Stationary generalized Kerr–Schild spacetimes Carlos F. Sopuerta ∗ Departament de F´ısica Fonamental, Universitat de Barcelona 8 9 Diagonal 647, 08028 Barcelona, Spain. 9 1 February 7, 2008 n a J 9 2 Abstract 1 v Inthispaperwehaveappliedthegeneralized Kerr–Schildtransformationfinding 3 anewfamilyofstationaryperfect–fluidsolutionsoftheEinsteinfieldequations. The 0 1 procedure used combines some well–known techniques of null and timelike vector 1 fields, from which some properties of the solutions are studied in a coordinate–free 0 8 way. These spacetimes are algebraically special being their Petrov types II and 9 D. This family includes all the classical vacuum Kerr–Schild spacetimes, excepting / c the plane–fronted gravitational waves, and some other interesting solutions as, for q instance, the Kerr metric in the background of the Einstein Universe. However, the - r family is much more general and depends on an arbitrary function of one variable. g : v i X r a PACS Numbers: 04.20.Jb ∗Also at Laboratori de F´ısica Matema`tica. Societat Catalana de F´ısica. IEC. Barcelona. E–mail address: carlos@hermes.ffn.ub.es 1 I Introduction Since the discovery of the Kerr metric [1] until now, the Kerr–Schild Ansatz [2]- [4] has been a powerful tool to find new explicit solutions of the Einstein equations. Roughly speaking, this technique consists in generating new solutions from the Minkowski spacetime and its null geodesic vector fields, and it can be applied to several types of energy–momentum content: vacuum, electromagnetic field, pure radiation, etc. With regard to the results obtained, we should note that some of the solutions found have been of crucial importance in general relativity and they have received too much attention, specially in relation to the description of black holes or radiating bodies. Well known examples are the Kerr–Newman family of spacetimes, the Vaidya metric, etc. A complete review of the classical Kerr–Schild transformation can be found in reference [5] (chapter 28). Some generalizations of this generation technique have appeared, so that they over- come some of the limitations that the classical Kerr–Schild transformation had. In par- ticular, the energy–momentum tensor of the classical Kerr-Schild spacetimes generated have the null vector field used in the transformation as a null eigenvector and hence, they cannot be perfect–fluid solutions. A generalization which allows to find perfect–fluid spacetimes isthegeneralized Kerr–Schild(GKShereafter)transformation, which hasbeen studied widely in [6]–[15]. The GKS transformation is a generation technique in which the metrics of the initial and the final spacetimes are related by the following expression g˜ = g + 2 ℓ ℓ , (1) ab ab a b H where the starting metric g , which we shall call seed metric, is any metric, ~ℓ is a null ab ~ vector field for the metric g , and is a scalar field. In the case in which ℓ is geodesic, ab H an illuminating result was proved in [6, 7] (see also [8, 9, 10]). It states that the energy– ~ ~ momentum tensor of the GKS spacetimes has ℓ as an eigenvector if and only if ℓ is an eigenvector for the energy–momentum tensor of the seed spacetimes. Therefore, if we are interested infinding perfect–fluidspacetimes, we shouldstart withaseed spacetime whose energy–momentum tensor does not have null eigenvectors. In this line, a particularization of the GKS transformation for the search of perfect–fluid solutions was developed in [6, 7], where the seed spacetime was taking to be conformally–flat. From the application of this technique to several cases, new perfect–fluid solutions of Einstein equations have been obtained. In some of these applications the solutions represent inhomogeneous cosmolog- ical models while in others, they are static and stationary spacetimes (see [6]–[9], [12] and [16]). The subject of this paper is to apply the GKS transformation, in combination with techniques of null and timelike vector fields, in order to find new perfect–fluid solutions of Einstein’s equations. The result found is a wide new family of stationary perfect–fluid spacetimes. 