Integrability of Five Dimensional Minimal Supergravity and Charged Rotating Black Holes
aa r X i v : . [ h e p - t h ] D ec ArXiv: 0912.3199 [hep-th]DCPT-09/87ULB-TH/09-43
Integrability of Five Dimensional Minimal Supergravity and ChargedRotating Black Holes
Pau Figueras (cid:7) , Ella Jamsin ♦ , Jorge V. Rocha ♯ , and Amitabh Virmani ♦(cid:7) Centre for Particle Theory, Department of Mathematical SciencesUniversity of Durham, South Road, Durham, DH1 3LE, UK [email protected] ♦ Physique Th´eorique et Math´ematiqueUniversit´e Libre de BruxellesandInternational Solvay InstitutesCampus Plaine C.P. 231, B-1050 Bruxelles, Belgium ejamsin, [email protected] ♯ Centro Multidisciplinar de Astrof´ısica - CENTRA,Dept. de F´ısica, Instituto Superior T´ecnico,Av. Rovisco Pais 1, 1049-001 Lisboa, Portugal [email protected]
Abstract
We explore the integrability of five-dimensional minimal supergravity in the pres-ence of three commuting Killing vectors. We argue that to see the integrabilitystructure of the theory one necessarily has to perform an Ehlers reduction to twodimensions. A direct dimensional reduction to two dimensions does not allow us tosee the integrability of the theory in an easy way. This situation is in contrast withvacuum five-dimensional gravity. We derive the Belinski-Zakharov (BZ) Lax pair forminimal supergravity based on a symmetric 7 × G / ( SL (2 , R ) × SL (2 , R )). We elucidate the relationship between ourBZ Lax pair and the group theoretic Lax pair previously known in the literature.The BZ Lax pair allows us to generalize the well-known BZ dressing method to five-dimensional minimal supergravity. We show that the action of the three-dimensionalhidden symmetry transformations on the BZ dressing method is simply the groupaction on the BZ vectors. As an illustration of our formalism, we obtain the doublyspinning five-dimensional Myers-Perry black hole by applying solitonic transforma-tions on the Schwarzschild black hole. We also derive the Cvetiˇc-Youm black hole byapplying solitonic transformations on the Reissner-Nordstr¨om black hole. ontents ˇc -Youm solution in canonical coordinates 30B Generalities on G
33C Representation of g ( ) and coset representative 34D Invariant 3-form and the basis of octonions for our representation 36 The presence of D − D -dimensional gravity and supergravity theories into the form of three-dimensional non-linear sigma models coupled to three-dimensional gravity. The construction works fora large number of theories, ranging from the simplest case of four dimensional vacuumgravity giving rise to the SL (2 , R ) /SO (2) coset model to eleven dimensional supergravitygiving rise to an E /SO (16) coset model. A classification of four dimensional theoriesleading to three-dimensional coset models was given in [1]. The higher dimensional originof these coset models was explored in [2]. 1urther reduction on another commuting Killing direction gives a two-dimensionaldilaton gravity coupled to the corresponding non-linear sigma model now living in twodimensions [1]. The non-linear sigma model schematically represents the metric on theKilling directions of the higher dimensional spacetime and the matter sector of the higherdimensional theory. This sigma structure plays a central role in the proofs of the celebratedblack hole uniqueness theorems in four and five spacetime dimensions [3, 4, 5, 6, 7, 8, 9,10, 11, 12] (see also [13, 14]), and their recent generalisations to minimal supergravity infive dimensions [10, 15, 16]. It is well known in the literature [17, 18, 19, 20] (see [21] fora comprehensive review and further references) that these two-dimensional coset modelsare classically completely integrable. The underlying group theoretic structure is a crucialingredient in establishing the integrability of the dimensionally reduced models. Except ina few isolated cases, the integrability of these models has not yet been used as a solutiongenerating technique. The notable exceptions are vacuum gravity in various dimensions[17] and four dimensional Einstein-Maxwell theory [22].For five-dimensional vacuum gravity this line of investigation has led to an impressiveprogress in our understanding of stationary black holes with two rotational Killing vectors.It is expected that such an investigation for various supergravity theories will help us betterunderstand the spectrum of black hole solutions in these theories. Furthermore, it is likelythat such an investigation will let us discover novel charged black hole solutions. In thispaper we take steps in this direction. Although our considerations are very general, andapply to any gravity theory that upon dimensional reduction gives rise to a coset model,for concreteness, we concentrate on the case of five-dimensional minimal supergravity.A motivation behind focusing on five-dimensional minimal supergravity comes from thediscovery of black rings [23], and their subsequent supersymmetric generalizations [24, 25](see [26, 27] for reviews and further references). A non-supersymmetric non-extremal fiveparameter family of black ring solutions characterized by the mass, two angular momenta,electric charge, and dipole charge is conjectured to exist in minimal supergravity [28].At present, though, all known smooth black rings have no more than three independentparameters [25, 28] . The solution of [28] does not admit any smooth supersymmetric limitto the BPS black ring [25]. It is likely that the integrability of minimal supergravity andthe related inverse scattering technique explored in this paper will allow us to construct themost general black ring that will describe thermal excitations above the supersymmetricring.Another motivation for exploring integrability of supergravity theories comes from thefuzzball proposal. According to this proposal, a black hole geometry is a coarse graineddescription of its microstates. Some of theses microstates can be identified with smoothhorizonless geometries with the same asymptotic charges as the black hole (see [30] forreviews). The fuzzball proposal has been mostly explored for BPS black holes, and largeclasses of smooth geometries