature singu- larity in the Kerr spacetime, smooth limit that rs, > 2m, the location of the horizon in the Schwarzschild ge- ometry. We will therefore tentatively identify rs as the locations of the inner and outer horizons. (I shall restrict our attention to the case a < m to avoid having to deal with cither the extremal case a = m or the naked singularities that occur for a>m) Consider now the set of three-dimensional surfaces formed by fixing the r coordinate to take on some specified fixed value r = r,, and letting the other three surfaces the induced metric is found by formally setting dr — 0 and r > ry 80 that 53 surace = —St? + (7? + a? cos 6) dO? + (r2 +a) sin? @ dg? 2m [dt —a sin? dp? r. (14a? cos? 6/72)” (1.105) Now we could calculate the determinant of this 3-metric directly. Alternatively we could use the the block-diagonal properties of the metric in Boyer-Lindquist coordinates plus the fact that for the full (3 + 1)-dimensional metric we have coordinates (¢, @, ) run over their respective ranges. On these three-dimensional24 Matt Visser already observed det(gap) = —sin? 6 (r? + a? cos? @)?, (1.106) to allow us to deduce that on the 3-surface r = r,: det(gi;)s.surface = —sin? 6 (r? + a? cos? 8) (r? — 2mry + a”). (1.107) Now forr, > ry orr, < r_ the determinant det(g;;)3-surface is negative, which is a necessary condition for these 3-surfaces to be (2+ 1) dimensional (two space plus one time dimension). For r. € (r_, 7.) the determinant det(¢;)3 uae i8 positive, indicating the lack of any time dimension — the label “t” is now misleading and “t” actually denotes a spacelike direction. Specifically at what we shall soon see are the inner and outer horizons, r, = r+, the determinant is zero — indicating that the induced 3-metric [gi Js.urtce is singular (in the sense of being represented by a singular matrix, not in the sense that there is a curvature singularity) everywhere on both of these 3-surfaces. In particular, since [g,;] is a singular matrix, then at each point in either of the 3-surfaces at r = rz there will be some 3-vector L' that lies in the 3-surface r = rs. such that Igij] Li = 0, (1108) which implies in particular [gi] L' L/ =0. (1.109) Now promote the 3-vector L! [which lives in three-dimensional (t, 8, @) space] to a 4-vector by tacking on an extra coefficient that has value zero: Li La (L',0, L°, L® (1.110) ~ Then in the (3 + 1)-dimensional sense we have Sav L° L° =0 — (L* only defined at r = ry), (1.111) i.e, there is a set of curves, described by the vector L’, that lie precisely on the 3-surfaces r = rs. and which do not leave those 3-surfaces. Furthermore on the surfaces r = ry. these curves are null curves and L* is a null vector, Note that these null vector fields L” are defined only on the inner and outer horizons r = rs, and that they are quite distinct from the null vector field €* occurring in the Kerr—Schild decomposition of the metric, that vector field being defined throughout the entire spacetime. The Kerr spacetime ~ a brief introduction 25 Physically, the vector fields L* correspond to photon “orbits” that skim along the surface of the inner and outer horizons without either falling in or escaping to infinity. You should then be able to easily convince yourself that the outer horizon is an “event horizon” (“absolute horizon”) in the sense of being the boundary of the region from which null curves do not escape to infinity, and we shall often concentrate discussion on the outer horizon The Indeed if we define quantities Q. by a a Q= — 8 Omir, Ea or and now define L4 = (L4, 0,0, £4) = (1, 0,0, 24), (1.113) then it is easy to check that Banlrs) LE LE =0 (1.114) at the 3-surfaces r = ry. respectively. That is, the spacetime curves XO =O 1, 00), $7) =, re, O, G0 +24 0) (1.115) are an explicit set of null curves (so they could represent photon trajectories) that skim along the 3-surfaces r = rs, the outer and inner horizons. In fact r, is the “event horizon” of the Kerr black hole, and Q,, is a constant over the 3-surface r = r,.. This is a consequence of the “rigidity theorems”, and the quantity ., is referred to as the “angular velocity” of the event horizon. The event horizon “rotates” as though it were a solid body. (Somewhat counter-intuitively, the inner horizon also rotates as though it were a solid body, but with a different angular velocityQ_.) a Because the inner and outer horizons are specified by the simple condition " = rs, you might be tempted to deduce that the horizons are “spherical”. Disabuse yourself of this notion. We have already looked at the induced geometry of the 3-surface r= rs and found the induced metric to be described by a singular matrix. Now let’s additionally throw away the t direction, and look at the 2- surface r = r4,t = 0, which is a two-dimensional “constant time” slice through the horizon. Following our previous discussion for 3-surfaces the intrinsic 2-geometry of this slice is described by Smarr (1973) aaa 83 sartace = (73 + a cos? 0) w+ ( mre ) sin?6 d@?, (1.116) ri +a? cos?626 Matt Visser or equivalently be+eP ri +a? cos?@ 83 tase = E+ a? cos” 8) db? + ( ) sin? dg?. (1.117) So while the horizons are topologically spherical, they are emphatically not geo- metrically spherical. In fact the area of the horizons is (Smarr 1973) Ag = 4a(r} +a?) = 82m? + Vm — ma?) = 8m? + Vt — 12), (1.118) Furthermore, the Ricci scalar is (Smarr 1973) (rk. + a”) (72, — 3a? cos? 4) Ro-sutace = —=—5-—$ L119 aa (ri + a cos? 6)> (1.119) At the equator 22 +a? ea) 0, 1.120) A ¢ )) Ro surface > but at the poles 2(r2 +.) (72. — 3a?) Re-surtace > (+a) (1.121) So the intrinsic curvature of the outer horizon (the event horizon) can actually be negative near the axis of rotation if 3a? > r?, corresponding to a > (3/2) m — and this happens before one reaches the extremal case a = m where the event horizon vanishes and a naked singularity forms. The Ricci 2-scalar of the outer horizon is negative for @ < cos"! (=z). or 0 > ~ cos"! Car (1.122) a a which only has a non-vacuous range if a > (./3/2) m. In contrast, on the inner horizon the Ricci 2-scalar goes negative for Ga “G which always includes the region immediately surrounding the axis of rotation. In fact for slowly rotating Kerr spacetimes, a < m, the Ricci scalar goes negative on 0 < cos” ). or = 6 >a —cos ). (1.123) The Kerr spacetime ~ a brief introduction 7] the inner horizon for the rather large range la la 6 rf(m,a,0) > ry =m > r_ > rg(m,a,8) > 0. (1.141) In fact rm, a, 6) > r+ with equality only at @ = 0 and 6 = x (corresponding to the axis of rotation). Similarly r;(m, a, 6) < r_ with equality only at the axis of rotation. (This inner stationary limit surface touches the inner horizon on the axis of rotation, but then plunges down to the curvature singularity r = 0 at the equator, 6 = 7/2.) Restricting attention to the “outer” region: There is a region between the outer stationary limit surface and the outer event horizon in which it is impossible to “stand still”, but it is still possible to escape to infinity. This region is known as the “ergosphere”. Note that as rotation is switched off, a — 0, the stationary limit surface moves to lie on top of the event horizon and the ergosphere dis- appears. (Sometimes one sees authors refer to the stationary limit surface as the “ergosurface”, and to refer to the ergosphere as the “ergoregion”. Some authors furthermore use the word “ergoregion” to refer to the entire region between the stationary limit surfaces — including the black hole region located between the horizons.) * By setting r = rf (m, a, 0) and then replacing 2 Berry Ga (os) oe (1.142) & = \ we Vint — at e0s'8 one can find the intrinsic induced 3-geometry on the ergosurface. One form of the result is dS} geometry = —2a sin? @ dr dp (1.143) = 4 DME 40? +2 [mri +a sin? 6] sin? 0 46, i cos? 030 Matt Visser This result is actually quite horrid, since if we wish to be explicit 55 geometry = 24 sin? @ dr dp (1.144) a 2m3(m + Vm? = a? cos? 4) 9? a* cos? +2[m(m a vm? — a? cos? 