2 In the development of this work we follow the spirit of the classical Kerr–Schild trans- formation, where the seed spacetime was very simple (the Minkowski spacetime) and the ~ richness of the transformation lay on the choice of the null vector fields ℓ used. In this sense, we apply the GKS transformation to the FLRW models with constant scale factor and taking their most general shear–free geodesic null (SFGN hereafter) vector field as ~ the vector field ℓ. These objects, which are the basic ingredients to carry out the GKS transformation, are studied in section II. In that situation, we study the perfect–fluid Einstein field equations for the GKS metrics, from which we obtain the form of the GKS energy–momentum tensor (energy density, pressure and fluid velocity) and some partial differential equations for the function (section III). Then, through the study of the H kinematical quantities of the GKS fluid velocity we impose some rectrictions in order to find stationary perfect–fluid solutions. Moreover, these restrictions together with the choice of the null basis will allow us to integrate the equations for without the explicit H use of a coordinatesystem. In addition, we deduce the Petrov type of the GKSspacetimes as well as expressions for the non–zero kinematical quantities of the fluid velocity (section IV). In section V, we find the explicit form of the solutions. As far as we know, this family was previously unknown. It is noted that it contains all the classical vacuum Kerr–Schild spacetimes, excepting the plane–fronted gravitational waves, and hence the Kerr metric [1] is included too. It is also pointed out that the generalization of the Kerr metric in the Einstein background, due to Vaidya [16], is also included. However, the family of stationary spacetimes found is much more general and depends on an arbitrary function of one variable. On the other hand, the efficiency of the procedure is shown in section VI, where the Killing equations are completely solved without using coordinates. It is remarkable that, in general, these spacetimes have only one symmetry. Moreover, we discuss the conditions in which other symmetries can appear and their consequences on the GKS spacetimes. The main equations in this paper are written in the Newman–Penrose formalism [17]. The conventions that we use here are the same of [5] with the only exception that the ~ ~ name for the main vector field of the null basis here is ℓ whereas in [5] it is k. Moreover, throughout this paper we have used units in which 8πG = c = 1. Latin indexes run from 0 to 3. The abbreviation “c.c.” will stand for complex conjugate. II Basic ingredients for the GKS transformation In the classical Kerr-Schild transformation the seed spacetime was the simplest one, the Minkowski spacetime, being the important object the null vector field of the transfor- mation,whichwasthemostgeneralSFGNvectorfieldforMinkowskispacetime. Following this line, we take as seeds the subclass of the FLRW spacetimes with constant scale factor ~ and their most general SFGN vector field as the vector field ℓ for the transformation. 3 In what follows, we give the characterization and the explicit form of the seed space- ~ times g as well as of the SFGN vector fields ℓ. ab The seed metrics As we have said, the seed spacetimes are the FLRW spacetimes with constant scale factor, or equivalently, without expansion. Taking into account the form of the FLRW spacetimes given in appendix A, the line element for the seed spacetimes can be written in the following explicitly conformally–flat form a2 V U 2 ds2 = 4dUdV +4 − dξdξ¯ , (2) (1+εU2)(1+εV2) − 1+ξξ¯!    