PHYSICAL REVIEW D VOLUME 9, NUMBER 6 45 APRIL 1974 Rotating cylinders and the possibility of global causality violation * Frank J. Tipler Department of Physics and Astrotomy, University of Maryland, College Park, Maryland 20742 (Gecelved 8 November 1973) 1 1996 van Stocnim sotvad the Basten equations Gyy ae} ‘rapidly rotating infinite cylinder. Iti shown that such fed violates causality, inthe senae: "Scat that it allows closed timelike line to connect any two events in spacetime. Thle suggests hal, 2 finite rotting elinder would also act a8 a Hime machine. Since the work of Hawking and Penrose," it has become accepted that classical general relativity predicts some sort of pathological behavior. How- fever, the exact nature of the pathology is under intense debate at present, primarily because so- lutions to the field equations ean be found which ex- ‘iit vireually any type of bizarre behavior ** It {s thus of utmost importance to know what types of pathologies might be expected to occur in actual physical situations. One of these pathologies is causality violation, and in this paper I shall argue that if we make the assumptions concerning the be~ havior of matter and manifold usual in general rel- ativity, then it should be possible in principle to set up an experiment in which this particular pa~ thology could be observed. Because general relativity is a local theory with no @ priori restrictions on the global topology, causality violation can be Introduced into solutions quite easily by injudicious choices of topology; for example, we could assume that the timelike coor~ dinate in the Minkowski metric is periodic, or we could make wormhole identifications in Reissner= Nordstrim space.* In both of these cases the cau- sality violation takes the form of closed timelike ines (CTL) which are not homotopic to zero, and these need cause no worries since they can be re~ moved by reinterpreting the metric in a covering space (following Carter,* CTL removable by such means will be called trivial—others will be called ‘nontrivial. : In 1949, however, Gtdel* discovered a solution to the fleld equations with nonzero cosmological ‘constant that contained nontrivial CTL. Still, it ‘could be argued that the Gédel solution is without physical significance, since it corresponds to a rotating, stationary cosmology, whereas the actual universe is expanding and apparently nonrotating ‘The low-angular-momentum Kerr field, on the other hand, cannot be claimed to be without physi- cal relevance: It appears to be the unique final state of gravitational collapse," and so Kerr black ‘holes probably exist somewhere, possibly in the center of our galaxy." This field also contains 9 % 81 T yy for the gravitational fold 9 nontrivial CTL, though the region of causality vi- olation is confined within an event horizon; causal- ity violation from this source could never be ob- served by terrestrial physicists." In addition, since the CTL must thread their way through @ re~ gion near the singularity, it is quite possible that matter of a collapsing star will replace this r ion, as matter replaces the past horizon inthe ‘ease of spherical coliapse."® The final Kerr feld with collapsed star could be causally well behaved, so the CTL pathology might still be eliminated from general relativity’s physical solutions. T doubt this, because nontrivial causality viola- tion also occurs in the feld generated by a rapidly rotating infinite cylinder. The field of such a cylinder in which the centif- ugal forces are balanced by gravitational attraction was discovered by van Stockum in 1936."" The metric Is expressed in Weyl-Papapetrou form: dst =H (dr? +424) Ldg? +2Mdodt = Fat*, (1) ‘where z measures distance along the cylinder axis, 7s the radial distance from the axis, o is the an- gle coordinate, and ¢ 1s required tobe timelike at P20, (mw cae, 00, provided H0. van Stockum assumes the Einstein equations Gh --8er4, ae 8p ay a’ where p is the particle mass density. Also & aa de / A. ceonss ds/ ds tant, Tet ,--p 2208 CCopyight © 1974 by The American Physical Society.2204 FRANK (particle paths required to be timelike). ha frame in which the mattor is at rest, the equations give for the interior field Pr peytinar?), padate”, Mear?, Fat, ‘where a is the angular velocity of the