6) +a? 6] sin? 6 dé’. + By now additionally setting dr = 0 one can find the intrinsic induced 2-geometry on an instantaneous constant time-slice of the ergosurface (Pelavas, Neary, and Lake 2001). One obtains 2mm + Vin? m? = a? cos? 8 +2[m(m + ym? — a? cos? 4) + a? cos? @) 6] sin? 9 ag. * The intrinsic area of the ergosurface can then be evaluated as an explicit expression involving two incomplete elliptic integrals (Pelavas, Neary, and Lake 2001). A more tractable approximate expression for the area of the outer ergosurface is (Pelavas, Neary, and Lake 2001): Ab =4n [come +a + gaa, 19a: al? + + Sams +0(S)]. (1.145) + The two-dimensional Ricci scalar is easily calculated but is again quite horrid (Pelavas, Neary, and Lake 2001). A tractable approximation is ~ 6 cos? 6)a2 aoa? cos? 6)a4 6 . i [i+ cos? O)a” _ 3(1 ~ Sos? + 3cos* A)a’ +o(4 )I: 2m? am? 8m* ms (1.146) On the equator of the outer ergosurface there is a simple exact result that | Am? + Sa? REO = 2/2) = (1.147) 4m? (2m? + a?)” At the poles the Ricci scalar has a delta function contribution coming from conical singularities at the north and south poles of the ergosurface, Near the north pole the metric takes the form ? dd? + oe] (1.148) where K is an irrelevant constant. The Kerr spacetime — a brief introduction 31 Fig. 1.2. Isometric embeddings in Euclidean space of the outer ergosurface and a portion of the outer horizon. For these images a = 0.90 m. A cut to show the (partial) outer horizon, which is not isometrically embeddable around the poles, is shown. The polar radius of the ergosurface diverges for a > m as the conical singularities at the poles become more pronounced. ‘An isometric embedding of the outer ergosurface and a portion of the outer horizon in Euclidean 3-space is presented in Fig. 1.2. Note the conical singularities at the north and south poles. The same techniques applied to the inner ergosurface yield 292/2 — 283a° a in m 25 = 20s?) _ 76 = 198cos*d + 128cos*# ~ 25cos°O |, (5) (2—cos? Pa? (8 — 12. cos? 6 + 6 cos* 6 — cos® A)m? mt }* (1.150) At the equator, the inner ergosurface touches the physical singularity. In terms of the intrinsic geometry of the inner ergosurface this shows up as the two-dimensional Ricci scalar becoming infinite. The metric takes the form ds? = a9 — 17/2) dO + a?(2 — 30 — 2/2)"} dg? + O[ — 7/2)", (1-151)—ant the regiorr immediately betow the event horizon 32 Matt Visser and after a change of variables can be written ds? = dh? + (2a? — 6alh|} dd? + O1A7]. (1.152) ‘At the poles the inner ergosurface has conical singularities of the same type as the outer ergosurface. See Equation (1.148). « Working now in Kerr-Schild Cartesian coordinates, since the ergosurface is defined by the coordinate condition g,, = 0, we see that this occurs at 2 4 mr +a (1.153) But recall that r(x, y, 2) is a rather complicated function of the Cartesian coordinates, so even in the “background” geometry the ergosurface is quite tricky to analyse. In fact itis better to describe the ergosurface parametrically by observing /x2. + yp = Vre@P +a? sind; ze = re(@) cosé. (1.154) ‘That is noe [24y= ([x + Vm? — a? cos 8 +0) sind; (1.155) ze = (m + yim? — a? cos?) cosd. (1.156) So again we see that the ergosurfaces are quite tricky to work with. See Fig. 1.3 for a polar slice through the Kerr spacetime in these Cartesian coordinates, Putting these particular mathematical issues aside: physically and astrophysi- cally, it is extremely important to realize that (assuming validity of the Einstein equations, which is certainly extremely reasonable given their current level of experimental and observational support) you should trust in the existence of the ergosurface and event horizon (that is, the outer ergosurface and outer horizon), However, you should not physically trust in the inner horizon or the inner ergosurface. Although they are certainly there as mathematical solutions of the exact vacuum Einstein equations, there are good physics reasons to suspect that the region at and inside the inner horizon, which can be shown to be a Cauchy horizon, is grossly unstable — even classically — and unlikely to form in any real astrophysical collapse. Aside from issues of stability, note that although the causal pathologies (closed timelike curves) in the Kerr spacetime have their genesis in the maximally extended r <0 region, the effects of these causal pathologies can reach out into part of the r > Oregion, in fact out to the inner horizon at r = r_ (Hawking and Ellis 1975) — so the inner horizon is also a chronology horizon for the maximally extended Kerr The Kerr spacetime ~ a brief introduction 33 | | Fig. 1.3. Polar slice through the Kerr spacetime in Cartesian KerrSchild coordi- nates. Location of the horizons, ergosurfaces, and curvature singularity is shown for a = 0.99 m and m = 1. Note that the inner and outer horizons are ellipses in these coordinates, while the inner and outer ergosurfaces are more complicated. The curvature singularity lies at the kink in the inner ergosurface. spacetime. Just what does go on deep inside a classical or semiclassical black hole formed in real astrophysical collapse is still being debated — see for instance the literature regarding “mass inflation” for some ideas (Poisson and Israel 1989; Balbinot and Poisson 1993). For astrophysical purposes it is certainly safe to discard the r < 0 region, and almost all relativists would agree that it is safe to discard the entire region inside the inner horizon r < r_. 1.11 Killing vectors By considering the Killing vectors of the Kerr spacetime it is possible to develop more invariant characterizations of the ergosurfaces and horizons. The two obvious34 Matt Visser Killing vectors are the time translation Killing vector K@ and the azimuthal Killing vector R%; in addition any constant coefficient linear combination a K* + b R4 is also a Killing vector — and this exhausts the set of all Killing vectors of the Kerr spacetime. ‘The time translation Killing vector K“ is singled out as being the unique Killing vector that approaches a unit timelike vector at spatial infinity. The vanishing of the norm Bap KO Kb =0 (1.157) is an invariant characterization of the ergosurfaces — the ergosurfaces are where the time translation Killing vector becomes null. Similarly, the horizon is a null 3-surface (since it is defined by a set of photon trajectories) that is further characterized by the fact that it is invariant (since the photons neither fall into nor escape from the black hole). This implies that there is some Killing vector a K* + b R* that becomes null on the horizon (a result that holds in any stationary spacetime, and leads to the “rigidity theorem”). Without loss of generality we can set a > 1 and define b > Qy so that the horizon is invariantly characterized by the condition Ban (K* + Qy RY) (KY + Qy RY) =0. (1.158) It is then a matter of computation to verify that this current definition of Qy coincides with that in the previous discussion. 