where a is the constant scale factor anda bar denotes complex conjugation. In the case ε = 0thislineelement correspondstotheMinkowski spacetime indoublenullcoordinates; in the case ε = 1 to the Einstein static Universe and, in the case ε = 1, it corresponds − to a spacetime usually not considered because the energy density is negative, however we consider it as a seed spacetime because it may lead to GKS solutions with good physical properties. These conformally–flat spacetimes have a perfect–fluid matter content and therefore, its energy–momentum tensor is given by T = (̺+p)u u +pg , g uaub = 1, ab a b ab ab − where the energy density ̺ and the pressure p are constant 3ε ̺ = = 3p, a2 − and in addition, the fluid velocity is ∂ ∂ 1 ∂ ∂ ~u = ua = = (1+εU2) +(1+εV2) , ∂xa ∂t 2a ( ∂U ∂V ) dU dV u = u dxa = dt = a + . a − − (1+εU2 1+εV2) It can be checked that ~u is a constant vector field, that is u = 0. On the other hand, ∇a b the case ε = 0 is Minkowski (̺ = p = 0), and although we can define ~u, it is not a preferred vector field. ~ The shear–free geodesic null vector field ℓ. In analogy with the pioneering works [4] on the classical Kerr–Schild transformation, we restrict ourselves to the class of shear–free geodesic null vector fields. Although the 4 Goldberg–Sachs theorem [18] tells us that such a kind of vector fields may not exist in general spacetimes, it is common knowledge that in the Minkowski spacetime there is a big family of such vector fields and all of them are known explicitly: this result constitutes the so–called Kerr’s theorem (see [5, 19]). Moreover, it is well known that a SFGN vector field for Minkowski is also a SFGN vector field for any conformally–flat spacetime, which means that we have the same large class of such vector fields for the seed spacetimes (2). Here, we are going to construct allof them and after this, we will construct anappropriate null basis associated with them. To begin with, the most general null one–form field for the seed spacetimes (2) can be written in the following form V U ℓ = F dU +Y Y¯ dV + − Y¯ dξ +Y dξ¯ , (3) ¯ ( 1+ξξ ) (cid:16) (cid:17) where F and Y are real and complex arbitrary scalar fields respectively. Then, in order that ℓ be geodesic and shear–free, the complex function Y must be a solution of the following system of non–linear partial differential equations ¯ ¯ (1+ξξ)Y Y = (V U)Y +Y (1+ξY), (4) ,ξ ,V − ¯ (1+ξξ)Y = (V U)Y Y Y (Y ξ), (5) ,ξ¯ ,U − − − where commas stand for partial derivative with respect to the subscript that follows. Now, in order to simplify further calculations, we take a null basis ℓ,k,m,m¯ for the { } ~ ~ seed spacetimes (2) associated with ℓ. We choose this null basis in such a way that ℓ be affinely parametrized and such that the unit timelike vector field ~u can be written in the following way 1 1 ~u = ~ℓ+~k , u = (ℓ+k) . √2 √2 (cid:16) (cid:17) Part of the remaining freedom is used for setting the imaginary part of the spin coefficient ǫ equals to zero. After some calculations, the explicit expressions for a such null basis ℓ,k,m,m¯ are (3) and { } 1+εV2 1+εU2 V U k = F Y Y¯ dU + dV − Y¯ dξ +Y dξ¯ , ( 1+εU2 1+εV2 − 1+ξξ¯ ) (cid:16) (cid:17) 1+εV2 1+εU2 m = F √Y Y¯ dU + dV + −  −s1+εU2 s1+εV2     V −U Y¯ 1+εU2 dξ Y2 1+εV2 dξ¯ , 1+ξξ¯sY s1+εV2 − s1+εU2      5 where the case Y = 0 can be obtained by taking Y/Y¯ = 1, and F is the following function √2a F = − . (1+εU2)+(1+εV2)YY¯ After long but straightforwardcalculations we find that the spin coefficients associated with this null basis satisfy the relations κ = σ = ǫ = π = λ = 0, (6) ¯ τ ν¯ = ρ µ¯ = α+β = γ +γ¯ = 0. − − Furthermore, the Newman–Penrose symbols for the Riemann tensor are Ricci tensor: 1 ε 1 ε Φ = Φ = Φ = 0, Φ = 2Φ = Φ = (̺+p) = , Λ = (̺ 3p) = , 01 02 12 00 11 22 4 2a2 24 − 4a2 Weyl tensor: Ψ = Ψ = Ψ = Ψ = Ψ = 0. 