cylinder. For r>1/a, the lines y =constant, = constant, z=constant are CTL (in fact, by a theorem due to Carter,? nontrivial CTL ean be found which inter~ sect any two events in the manifold), but one could hope that the causality violation could be elimi nated by requiring the boundary of the eylinder to be at r=R<1/a, Here the interior solution would bbe joined to an exterior solution which would be (hopefully) causally well behaved; indeed, the re- sulting upper bound to the “velocity” aR would ‘equal 1, the speed of light in our units (though the orbits of the particles creating the field are time- ke for all 7). ‘van Stockum has developed a procedure which generates an exterior solution for all aR >0, When OcaR4, (a) Hee Moan, _ Brsin(3a ty) Bsinzscoss * rsin(@ +7) singe _7sin8~y) Reins? (Ge) M F with 3. TIPLER ° y=(4e? Re =1)""In(e/R), petan"*(dat RY 1)! [as in the interior solution, FL +M?=7?, s0 the metric signature is (+++-) for Rsr<=). We see that causality violation is avoided for aR =}, but Carter's theorem tells us that itis possible to connect any two events by nontrivial CTL when az >} ‘There are several objections to be met before ‘this result can be Interpreted physically. First of all, Bs. (3), which van Stockum derived by as~ ‘suming a special functional form for the zy.» ‘might not be the only candidates for the exterior field; it is known, for instance, that the gravita~ tional field (Sa) is sfatic™ in the sense that a “transformation” of the form A,B,C, D constants treateBe, “ of=ct+De will eliminate the gy, component. (Transformation is placed in quotes Since 1 is a periodic coordi~ nate: {'=1'+B2n, Interpreted globally, the new metric covers a manifold with topology x (half plane). We can return to the original topology by talking a covering space, an operation which is not equiva lent to changing a coordinate system. | Fortunately, it 1s easy to prove that (3) are the only possible exterior fields for a rotating infinite tylinder. Levy and Robinson" have shown that in this ease, the Weyl-Papapetrou metric can be ‘written [modulo (4)] in the form dst =e (dt sadg? +89 (dr? +de*) srtemdgt, © where u,a,8 are functions of 7 only. A procedure developed by Davies and Caplan’* and myself al- tows the equations R,, =0 to be integrated; the s0- lations are equivalent to (8). (Details of the ‘uniqueness proof can be found in the Appendix.) ‘Sinee the eausality problems come trom the si- nusoid factors of (3c), we might hope to avoid these factors by “transforming” (8a) via (4) and then attempting to join the Interior field to the “new” (topologically distinct) field. ut the “transformation” (4) will not change the exponents ofr, which for aR > become imaginary—in fact, for aR> 3, (2a) és (Se) with the substitutions « =i8 and 8=iy. "Thus we expect causality violation to occur in the matter-free space surrounding a rapidly rotat~ Ing infinite eylinder, As Thorne"* has emphasized, however, it is risky to claim that the properties ‘of such a eylinder also hold for realistic cylinders.2 ROTATING CYLINDERS AND THE POSSIBILITY OF GLOBAL... 1n addition to the already mentioned static nature of the fleld, there Is the fact that it is not even ssymptotially Miskowskian (especialy when af Stil, the gravitational potential of the cylin der’s Newtonlan analog also diverges at radial in- finity, yet this potential is a good approximation near the surface in the middle of a long but finite feylinder, and if we shrink the rotating cylinder down to a “ring” singularity, we end up with the Kerr field, which also has CTL. These facts suz- gest that there is a region near the surface of a (finite cylinder where g,, becomes negative, im- plying causality violation. . ‘Since #0 for r#0, there are no event horizons around the infinite cylinder. By analogy with the static case,"* I expect this to be true for a finite cylinder; if so, then a timelike line from any event in the universe could enter the region where Gey 18 negative and return to any other event.” ‘in short, general relativity suggests that if we ‘construct a sufficiently large rotating cylinder, wwe create a time machine. {I would like to thank Dr. D. Schmidt for helpful ‘lscussions, and Professor D. R. Brill for reading the manuscript. APPENDIX: PROOF THAT VAN STOCKUN'S EXTERIOR