1.12 Comments There are a (large) number of things I have not discussed in this brief introduction, but [have to cut things off somewhere. In particular, you can go to several textbooks, the original literature, and the rest of this book to see discussions + The Kerr-Newman geometry (a charged rotating black hole). * Carter-Penrose diagrams of causal structure, Maximal analytic extensions of the Kerr spacetime. Achronal regions and chronology horizons (time travel) in the idealized Kerr spacetime. Extremal and naked Kerr spacetimes; cosmic censorship. Particle and photon orbits in the Kerr black hole. Geodesic completeness. Physically reasonable sources and possible “interior solutions” for the Kerr spacetime. * Penrose process ~ energy extraction from a rotating black hole. Black hole uniqueness theorems. Singularity theorems guaranteeing the formation of black holes in certain circumstances. The Kerr spacetime — a brief introduction 35 * The classic laws of black hole mechanics; area increase theorem. * Black hole thermodynamics (Hawking temperature and Bekenstein entropy). ‘These and other issues continue to provoke considerable ongoing research — and I hope this brief introduction will serve to orient the reader, whet one’s appetite, and provoke interest in the rest of this book. Acknowledgments Supported by the Marsden Fund administered by the Royal Society of New Zealand. Figure 1.1 courtesy of Silke Weinfurtner. Figure 1.2 courtesy of Kayll Lake. Ad- ditionally, I wish to acknowledge useful comments from Silke Weinfurtner, Kayll Lake, Bartolome Alles, and Roy Kerr. References Adler, R. J., Bazin, M. & Schiffer, M. (1975), Introduction to General Relativity, Second edition (McGraw-Hill, New York). [It is important to acquire the 1975 second edition, the 1965 first edition does not contain any discussion of the Kerr spacetime.] Balbinot, R. & Poisson, E. (1993), “Mass inflation: the semiclassical regime”, Phys. Rev, Lett. 70, 13. Birkhoff, G. D. (1923), Relativity and Modern Physics (Harvard University Press, Cambridge). Burinskii, A. & Kerr, R. P. (1995), “Nonstationary Kerr congruences”, arXiv:gr-qc/ 9501012. Carroll, S. (2004), An Introduction to General Relativity: Spacetime and Geometry (Addison Wesley, San Francisco). Chandrasekhar, S. (1998), The Mathematical Theory of Black Holes (Oxford University Press, Oxford). Debney, G. C., Kerr, R. P, & Schild, A. (1969), “Solutions of the Einstein and Einstein-Maxwell equations”, J. Math. Phys. 10, 1842. 261 [arXiv:gr-qo/0408067). D'Inverno, R. (1992), Introducing Einstein’ s Relativity (Oxford University Press, Oxford). Doran, C. (2000), “A new form of the Kerr solution”, Phys. Rev. D 61, 067503 [arXiv:gr-qc/9910099}. Einstein, A. (1915), “Zur allgemeinen Relativitatstheorie”, Sitzungsber. Preuss. Akad. Wiss., 778; Addendum ibid., 799. Fackerell, E. D. & Kerr, R. P. (1991), “Einstein vacuum field equations with a single non-null Killing vector”, Gen. Rel. Grav. 23, 861. Goldberg, J. N. & Kerr, R. P. (1964), “Asymptotic properties of the electromagnetic field”, J. Math. Phys. 5,172. Gullstrand, A. (1922), “Allgemeine Lésung des statischen Einkérperproblems in der Einsteinschen Gravitationstheorie”, Ark. Mat. Astron. Fys. 16(8), 1. Harte, J. (2003), Gravity: an Introduction to Einstein's General Relativity (Addison Wesley, San Francisco).36 Matt Visser Hawking, S. W. & Ellis, G. F. R. (1975), The Large Scale Structure of Space-time (Cambridge University Press, Cambridge). Hilbert, D. (1915), “Die Grundlagen der Physik (Erste Mitteilung)”, Nachr. Gesell. Wiss. Gottingen Math. Phys. = Jebsen, J. T (1921), “Uber die allgemeinen kugelsymmetrischen Lisungen der Einsteinschen Gravitationsgleichungen im Vakuum”, Ark. Mat. Ast. Fys. 15(18), 1. Johansen, N. V. & Ravndal, F. (2006), “On the discovery of Birkhoff’s theorem”, Gen, Rel. Grav. 38, 537 {arXiv: physics/0508163). Kerr, R. P, (1959), “The Lorentz-covariant approximation method in general relativity”, "Nuovo Cimento 13, 469. Kerr, RP. & Goldberg, J. N. (1961), “Einstein spaces with four-parameter holonomy groups”, J. Math. Phys. 2, 332. Kerr, R. P. (1963), “Gravitational field of a spinning mass as an exam special metrics”, Phys. Rev: Lett. 11, 237. Kerr, R. P. (1965), “Gravitational collapse and rotation”, published in: Quasi-stellar sources and gravitational collapse: including the proceedings of the First Texas Symposium on Relativistic Astrophysics, ed. Robinson, 1. Schild, A. & Schiicking, E. L. University of Chicago Press, Chicago), pp. 99-102. ‘The conference was held in Austin, Texas, on 16-18 December 1963. Kerr, R. P. & Schild, A. (1965), “Some algebraically degenerate solutions of Binstein’s gravitational field equations”, Proc. Symp. Appl. Math. 17, 199. Kerr, R. P. & Debney, G. C. (1970), “Einstein spaces with symmetry groups”, J. Math Phys. 11, 2807. Kerr, R.P. & Wilson, W. B. (1977), “Singularities in the Kert-Schild metrics”, ‘Proceedings of the 8th International Conference on General Relativity and Gravitation, 7-12 August 1977, Waterloo, Ontario, Canada, p. 378. Lake, K. (2003), “Comment on negative squares of the Weyl tensor”, Gen. Rel. Grav. 35, 2271 [arXiv:gr-qc/0302087). Lake, K, (2004), “Differential invariants of the Kerr vacuum”, Gen. Rel. Grav. 36, 1159 [arXiv:gr-qc/0308038]. Misner, C. A., Thome, K. & Wheeler, J. A. (1973), Gravitation (W. H. Freeman, San Francisco). O'Neill, B. (1995), Geometry of Kerr Black Holes (A. K. Peters, Wellesly, MA). Oppenheimer, J. R. & Snyder, H. (1939), “On continued gravitational contraction”, Phys. Rey, 56, 455. é nc Se ple of algebraically “La mécanique classique et la theorie de la relati 173, 677. Pelavas, N., Neary, N. & Lake, K. (2001), “Properties of the instantaneous ergo surface of a Kerr black hole”, Class. Quantum Grav. 18, 1319 [arXiv:gr-qo/0012052]. Pfister, H. (2005), “On the history of the so-called Lense-Thirting effect”, hhttp://philsci-archive.pitt.edu/archive/0000268 1/01 /lense.pdf Plebasiski, J. & Krasitiski, A. (2006), An Introduction to General Relativity and Cosmology (Cambridge University Press, Cambridge). Poisson, E. & Israel, W. (1989), “Inner-horizon instability and mass inflation in black holes”, Phys, Rev. Lett. 63, 1663-6. Schwarzschild, K. (1916a), “Uber das Gravitationsfeld eines Massenpunktes nach der Einsteinschen Theorie”, Sitzungsber. Preuss. Akad. Wiss., 189. Schwarzschild, K. (19166), “Uber das Gravitationsfeld ciner Kugel aus inkompressibler Flassigkeit nach der Einsteinschen Theorie”, Sitzungsber. Preuss. Akad. Wiss. 424. Smarr, L. (1973), “Surface geometry of charged rotating black holes”, Phys. Rev. D 7, 269. The Kerr spacetime ~ a brief introduction 37 Stephani, H., Kramer, D., MacCallum, M. A. H., Hoenselaers, C. & Herlt, E. (2002), Exact ‘Solutions of Einstein's Field Equations (Cambridge University Press, Cambridge). Thirring, H. & Lense, J. (1918), “Uber den Einfluss der Eigenrotation der Zentralkérper ‘auf die Bewegung der Planeten und Monde nach der Einsteinschen Gravitationstheorie”, Phys. Z. 19, 156; English translation by B. Mashoon, F. W. Hehl & D. S. Theiss, “On the influence of the proper rotations of central bodies on the motions of planets and moons in Einstein's theory of gravity”, Gen. Rel. Grav. 16, 727 (1984). Visser, M. (1998), “Acoustic black holes: horizons, ergospheres, and Hawking radiation”, Class. Quantum Gray. 1, 1767 {arXiv:gr-qc/9712010}. Visser, M, & Weinfurtner, S. Ch. (2005), “Vortex geometry for the equatorial slice of the Kerr black hole”, Class. Quantum Grav. 22, 2493 [arXiv:gr-qc/0409014), Weir, G. & Kerr, R. P. (1977), “Diverging type-D metrics”, Proc. Roy. Soc. Lond., A 355, 31.2 Discovering the Kerr and Kerr—Schild metrics Roy P. Kerr 2.1 Introduction ‘The story of this metric begins with a paper by Alexei Zinovievich Petrov (1954) where the simultaneous invariants and canonical forms for the metric and conformal tensor are calculated at a general point in an Einstein space. This paper took a while to be appreciated in the West, probably because the Kazan State University journal was not readily available, but Felix Pirani (1957) used it as the foundation of an article on gravitational radiation theory. He analysed gravitational shock waves, calculated the possible jumps in the Riemann tensor across the wave fronts, and related these to the Petrov types. I was a graduate student at Cambridge, from 1955 to 1958. In my last year I was invited to attend the relativity seminars at Kings College in London, including one by Felix Pirani on his 1957 paper. At the time I thought that he was stretching when he proposed that radiation was type N, and I said so, a rather stupid thing for a graduate student with no real supervisor to do.' It seemed obvious that a superposition of type N solutions would not itself be type N, and that gravitational ___waves near a macroscopic body would be of general type. not type N. Perhaps I did Felix an injustice. His conclusions may have been oversimplified but his paper had some very positive consequences. Andrzej