0 1 2 3 4 III The construction of the GKS spacetimes In this section we study the GKS transformation (1) for the seed spacetimes (2) with ~ (3) as the SFGN vector field ℓ. To that end, we use the fact that given a null basis ℓ,k,m,m¯ associated with the seed metric, we can construct a null basis ℓ˜,k˜,m˜,m˜¯ { } { } associated with the GKS metric by taking (see [12, 6]) ℓ˜a = ℓa, ℓ˜ = ℓ , a a k˜a = ka + ℓa, k˜ = k ℓ , (7) a a a H −H m˜a = ma, m˜ = m , a a where, fromnowon, objects withtildeareassociated withtheGKSspacetimes incontrast with objects without tilde which are associated with the seed spacetimes. These relationships between null bases allow us to compute all objects associated with the GKS spacetimes as functions of the same objects associated with the seed spacetimes, and the function and their derivatives. In this sense, we can compute all the Newman– H Penrose symbols for the GKS spacetimes. Taking into account the expressions of the previous section, the result for the components of the Ricci tensor is the following ε Φ˜ = , Φ˜ = Φ˜ = 0, (8) 00 2a2 01 02 6 ε 1 ε 1 1 Φ˜ = + (ρ ρ¯)2 + + (ρ+ρ¯) D , (9) 11 4a2 2 − 2a2 H 4 V − 4 V (cid:20) (cid:21) 1 Φ˜ = [δρ¯+(ρ ρ¯)τ] +ρ¯ + (τ δ ) , (10) 12 − H H 2 V − V ε 1 1 1 Φ˜ = + 2 (ρ+ρ¯) (ρ ρ¯)2 + (ρ+ρ¯) + (ρ+ρ¯) + 22 2a2 2 △ − − H 2 △H 2 V h 1 i ε (τ¯+α) +(τ +α¯) ¯ δ ¯ +δ¯ + 2, (11) U U − 2 U U 2a2 H (cid:16) (cid:17) ε 1 3ε 1 1 Λ˜ = + (ρ+ρ¯)2 +2ρρ¯ (ρ+ρ¯) + D , (12) 4a2 6 − 2a2 H− 4 V 12 V (cid:20) (cid:21) where and are defined by V U D +(ρ+ρ¯) , δ . V ≡ H H U ≡ H In the same way, we can obtain the components of the Weyl tensor. Their expressions are Ψ˜ = 0, Ψ˜ = 0, (13) 0 1 3 1 3Ψ˜ = (ρ ρ¯)2 (ρ ρ¯) D , (14) 2 − − H− 2 − V − 2 V 1 Ψ˜ = (ρ ρ¯) ¯ + (τ¯ δ¯ ), Ψ˜ = 2(τ¯ α) ¯ δ¯ ¯. (15) 3 4 − − U 2 V − V − U − U As we can see directly from these equations, all the GKS spacetimes that we can obtain ~˜ ~ are algebraically special because the null vector field ℓ = ℓ is a multiple null eigenvector of the Weyl tensor. Now, we must study the Einstein field equations for the GKS metrics with a perfect– fluid source. They read as follows G˜ +Cg˜ = T˜ , (16) ab ab ab where G˜ is the Einstein tensor for the GKS metrics, C is the cosmological constant, and ab T˜ is the energy–momentum tensor, which has the form ab T˜ = (̺˜+p˜)u˜ u˜ +p˜g˜ , g˜ u˜au˜b = 1, (17) ab a b ab ab − where ̺˜, p˜ and ~u˜ are the energy density, the pressure and the fluid velocity of the GKS perfect fluid, respectively. In this situation, we project the Einstein equations (16) onto the null basis ℓ˜,k˜,m˜,m˜¯ and then, using (17) we obtain other expressions for the Ricci { } tensor components 1 Φ˜ = (̺˜+p˜)(ℓ˜au˜ )2, (18) 00 a 2 7 1 1 Φ˜ = (̺˜+p˜)(ℓ˜au˜ )(m˜bu˜ ), Φ˜ = (̺˜+p˜)(m˜au˜ )2, (19) 01 a b 02 a 2 2 1 Φ˜ = (̺˜+p˜) (ℓ˜au˜ )(k˜bu˜ )+(m˜au˜ )(m˜¯bu˜ ) , (20) 11 a b a b 4 h i 1 1 Φ˜ = (̺˜+p˜)(k˜au˜ )(m˜au˜ ), Φ˜ = (̺˜+p˜)(k˜au˜ )2, (21) 12 a a 22 a 2 2 1 1 Λ˜ = (̺˜ 3p˜)+ C. (22) 24 − 6 The next step in this process is to compare the expressions for the Ricci tensor (8–12) with the Einstein equations for the GKS metrics (18–22). We do this for the seed metrics with ε = 0, so that ̺˜+p˜= 0, and later we will extend the results for the ε = 0 case. The 6 6 outcome of this comparison is The components, in the null bases ℓ,k,m,m¯ and ℓ˜,k˜,m˜,m˜¯ , for the fluid velocity • { } { } ~u˜ of the GKS perfect fluid: ε ε ε 1 (ℓ˜au˜ )2 = (ℓau˜ )2 = +2 (ρ ρ¯)2 + +(ρ+ρ¯) D − , (23) a a 2a2 a2 − 2a2 H V − V (cid:20) (cid:18) (cid:19) (cid:21) k˜au˜ = kau˜ + ℓau˜ = 2(ℓ˜au˜ ) −1 , (24) a a a a H m˜au˜ = mau˜ =h 0. i (25) a a The energy density and the pressure of the GKS perfect fluid: • 3ε ̺˜= C +2(2ρ2 ρρ¯+2ρ¯2) D , (26) a2 − − H− V ε ε p˜= +C +2 3ρρ¯+ +2(ρ+ρ¯) D , (27) −a2 − a2 H V − V (cid:18) (cid:19) Two second order partial differential equations for : • H δ = τ +2ρ¯ +2 [δρ¯+(ρ ρ¯)τ] , (28) V V U − H ε (ρ+ρ¯) = (ρ ρ¯)2 + 2 (ρ+ρ¯) (ρ+ρ¯) + △H − a2 − △ H− V (cid:20) (cid:21) ε δ ¯ +δ¯ 2(τ¯+α) 2(τ +α¯) ¯ +( ) 1 , (29) − U U − U − U a2 W where is given by W = D (ρ+ρ¯) 2(ρ ρ¯)2 W V − V − − H × h i2ε D (ρ+ρ¯) 2(ρ ρ¯)2 (1+ ) . (30) V − V − − H− a2 H (cid:20) (cid:21) As we can see, equation (29) is a non–linear partial differential equation for due to the H term with . W 8 IV Characterization of the GKS spacetimes through the kinematical quantities Until now, the only assumption we have made on the GKS spacetimes is that their energy–momentum