SOLUTIONS (3) ARE THE ONLY POSSIBLE EXTERIOR FIELDS FOR AN INFINITE ROTATING CYLINDER, Davies and Caplan have shown that the field ‘equations &,,=0 for the Levy-Robinson metric (Bq. (6)] reduce to (a1) (a2) (as) ‘We have three coupled equations for three fune~ tons: second order in w, second order ina, first order in k. ‘Thus we expect five arbitrary con~ stants, A general physical solution to the above system will be defined to be a set of functions 2, 4, & in which the five constants are allowed to assume all real values from —« to. [ will show that this general solution Is given by Bas. (3a)~ (8c), ‘equation (A2) can be written af late ralte®) hun (/ri afr =D (ohare i conta Stung so (a we tain at Suppose first that w =0, ‘Then a little manipulation yields u=A(nr)+B, k= A‘lnr)+C, a=D, where A,B,C, D are constants. By the transformation f=t'—ag, g=g’, 2=2', rer!, we discover that except for global topology ‘this golution is just the Woyl solution (9a). Suppose now that 0. It is at this point that Davies and Caplan err; thelr “geseral” solution in fact places implicit restrictions on the value of their constant A. The complete general solution is obtained via the following procedure. Let v=". palor}, 80 that w= —fIn(o), and d/ar = 2u*r(d/ép), which gives dy idy tute 1 0. ae Fe dp 2 uation (AS) becomes pap i) are Bg ( MEG) were or a(es)-* ws Let w=pu, giving du/dp=v'=1'/p= sw /p*. (AS) be~ ae a6 Let t=In(p), d/dp=(1/p\d/dt. (A8) becomes f(8)-20-0 ww Let Q=id=dw/dt, d/dt =(Q¥//dw. (87) become: od (2)-10 js AP, which can be written2208 FRANK J. TIPLER 2 woe w(dwe at}, (as) ‘The next integral depends on the sign choice in (A8). First choose the + sign. Then performing the integration, we obtain 1 jsf wea A) gL ae Geecans |i ge. [he constant of integration (1/A) ln has values from —e to =, though 01/e ia @) and a> {In 6), tere 10 no near combization y= At +B0+ C2 (where, B,C are constants) auch that te {orm 4 is everywhere timelike. $i. Gide, Rev. Mod. Piya. 21, 447 (4 RM, Wald, Pays, Rev. Les 26, 1659 107). $c. W.Mlsnos, Phys. Rev. Lot 28, 994 1072). Porat ve! >m! there are a0 even! horizons and 80 causality violation fs global, Dutt ls not clear that a Star with such high valuos of angular momentum and/or charge would collapse sufflctenty far to wacover the fun FM, fap2 rt Bolna"y 9D), fu=orln(u'rD?), foe *-r8Q2+B0lne"r'D), where E,D, F are constants. These solutions are Identical to (8b), with a suitable choice of con- stants, Suppose now that the integration constant is -A*, We obtain (as) gut Zeosline*y) +l, bate Pro, { inf nw 4) +c] +Deos{ nwt) +c]}. ey 1s determined by the relation FL +A? where A, C, D, F are constants. ‘Thus the general exterior field is given by (3). (ato) Be region where ggg changes sign (600 Ref. 7). Penrose has argued lin Proceedings of the Sixth ‘Texas Sym- postum on Relativistic Astrophysics, 1972 tunpub- lished)| that a naked Kerr singularity would be a good model for a rapldly rotating star which has collapsed into adisk, CTL would be expected when e 0, but ‘one might contend that these occur 8 close to the Singularity fand hence sn regions where we expect general relativity o break down anyway) that they are Without physical aigificance. ‘van Stockum's work shows, however, that CTL are not necessarily assoc. ated with extrome curvature in physically significant situtions 10. W. Wisner, tn Astrophysics and General Relativity, ‘edited by a. Chrétien, S. Deser, ani J. Goldstein (Gordon and Breach, New York, 1969), Vol. i. ‘SW, 4. van Stocium, Proc. R. Soe. Balnb, 57, 195 (1987). Other authors, such as 5. C. Maitra[5. ath, Phys. {Z, 1028 (1966)} have noted that the van Stocinim interior Solution possesses CTI.. "85, Ehlers and W. Kundt, in Gravitaion: A Introdue~ tow to Curvent Research, edited by L. Witten (Wiley, New York, 1962) p. 84. ‘SH, Levy and W. J. Robinson, Pree. Camb. Philos. Soe, 60, 279 (1984). 44H, Davies and 7. A. Caplan, Proe. Camb. Philos. Soe. ‘88, 825 (071), B. Prehland, Commun, Math. Phys 35, 127 aT), K.'8, Thome, Comments Astrophys. Space Phys. 2, st 1970, “yr Ternel, Nature 216, 148 (1967); 216, 512 (960. “Agsuming, of course, that the eylinder has existed ‘or alltime. If it is rested, then thi statement will have to be qualified somewhat, but observable causality violation will still ocean.