Trautman computed the asymptotic properties of the Weyl tensor for outgoing radiation by generalizing Sommerfeld’s work on electromagnetic radiation, confirming that the far field is type N. Bondi, van der Burg and Metzner (1962) then introduced appropriate null coordinates to study gravitational radiation in the far zone and related this to the results of Petrov and Pirani. My nominal supervisor was a particle physicist who had no interest in general relativity. ‘The Kerr Spacetime: Rotating Black Holes in General Relativity. Ed. David L.. Wiltshire, Maat Visser and Susan M. Scott. Published by Cambridge University Press, © Cambridge University Press 2009. The Kerr and Kerr—Schild metrics 39 In 1958 I went to Syracuse University as a research associate of Peter Bergmann. While there I was invited to join Joshua Goldberg at the Aeronautical Research Laboratory in Dayton Ohio.? There was another relativist at the lab, Dr Joseph Schell, who had studied Einstein’s unified field theory under Vaclav Hlavaty. Josh was about to go on study leave to Europe for a few months, and did not want to leave Joe by himself. Before he left, Josh and I became interested in the new methods that were entering general relativity from differential geometry at that time. We did not have a copy of Petrov’s paper so our first project was to re-derive his classification using projective geometry, something which was being done by many other people throughout the world at that time. In each empty Einstein space, €, the conformal tensor determines four null “eigenvectors” at each point. The metric is called algebraically special (AS)° if two of these eigenvectors coincide. This vector is then called a principal null vector (PNY) and the field of these is called a principal null “congruence”. After this, we used a tetrad formulation to study vacuum Einstein spaces with de- generate holonomy groups (Goldberg and Kerr (1961), Kerr and Goldberg (1961)). The tetrad used consisted of two null vectors and two real orthogonal space-like vectors, ds? =P + @Y + 20% ot We proved that the holonomy group must be an even-dimensional subgroup of the Lorentz group at each point, and that if its dimension is less than six then coordinates can be chosen so that the metric has the following form: ds? = dx? + dy? + 2du (dv + pdx + 4(@— pv) du), where both p and w are independent of v,* and Pax + Py = xx + ®xy = Wx — 2PP.xx — (P.xY + (3? This coordinate system was not quite uniquely defined. If p is bilinear in x and y then it can be transformed to zero, giving the well-known plane-fronted wave 2 There isa claim spread on internet that we were employed to develop an antigravity engine to power spaceships. ‘This is rubbish! The main reason why the US Air Force had created a General Relativity section was probably to show the navy that they could also do pure research, The only real use that the USAF made of us was when some crackpot sent them a proposal for antigravity or for converting rotary: motion inside a spaceship to & translational driving system. These proposals typically used Newton's equations to prove non-conservation of ‘momentum for some classical system. The term algebraically degenerate is sometimes used instead. 4 The simple way that the coordinate v appears was to prove typical of these algebraically special metrics.40 Roy P. Kerr solutions. These are type N, and have two-dimensional holonomy groups. The more general metrics are type III with four-dimensional holonomy groups. In September 1961 Joshua joined Hermann Bondi, Andrzej Trautman, Ray Sachs and others at King’s College in London, By this time it was well known that all such AS spaces possess a null congruence whose vectors are both geodesic and shearfree. These are the degenerate “eigenvectors” of the conformal tensor at each point, Andrzej suggested to Josh and Ray how they might prove the converse. This led to the celebrated Goldberg-Sachs theorem (see Goldberg and Sachs (1962)). Theorem 2.1. A vacuum metric is algebraically special if and only if it contains a geodesic and shearfree null congruence Either properties of the congruence, being geodesic and shear-free, or a property of the conformal tensor, algebraic degeneracy, could be considered fundamental with the others following from the Goldberg-Sachs theorem. It is likely that most thought that the algebra was fundamental, but I believe that Ivor Robinson and Andrzej Trautman (1962) were correct when they emphasized the properties of the congruence instead. They showed that for any Einstein space with a shear-free null congruence which is also hypersurface orthogonal there are coordinates for which ds? = 2r?P-2 de dé — 2du dr — (A In P — 2r(In P),y — 2m(u)/r) du?, where £ is a complex coordinate, ¢ = (x + iy)/-V2, say, so that 2de¢ dé = dx? + dy’. The one remaining field equation is, A = 2P Adz AA(In P) + 12m(in P),, — 4m, The PNVS is k = k“@,, = 8,, where r is an affine parameter along the rays. The corresponding differential form is k = k,dx" = du, so that k is the normal to the surfaces of constant u. The coordinate u is a retarded time, the surfaces of constant r, ware distorted spheres with metric ds? = 2r?P~? dg dg and the parameter m(u) is loosely connected with the system’s mass. This gives the complete solution to the Robinson—Trautman problem.’ In 1962 Goldberg and I attended a month-long 5 ‘The letter & will be used throughout this article to denote the PNY. In the study of exact solutions, “solving” a problem usually means introducing 2 useful coordinate system, solving the easier Einstein equations and replacing the ten components of the metric tensor with a smaller hhumber of functions, preferably of less than four variables. These will then have (o satisfy a residual set of differential equations, the harder ones, which usually have no known complete solution. For example, the remaining field equation for the Robinson-Traulman metrics is highly non-linear and has no general solution. The Kerr and Kerr—Schild metrics 41 Fig. 2.1. Ivor Robinson and Andrzej Trautman constructed all Einstein spaces possessing a hypersurface orthogonal shearfree congruence. Whereas Bondi and his colleagues were looking at spaces with these properties asymptotically, far from any sources, Robinson and Trautman went a step further, constructing exact solutions. (Images courtesy of Andrzej Trautman and the photographer, Marek Holzman.) meeting in Santa Barbara. It was designed to get mathematicians and relativists talking to each other. Perhaps the physicists learned a lot about more moder mathematical techniques, but I doubt that the geometers learned much from the relativists, All that aside, I met Alfred Schild at this conference. He had just per- Rel: at the Universit of Texas, and had arranged for an outstanding group of relativists to join. These included Roger Penrose and Ray Sachs, but neither could come immediately and so I was invited to visit for the 1962-1963 academic year. After Santa Barbara, Goldberg and I flew to a conference held at Jablonna near Warsaw, This was the third precursor to the triennial meetings of the GRG society and therefore it could be called GR3. Robinson and Trautman (1964) presented a paper on “Exact Degenerate Solutions” at this conference. They spoke about their well-known solution and also showed that when the rays are not hypersurface orthogonal coordinates can be chosen so that ds? = —P?