tensor must be of the perfect–fluid type. Then, once we have a func- tion solution of the system (28–29), we have the explicit form of the GKS metric as H well as of the matter content variables. However, there is not a systematic way of solving (28–29), specially because (29) is a non–linear equation. For this reason, we are going to introduce some additional assumptions on the GKS spacetimes which can help us to integrate these equations. Of course, it would be advisable to impose conditions which have a physical meaning. In this sense, an interesting way to control these conditions is through the study of the kinematical quantities for the fluid velocity~u˜ of the GKS perfect fluid. From equations (23–25) it follows that ~u˜ lies in the two–planes generated by~ℓ˜and ~k˜. Then, by making the following change of basis ~ℓ˜ L~ = ~ℓ˜, ~k˜ K~ = 1~k˜, m~˜ m~˜ , (31) − → A → A → where 1 2 2(ℓau˜ )2 − , (32) a A ≡ h i we get the following form for ~u˜ 1 ~u˜ = (L~ +K~). √2 Therefore, we can now use the formulas given in the appendix B to obtain the kinematical quantities of ~u˜. To that end, we compute the spin coefficients associated with the null basis L,K,m˜,m˜¯ , which we denote with a hat. The most useful results are { } ρˆ+ρˆ¯= (ρ+ρ¯) , µˆ+µˆ¯ = (ρ+ρ¯)(1 ) 1, ǫˆ+ˆǫ¯= D , − A −H A A γˆ +γˆ¯ = 1D +( + D ) 2, αˆ +βˆ¯ = 1δ¯ , − − − −A H △A H A A A A πˆ +τˆ¯ νˆ κˆ¯ = τ¯+ δ¯ τ¯(1+ ) 2, σˆ = λˆ = 0, − − − H− H A h i ρˆ ρˆ¯= (ρ ρ¯) , µˆ µˆ¯ = (ρ ρ¯)(1+ ) 1, − − − A − − − H A In this situation, we are going to impose conditions which can lead to stationary spacetimes. To that end, we impose the vanishing of the shear and expansion of the fluid velocity ~u˜, that is, ˜ θ = σ˜ = 0. (33) ab 9 These conditions are necessary in order to have ~u˜ proportional to a time–like Killing vector field. Then, from the vanishing of the expansion and the m m¯ –component of the (a b) shear tensor (see appendix B) we have (ρ+ρ¯) 2 1+ = 0, A − H (cid:16) (cid:17) so that, two different cases may appear: ρ+ρ¯= 0 or 2 = 1 . If ρ+ρ¯= 0 the SFGN ~ A −H vector field ℓ is expansion–free. In addition, from the Newman–Penrose equations [17] we obtain that ρ2+ε/(2a2) = 0, which means that only the case ε = 1 is possible. However, further analysis of the equations for this case shows that there are not solutions in this case. Therefore, from now on we will consider that ρ+ρ¯ = 0, and then we must follow 6 through the second possibility 2 = 1 . (34) A −H From this equation and using the expressions (23, 32) we get another differential equation for H ε D = (ρ+ρ¯) +2 (ρ ρ¯)2 + . (35) V V − a2 H (cid:20) (cid:21) If we use this equation, we can see from (30) that = 4(ε2/a2) , and therefore W − H (29) is a linear differential equation for . Moreover, from the remaining components of H equations (33) we have the following two additional conditions on and Y, H (D + ) = 0, (36) △ H τ = 0, (37) respectively. The first is a further differential equation for , while the second one is a H constraint on the possible complex functions Y that we can use to construct the SFGN vector fields ~ℓ. Then, replacing (34) into expressions (23–25) for the fluid velocity ~u˜ we obtain ~u˜ = 1 (1 )~ℓ˜+~k˜ = 1 ~ℓ+~k = 1 ~u, (38) 2(1−H) (cid:20) −H (cid:21) 2(1−H) (cid:16) (cid:17) √1−H q q that is to say, ~u˜ is proportional to ~u. On the other hand, it can be shown that the vector field ~u, which is a Killing vector field for the seed spacetimes (2), is also a Killing vector field for the GKS spacetimes provided that (36) and (37) hold. Thus, ~u˜ is parallel to a Killing vector field. We can see this fact from the following expression ˜ u˜ + ˜ u˜ = √2 (D + ) ℓ ℓ +2 τk m +c.c. = 0, (39) ∇a ′b ∇b ′a △ H a b (a b) H n h i o where u˜ = g˜ ub. Here, we have used (6) and the second equality follows from (36,37). ′a ab Notice that this is not a trivial result because, even though conditions (33) are necessary for~u˜ to be proportional to a Killing, they are not sufficient (we also need the acceleration 10