[(d& — ak)? + (dn — bk)?] + 2dpk + ck’,42 Roy P. Kerr Fig. 2.2. Ezra T. Newman, who with T. Unti and L. A. Tambourino, studied the field equations for diverging and rotating algebraically special Einstein spaces. where, as usual, k is the PNV. Its components, k., are independent of p, but a, b,c and P may be functions of all four coordinates. I was playing around with the structure of the Einstein equations during 1962, using the new (to physicists) methods of tetrads and differential forms. I had written out the equations for the curvature using a complex null tetrad and self- dual bivectors, and then studied their integrability conditions. In particular, I was interested in the same problem that Robinson and Trautman were investigating where k was not a gradient, ie. twisting, but there was a major road block in my way. Alan Thompson had also come to Austin that year and was also interested in these methods. Although there seemed to be no reason why there should not be many algebraically special spaces, Alan kept quoting a result from a preprint _of a paper by Newman, Tambourino and Unti (1963) in which they had “ ‘that the only possible space with a diverging and rotating PNV is NUT space, a one parameter generalization of the Schwarzschild metric. They derived this result using the new Newman-Penrose (N-P) spinor formalism (Newman and Penrose, 1962). Their equations were essentially the same as those obtained by people such as myself using self-dual bivectors: only the names are different. T could not understand how the equations that I was studying could possibly lead to their claimed result, but could only presume it must be so since I did not have a copy of their paper. In the spring of 1963 Alan obtained a preprint of this paper and loaned it to me. I thumbed through it quickly, trying to see where their hunt for solutions had died. The N—P formalism assigns a different Greek letter to cach component of the The Kerr and Kerr—Schild metrics 43 connection, so I did not try to read it carefully, just rushed ahead until I found what appeared to be the key equation, (ny trp +:n3)a? = (2.1) where the n; were all small integers, between —4 and +4. Their sum was not zero so this gave a = 0. [had no idea what a represented, but its vanishing seemed to be disastrous and so I looked more carefully to see where this equation was coming from. Three of the previous equations, cach involving first derivatives of some of the connection components, had been differentiated and then added together. All the second derivatives cancelled identically and most of the other terms were eliminated using other N~P equations, leaving (2.1). ‘The mistake that Newman er al. made was that they did not notice that they were simply recalculating one component of the Bianchi identities by adding together the appropriate derivatives of three of their curvature equations, and then simplifying the result by using some of their other equations, undifferentiated. The result should have agreed with one of their derived Bianchi identities involving derivatives of the components of the conformal tensor, the V; functions, giving ny +m +n3=0. (2.2) In effect, they rediscovered one component of the identities, but with numerical errors, The real fault was the way the N-P formalism confuses the Bianchi identities with the derived equations involving derivatives of the W; variables. Alan Thompson and myself were living in adjoining apartments, so I dashed next door and told him that their result was incorrect. Although it was unnecessary, we recalculated the first of the three terms, 71, obtained a different result to the one in the preprint, and verified that (2.2) was now satisfied. Once this blockage was out of the way, I was then able to continue with what I had been doing and derive the metric and field equations for twisting algebraically special spaces. The coord £ s ted turned out to-be- essentially the-sat sth givenin, Robinson and Trautman (1962). This shows that they are the “natural” coordinates for this problem since the methods used by them were very different to those used by me. Ivor loathed the use of such things as N-P or rotation coefficients, and Andrzej and he had a nice way of proving the existence of their canonical complex coordinates ¢ and &. I found this same result from one of the Cartan equations, as will be shown in the next section, but I have no doubt that their method is more elegant. Ivor explained it to me on more than one occasion, but unfortunately T never understood what he was saying!” ? While writing this article I read their 1962 paper and finally understood how they derived their coordinates. It only took me 45 years,44 Roy P. Kerr At this point I presented the results at a monthly Relativity conference held at the Steven’s Institute in Hoboken, N. J. When I told Ted Newman that (2.1) should have been identically zero, he said that they knew that 7; was incorrect, but that the value for n given in the preprint was a misprint and so (2.2) was still not satisfied. I replied that since the sum had to be zero the final term, n3 must also be incorrect. Alan and I recalculated it that evening, confirming that (2.2) was satisfied.® 2.2 Discovery of the Kerr metric When I realized that the attempt by Newman et al. to find all rotating AS spaces had foundered and that Robinson and Trautman appeared to have stopped with the static ones, I rushed headlong into the search for these metrics. Why was the problem so interesting to me? Schwarzschild, by far the most significant physical solution known at that time, has an event horizon. A spherically symmetric star that collapses inside this is forever lost to us, but it was not known whether angular momentum could stop this collapse to a black hole. Unfortunately, there was no known metric for a rotating star. Schwarzschild was an example of the Robinson-Trautman metrics, none of which could contain a rotating source as they were all hypersurface orthogonal. Many had tried to solve the Einstein equations assuming a stationary and axially symmetric metric, but none had succeeded in finding any physically significant rotating solutions. The equations for such metrics are complicated non-linear PDEs in two variables. What was needed was some extra condition that would reduce these to ODEs, and this might be the assumption that the metric is AS. The notation used in the rest of this chapter is fairly standard. There were two competing formalisms being used around 1960, complex tetrads and spinors. I used the former,? Newman et al. the latter. The derived equations are essentially identical, but each approach has some advantages. Spinors make the the Petrov ~ classification triviat once it has been shown that a tensor with the symmetries of— the conformal tensor corresponds to a completely symmetric spinor, Vagcp. The standard notation for the components of this tensor is Wo = Wooo, Yi = Yooor.--. Y4 = Win. If ¢4 is an arbitrary spinor then the equation Waser S45%E°S? =O * Robinson and Trautman (1962) also doubted the original claim by Newman et al. since they knew that the « linearized equations had many solutions. Robinson and Trautman also had a fairly natural complex tetrad approach. The Kerr and Kerr—Schild metrics 45 is a homogeneous quartic equation with four complex roots, ¢/4. The related real null vectors, Z¢* = ¢%¢%, are the four PNVs of Petrov. The spinor ¢% = 8% gives a PNV if Wp = 0. It is a repeated root and therefore it is the principal null vector of an AS spacetime if Y, = 0 as well. ‘The Kerr (1963) letter presented the main results of my calculations but gave few details.!° The methods that I used to solve the equations for AS spaces are essentially those used in “Exact Solutions” by Stephani et al. (2003), culminating in their Equation (27.27). I will try to use the same notation as in that book since it is almost identical to the one I used in 1963, but I may get some of the signs wrong. Beware! Suppose that (eg) = (m, m, I, k) is a null tetrad, i.e. a set of four null vectors where the last two are real and the first two are complex conjugates. The corre- sponding dual forms are (w) = Gm, m, —k, —1) and the metric is!! ds? = 20mm — kl) = 2A(o!w - wo). (2.3) The vector k is a PNV and so its direction is uniquely defined, but the other directions are not. The form of the metric tensor in (2.3) is invariant under a combination of a null rotation (B) about k, a rotation (C) in the m A m plane and a Lorentz transformation (A) in the / A k plane, kK=k, m' = m-+ Bk, U =14+ Bm+ Bm+ BBk, (2.4a) K=k, m' =m, V=l, (2.4b) K = Ak, m' =m, Aer (2.4c) The most important connection form is Tay = Pq1g@" = m" kaip dx? ‘The optical scalars of Ray Sachs (Sachs, 1961, 1962) for k are just the components of this form with respect to the w*. o = Pau = shear; p =a = complex divergence; k = Pane = geodesy. 0 | spent many years trying to write up this research but, unfortunately, I could never decide whether (0 use Spinors or complex tevad. and thus it did not gt witen up unl Ker nd Debmey (1970). Geonge Debey also collaborated with Alfred Schild and myself on the Kerr—Schild metrics in Debney et al. (1969). 11 [ personally hate the minus sign in this expression and did not use it in 1963, but it seems to have become standard. By the time I finish this article I am sure that I will wish I had stuck with all positive signs!46 Roy P. Kerr The fourth component, "413, is not invariant under a null rotation about k, Pas = Pais + Bp, and has no real geometric significance. It can be set to zero using an appropriate null rotation. Also, since k is geodesic and shearfree for AS metrics, both « and a are zero and therefore Ta = po’. (2.5) If we use the simplest field equations, 0 Rig = 2Rag =0, Ra = Raz — Raza =O, Ri = 2Ra3 = 0, a total of five real equations, and the fact that the metric is AS, Yo = -2Rygt =0, 281 = Rata ~ Raras = 0, then the most important of the second Cartan equations simplifies to APs) Par A Pin t+ Pau) = Ratan" A@” = Rage Aw. (2.6) Taking the wedge product of (2.6) with T's; and using (2.5), ee (2.7) This was the key step in my study of these metrics but this result was not found in quite such a simple way. At first, I stumbled around using individual component equations rather than differential forms to look for a useful coordinate system. It was only after I had found this that I realized that using differential forms from the tions, ¢ and TT, such that Ty =dé/M, Ta =d¢/T. The two functions ¢ and its complex conjugate, &, will be used as (complex) coordinates. They are not quite unique since ¢ can always be replaced by an arbitrary analytic function &(¢). Using the transformations in (2.4b) and (2.4c), Tey = Ae Py = AeCde/M, => MW = Ate CN. The Kerr and Kerr—Schild metrics 47 Tl’ can be set to 1 by choosing Ae =I, and that is what I did in 1963, but it is also common to just use the C-transformation to convert it to a real function P, Ta = po? = dt /P. (2.8) This is the derivation for two of the coordinates used in 1963. Since w!k* = 0 —> k(t) = 0, these functions, ¢, ¢, are constant along the PNV. The other two coordinates were very standard and were used by most people considering similar problems at that time. The simplest field equation is Ry=0 > kp=py= so that the real part of —p~! is an affine parameter along the rays. This was the obvious choice for the third coordinate, r, pl =-(r +i). There was no clear choice for the fourth coordinate, so u was chosen so that Mug = 1, kuq = 0, a pair of consistent equations. Given these four coordinates, the basis forms are 1 o Mex = —dt/Pp = (r —id)do/P, & = figdx® = —do/Pp = (r +iZ)dE/P, ow =kgdx® = du t+ Ldg + Ldg, ot =lydx® = dr + We + Wi + Ho. where L is independent of R, and the coefficients £, W and H have still to be determined. the simplest component of the second Cartan equation, (2.43), © and W were calculated as functions of L and its derivatives," 2D=P(GL-dL), a= Wa +iX)Ly + idd. — Liu, The remaining field equations, the “hard” ones, were more complicated, but still fairly straightforward to calculate. Two gave H as a function of a real “mass” 2 In Kerr (1963) &, D and A were used instead of 1, @ and E, but the results were the same, mutatis mutandis.48 Roy P. Kerr function m(u, ¢, ) and certain functions of the higher derivatives of P and L,!? mr + MZ rP+y2 M = XK + P’Re[dE — 2L,,0E — Ea, dL], K =2P°Re(ad In P — L.)), H=3K—r(nP)y— Finally, the first derivatives of the mass function, m, are given by the rest of the field equations, R3; = 0 and Rs; = 0, a(m + iM) = 3m +iM)Lu, (2.9a) d@m —iM) = 3(m —iM)L,, (2.9b) [P3(m +iM)],, = Pld +2@In P — L,,)\al, (2.9¢) where dG In P—L,)+(@mP—-L,). (2.10) As was said in Kerr (1963), there are two natural choices that can be made to restrict the coordinates and simplify the final results. One is to rescale r so that P = | and L is complex, the other is to take L to be pure imaginary with P #1. chose to do the first since this gives the most concise form for M and the remaining field equations. It also gives the smallest group of permissible coordinate transformations, simplifying the task of finding all possible Killing vectors. The results for this gauge are M =Im(330L), (2.1a) am + iM) = 3(m +iM)Lu, (2.11b) TM = STE ATE a,[m — Re(daL)] = |0,3L/?. (2.114) Since all derivatives of m are given, the commutators were calculated to see whether the field equations were completely integrable. This gives m as a function of the higher derivatives of L unless both E,, and Lu, are zero, As stated in Kerr (1963), if these are both zero then there is a coordinate system in which P and L are independent of u and m = cu + A(¢, £), where c is a real constant. If it is zero 13 This expression for M was first published by Robinson et al. (1969). The corresponding expression in Kerr (1963) is for the gauge when P = 1. The same is true for Equation (2.9c). The Kerr and Kerr-Schild metrics 49 then the metric is independent of u and therefore stationary. The field equations are V[V(n P)] =c, V = P?a*/agae, (2.12a) M =2EV(InP)+ VE, m=cut A(E,E), (2.12b) cL=(A+iM), > VM =c¥. (2.12c) We shall call these metrics quasi-stationary. In Kerr (1963) I stated that the solutions of these equations include the Kerr metric (for which c = 0). This is true but it is not how this solution was found. Furthermore, in spite of what many believe, its construction did not use the Kerr— Schild ansatz. 2.3 Symmetries in algebraically special spaces The field equations, (2.9) or (2.11), are so complicated that some extra assumptions were needed to reduce them to a more manageable form. My next step in the hunt for physically interesting solutions was to assume that the metric is stationary. Fortunately, I had some tricks that allowed me to find all possible Killing vectors without actually solving Killing’s equation. The key observation is that any Killing vector generates a 1-parameter group which must be a subgroup of the group C of coordinate transformations that preserve all imposed coordinate conditions. Suppose that {x**, «o;} is another set of coordinates and tetrad vectors that satisfy the conditions imposed in the previous section. If we restrict our coordinates to those that satisfy P = 1 then C is the group of transformations x —> x* for which _— o! = (IO_1/,)o', uw" = [elu + Se, BD), aa and the basic metric functions, L* and m*, are given by L = (1 c|/ ML — Se — (Pee / Pe Nu + SC, OI, (2.13a) m* = |%,|3m. (2.13b) Let S be the identity component of the group of symmetries of a space. If we interpret these as coordinate transformations, rather than point transformations, then it is the set of transformations x > x* for which Sup (®") = Sap(x*).50 Roy P. Kerr It can be shown that S is precisely the subgroup of C for which m*(x*) = (x), L*Q@") = LGr*).'4 Suppose that x —> x*(x, t) is a I-parameter group of motions, ot = We), ut = [Wel + TG, 832)), rs |wel ir Since x*(x; 0) = x, the initial values of y and T are WE0) =F, TE,F;0) =0. ‘The corresponding infinitesimal transformation, K = K"0/dx"" is (=| Kts 0 at Ji-o _ [ae n-f% woz]. "60-[] then the infinitesimal transformation is Tf we define K = ad, + a; + Re(ae)(ud, — r8,) + Vu. (2.14) Replacing ©(¢) with (¢;) in (2.13), differentiating this w.r.t. 1, and using the initial values for y and T, K is a Killing vector if and only if | Vest borger + KE Sete — Gig) = 0, The Kerr and Kerr—Schild metrics 51 already simple without the (®, S) transformation being used. There are therefore two canonical types for K, typel: K= Vay, or type?: K= a, +i. (2.15) ‘These are asymptotically timelike and spacelike, respectively. 2.4 Stationary solutions The obvious and easiest way to simplify the field equations was to assume that the metric was stationary. The type 2 Killing vectors are asymptotically spacelike and so I assumed that € had a type 1 Killing vector K = Va,. The coordinates used in the last section assumed that P = 1. If we transform to the more general coordinates where P # 1, using an A-transformation (2.4c) with associated change in the (r, w) variables, we get K=Ak, U=ASLr = Ar w= Au, K=Va,=VAde =a ifVA=1. The metric can therefore be assumed independent of w, but P may not be constant. The basic functions, L, P and m are functions of (¢, &) alone, and the metric simplifies to Km +3Re(a;)m = 0. The transformation rules for K under an element (®, S) of C are aw=Oa, V= {\,|[V — Re(a;)S + KS]. Since a is itself analytic, if « # 0 for a particular Killing vector then, ® can be chosen so that a* = 1,!5 and then S so that V* = 0. If = 0 then so is @*, and K is "3 Note that cis implies that all derivatives of these functions are also invariant, and so gop itself is invariant. 'S Or any other analytic function of ¢ that one chooses. ds? = ds? + 2mr/(r? + EK, (2.16) where the “base” metric is (dso) = 2(r? + D?)P~? dg dF — 2lok, (2.17a) — 7 oa Mx _] 147 =a Fit FOE OTEK =a ee | 2470} Although the base metric is flat for Schwarzschild it is not so in general. £, K and M are all functions of the derivatives of L and P, = = P*lm(L;), K =2V?InP, ro (2.18) M=EK+VE, V? = Pad, The mass function, m, and M are conjugate harmonic functions, m, = —iM,, me = +iM;, (2.19)52 Roy P. Kerr and the remaining field equations are WK VM =0. (2.20) If m is a particular solution of these equations then so is m + mg where mo is an arbitrary constant. The most general situation where the metric splits in this way is when P, L and M are all independent of u but m = cu + A(¢, ). The field equations for these are given in (2.12) and in Kerr (1963). We can state this as a theorem. Theorem 2.2 if ds? is any stationary (diverging) algebraically special metric, or more generally a solution of (2.12), then so is 2mor dsg + =. et ae where mo is an arbitrary constant. These are the most general diverging alge- braically special spaces that split in this way. These are “generalized Kerr-Schild” metrics with base spaces ds? that are not necessarily flat. The field equations for stationary AS metrics are certainly simpler than the original ones, (2.9), but they are still PDEs, not ODEs, and their complete solution is unknown. 2.5 Axial symmetry We are getting close to Kerr. At this point I assumed that the metric was axially symmetric as well as stationary. I should have revisited the Killing equations to look for any Killing vector (KV) that commutes with the stationary one, 3,. However, I knew that it could not also be type 1'8 and therefore it must be type 2. It seemed The Kerr and Kerr-Schild metrics 53 where F is analytic. There is only one known solution of this equation for F not a constant, P= Po+e), VinP= —Z rue +), (2.21) where Po is an arbitrary constant. The mass function m is then constant and the last field equation, V?M = 0, can be solved for L. The final metric is given in Kerr and Debney (1970), Equation (6.14), but it is not worth writing out here since it is not asymptotically flat. If € is to be the metric for a localized physical source then the null congru- ence should be asymptotically the same as Schwarzschild. F(¢) must be regular everywhere, including at infinity, and must therefore be constant, 3K = PP ez — PePz = Ro =+PZ, (say). (2.22) As was shown in Kerr and Debney (1970), the appropriate Killing equations for aK that commutes with K are K=ak +a0;, a=a(¢), KL=-al, KE=0, (2.23) KP =Re(a)P, Km =0. Ido not remember the choice I made for the canonical form for kK in 1963, but it was probably 3). The choice in Kerr and Debney (1969) was. a=it, > Kile, £0), —fairly-clear-that-it-could-be-transformed-to-the-canonical form i(d¢ = ogy = 8 where £ = x + iy) or equivalently i(¢4, — £4;) (= a in polar coordinates where ¢ = Re’®), I was getting quite eager at this point so I decided to just assume such a KV and see what turned up.'7 From the first equation in (2.20), the curvature V2(In P) of the 2-metric P~ de dé is a harmonic function, V? In P = P'(In P).¢¢ = FC) + FQ), { Hit were it would be parallel to the stationary KV and therefore a constant multiple of it "” All possible symmetry groups were found for diverging AS spaces in George C. Debney’s Ph.D. thesis. My 1963 expectations were confirmed there. Kf=0 = £(G.E)= g(Z), where Z =f. Now Re(a,-) = 0, and therefore KP=0, > P=P(2), and therefore K is given by 7K = PUInP)zg= PPyg—PpPp=Zo = P=Z+Zo54 Roy P. Kerr after a &(¢)-coordinate transformation. Note that the form of the metric is invariant under the transformation r=Aor*, u=Ag'ut, = Aog*, (2.24 Zo = Ag?Zh, mo = a <3 nt ‘0 Mor where Ao is a constant, and therefore Zo is a disposable constant. We will choose it later. The general solution of (2.23) for L and © is L=itP?B(Z), © =ZB-(1— ZoP™')B, where B’ = dB/dZ. The complex “mass”, m + iM, is an analytic function of ¢ from (2.19), and is also a function of Z from (2.23). It must therefore be a constant, m+iM = po = mo+iMo. Substituting this into (2.18), the equation for £, IK+V2=M=M, — P?[ZE" 4+ E'] + 2ZE = Mo. The complete solution to this is r= a+z 201 a +C)lnZ], Z+Zo where (Co, a, C)) are arbitrary constants. This gives a four-parameter metric when these known functions are substituted into (2.16) and (2.17). However, if C is non-zero then the final metric is singular at r = 0. It was therefore omitted in Kerr (1963). The “imaginary mass” is then M = 2ZoCo and so Co is a multiple of the NUT parameter. It was oon in 1963 that the metric cannot be asymptotically at if this is non-zero also omitted. The only constants ret ‘mo, a and Zg. When a is zero and Zg is positive the metric is that of Schwarzschild. It was not clear that the metric would be physically interesting when a # 0, but if it had not been so then this whole exercise would have been futile. The curvature of the 2-metric, 2P~*d¢ dZ, had to have the same sign as Schwarzschild if the metric was to be asymptotically flat, and so Zy = +P}. The basic functions in the metric then became Zo=P3, P=t&+Pj, m=m, M=0, ° 0 55 Zo L=iagP?, 2= = So4+Zo the Kerr and Kerr—schiid metrics awe ‘The metric was originally published in spherical polar coordinates. The rela- tionship between these and the (£, £) coordinates is = Pocot$ e'#. At this point we choose Apo in the transformation (2.24) so that 2Pe=1, > k=du+asin’Odd Recalling the split of (2.16) and (2.17), ds? = ds2 + 2mr/(r? + acos?0)k? (2.25) where m = mo, a constant, and ds2 = (r? + a?cos?6)(d6? + sin?@ dp?) i 2.26) — (2dr + du — asin?@ dp)(du + a sin?6 dd). oy This is the original form of Kerr (1963), except that u has been replaced by —u to agree with current conventions, and a has been replaced with its negative. ae Having found this fairly simple metric, I was desperate to see whether it was ro- tating. Fortunately, knew that the curvature of the base metric, ds, was zero, and so it was only necessary to find coordinates where this was manifestly Minkowskian. These were (r+ iajel?sing =x+iy, rcos@=z, rt+u=—t. This gives the Kerr—Schild form of the metric, oo dx? 4 dy? + dz? — dt? + =——5 (2.21) where the surfaces of constant r are confocal ellipsoids of revolution about the c-axis, (2.28) Asymptotically, of course, r is just the distance from the origin in the Minkowskian coordinates, and the metric is clearly asymptotically flat. '8 We will see why later56 Roy P. Kerr Angular momentum ‘After the metric had been put into its Kerr—Schild form I went to Alfred Schild and told him I was about to calculate the angular momentum of the central body. He was just as excited as I was and so he joined me in my office while I computed. We were both heavy smokers at that time, so you can imagine what the atmosphere was like, Alfred puffing away at his pipe in an old arm chair, and myself chain-smoking, cigarettes at my desk. ‘The Kerr-Schild form is particularly suitable for calculating the physical pa- rameters of the solution. My PhD thesis at Cambridge was entitled “Equations of Motion in General Relativity”. It had been claimed previously in the literature that it was only necessary to satisfy the momentum equations for singular particles to be able to integrate the ETH quasi-static approximation equations at each order. One thing shown in my thesis was the physically obvious fact that the angular momentum equations were equally important. Some of this was published in Kerr (1958) and (1960). Because of this previous work I was well aware how to calculate the angular momentum in this new metric. It was first expanded in powers of R7!, where R = Euclidean distance, + y? +2? is the usual ds? ds? + 2.da,, dx". If we choose ydx) > h om aydx" = ~Fe dy 4a —— Qdajede= din EF (dy yaa) dR - then the approximation in (2.29) simplifies to ds? dx? 4 dy? +e? — a? 4 dr + aR R on (2.30) ~ py ody = ydx)dt + O(R™). ‘The leading terms in the linear approximation for the gravitational field around a rotating body were well known (for instance, see Papapetrou (1974) or Kerr (1960). The contribution from the angular momentum vector, J, is AR™ ep Six! dx* dt. The Kerr and Kerr—Schild metrics 57 A comparison of the last two equations showed that the physical parameters were!? Mass=m, J = (0,0,ma). When I turned to Alfred Schild, who was still sitting in the armchair smoking away, and said “Its rotating!” he was even more excited than I was. I do not remember how we celebrated, but celebrate we did! Robert Boyer subsequently calculated the angular momentum by comparing the known Lense~Thirring results for frame dragging around a rotating object in linearized relativity with the frame dragging for a circular orbit in a Kerr metric. This was a very obtuse way of calculating the angular momentum since the ap- proximation (2.30) was the basis for the calculations by Lense and Thirring, but it did show that the sign was wrong in the original paper! 2.6 Singularities and topology ‘The first Texas Symposium on Relativistic Astrophysics was held in Dallas De- cember 16-18, 1963, just a few months after the discovery of the rotating solution. It was organized by a combined group of Relativists and Astrophysicists and its purpose was to try to find an cxplanation for the newly discovered quasars. The source 3C 273B had been observed in March and was thought to be about a million million times brighter than the sun. Ithad been long known that a spherically symmetric body could collapse inside an event horizon to become what was to be later called a black hole by John Wheeler. However, the Schwarzschild solution was non-rotating and it was not known what would happen if rotation was present. I presented a paper called “Gravitational collapse and rotation” in which I outlined the Kerr solution and said that the topological and physical properties of the event horizon may change Kerr was the only possible stationary solution for such a rotating black hole and so I discussed it as an example of such an object. Although this was not pointed out in the original letter, Kerr (1963), the geometry of Kerr is even more complicated than the Kruskal extension of Schwarzschild. ‘The metric is everywhere non-singular, except on the ring 0 x+y? ° Unfortunately, I was rather hurried when performing this calculation and got the sign wrong, This is why the sign of the parameter ain Kerr (1963) is different to that in all other publications, including this one. This way of calculating J was explained at the First Texas Symposium at the end of 1963, see Kerr (1965), but I did not ccheck the sign at that time.58 Roy P. Kerr ‘The Weyl scalar, Rabea R24 —> oo near these points and so they are true singular- ities, not just coordinate ones. Furthermore, this ring behaves like a branch point in the complex plane. If one travels on a closed curve that threads the ring the initial and final metrics are different: r changes sign. Equation (2.28) has one non-negative root for r?, and therefore two real roots, rs, for r. These coincide where r? = 0, ie. on the dise D bounded by the ring singularity ety and vice versa. In E1, r > O and the mass is positive at infinity; in E,r < 0 and the mass is negative. The metric is then everywhere analytic except on the ring. It was trivially obvious to everyone that if the parameter a is very much less than m then the Schwarzschild event horizon at r = 2m will be modified slightly but cannot disappear. For instance, the light cones at r = m in Kerr all point inwards for small a. Before I went to the meeting I had calculated the behaviour of the time- like geodesics up and down the axis of rotation and found that horizons occurred at the points on the axis in E; where r?— mr +a? but that there are no horizons in Ez where the mass is negative. In effect, the ring singularity is “naked” in that sheet. I made a rather hurried calculation of the two event horizons in E; before I went to the Dallas Symposium and claimed incorrectly there, Kerr (1965), that the equations for them were the two roots of myc aa ee whereas z? should be replaced by r? in this and the true equation is P—Imrt+a= T attempted to calculate this using inappropriate coordinates and assuming that the equation would be: “y(r, z) is null for some function of both r and z”. I did not realize that this function depended only on r. The Kerr-Schild coordinates are a generalization of the Eddington—Finkel- stein coordinates for Schwarzschild. For the latter future-pointing radial geodesics are well behaved but not those travelling to the past. Kruskal coordinates were The Kerr and Kerr—Schild metrics 59 designed to handle both. Similarly for Kerr, the coordinates given here only handle ingoing curves. This metric is known to be type D and therefore it has another set of Debever-Penrose vectors and an associated coordinate system for which the outgoing geodesics are well behaved, but not the ingoing ones. ‘This metric consists of three blocks, outside the outer event horizon, between the two horizons and within the inner horizon (at least for m < a, which is probably true for all existing black holes). Just as Kruskal extends Schwarzschild by adding extra blocks, Boyer and Lindquist (1967) and Carter (1968) independently showed that the maximal extension of Kerr has a similar proliferation of blocks. However, the Kruskal extension has no application to a real black hole formed by the collapse of a spherically symmetric body and the same is true for Kerr. In fact, even what I call E2, the sheet where the mass is negative, is probably irrelevant for the final state of a collapsing rotating object. Ever since this metric was first discovered people have tried to fit an interior solution, One morning during the summer of 1964 Ray Sachs and myself decided that we would try to do so. Since the original form is useless and the Kerr— Schild form was clearly inappropriate we started by transforming to the canonical coordinates for stationary axisymmetric solutions. In Papapetrou (1966) there is a very elegant treatment of stationary axisymmetric Einstein spaces, He shows that if there is a real non-singular axis of rotation then the coordinates can be chosen so that there is only one off-diagonal component of the metric, We call such a metric quasi-diagonalizeable. All cross terms between {dr, d9} and {dz, df} can be eliminated by transformations of the type di =dt+Adr+Bd0, dd’ =de+Cdr + Dds. where the coefficients can be found algebraically. Papapetrou proved that di’ and dq’ are perfect differentials if the axis is regular? .d the coefficients A... D and transformed the metric to the 20 igshown in Weirand Kerr (1977) that if the metrics also algebraically special thn tis quasi-diagonaizeable precisely when i is type D. These metrics are the NUT parameter generalization of Kerr