corresponding to microstates of certain supersymmetric blackholes have been constructed. However, for most black holes a generic microstate need notadmit a supergravity description [31].The situation for non-extremal black holes is much less developed. If the fuzzballproposal were to apply to non-extremal black holes as well, one would need to con-struct smooth horizonless non-BPS geometries with the same asymptotic charges as a In vacuum, the doubly spinning black ring solution of [29] also has three independent parameters. • We derive the Belinski-Zakharov (BZ) Lax pair for minimal supergravity based on asymmetric 7 × G / ( SL (2 , R ) × SL (2 , R )).Our coset construction is largely based on the one used in [38], but it differs in oneimportant aspect, that the coset representative matrix M is symmetric — which isnot the case in [38]. In [38] the matrix M is symmetric under generalized transpo-sition, but not under the usual transpose. References [39, 40] gave another cosetconstruction for the coset G / ( SL (2 , R ) × SL (2 , R )) where the matrix M is alsosymmetric. Our matrix M shares several of the properties of the matrix M of [39, 40]though the two constructions are different. Refs [39, 40] use different field variablesand a different basis for the representation of g ( ) than ours. • We elucidate the relationship between our BZ Lax pair and the group theoreticLax pair previously known in the literature [17, 18, 19, 20, 21]. We generalize thewell-known BZ dressing method to five-dimensional minimal supergravity (modulocertain subtleties to be discussed in section 5.2). • We show that the action of the three-dimensional hidden symmetry transformationson the BZ dressing method is simply the group action on the BZ vectors. • As an illustration of our formalism, we obtain the doubly spinning five-dimensionalMyers-Perry black hole, seen as a solution of minimal supergravity, by applyingsolitonic transformations on the Schwarzschild black hole. We also derive the Cvetiˇc-Youm black hole by applying solitonic transformations on the Reissner-Nordstr¨omblack hole. Recall that G is the split real form of G . It is the only real form of G that is relevant for ourpurposes. At the level of Lie algebras, following the standard notation, we denote the split real form of g as g ( ) . We argue that to see the integrability structure of the theory one necessarily hasto perform an Ehlers reduction to two dimensions. An Ehlers reduction is a twostep reduction: first one reduces the theory to three dimensions, dualizes all three-dimensional vectors into scalars, and then further reduces to two dimensions. Adirect dimensional reduction to two dimensions does not allow us to see the inte-grability of the theory in an easy way. This situation is in contrast with vacuumfive-dimensional gravity.The rest of the paper is organized as follows. In section 2 we present the dimensionalreduction of five-dimensional minimal supergravity to two dimensions. The integrability ofthis theory in explored in section 3: in section 3.1 we derive the BZ Lax pair; in section 3.2we show (following [19]) that the BZ Lax pair is equivalent to the one used in [20, 19]. Insection 4 we generalize the BZ construction to minimal supergravity. The intuition behindhow to use this construction is developed in section 4.2. Section 5 contains certain generalresults pertaining to the generalized BZ construction and the hidden G symmetry. Insection 6 we construct rotating black holes using the generalized BZ construction. Theargument that to see the integrability of five-dimensional minimal supergravity one needsto perform an Ehlers reduction is presented in section 7. We close with a discussion ofopen problems in section 8. Various technical details are relegated to appendices. Inappendix A we present the Cvetiˇc-Youm solution in Weyl canonical coordinates. Wecollect some general results for the Lie algebra g ( ) in appendix B. In appendix C wegive the representation of g ( ) that we use and present a construction of a symmetriccoset representative matrix M . Finally, in appendix D we construct the three-form andthe octonion structure constants preserved by our representation of g ( ) . In this section we review the dimensional reduction of minimal five-dimensional supergrav-ity from five to two dimensions. We assume the existence of three mutually commutingKilling vectors, ξ a , a = 1 , ,
3, so that L ξ a g = 0 and L ξ a F = 0 for all a , where L ξ a denotes the Lie derivative along ξ a . We first reduce the theory to three dimensions usingtwo of these Killing vectors, dualize all three-dimensional one-forms into scalars, and thenreduce the resulting three-dimensional theory to two dimensions on the remaining Killingvector. Reference [1] derived a list of three-dimensional symmetric space non-linear sigma-modelsobtained by dimensional reduction from a class of four dimensional gravity theories cou-pled to abelian gauge fields and scalars . In this paper we concentrate on the particularcase of five-dimensional minimal (ungauged) supergravity, although we believe that ourconsiderations can be readily generalized to other theories — most likely to all theoriesconsidered in [1, 41, 2]! A detailed analysis of their group theoretical structure was given in [41], and the higher dimensionalorigin of these coset models was explored in [2]. G /SO (4) coset model. On the other hand, when the reduction is performed over onetimelike and one spacelike Killing direction one obtains three-dimensional Euclidean grav-ity coupled to the G / ( SL (2 , R ) × SL (2 , R )) coset model. In this section, we brieflyreview these dimensional reductions. This presentation follows closely the one given in[38] but contains a little less details.Five-dimensional minimal supergravity is the simplest supersymmetric extension offive-dimensional vacuum gravity. In the bosonic sector, it contains a metric g and agauge potential A whose field strength is F = dA . The Lagrangian has the form ofEinstein-Maxwell theory with a Chern-Simons term: L = R ⋆ − ⋆ F ∧ F + 13 √ F ∧ F ∧ A . (2.1)The reduction to three dimensions of the metric g is performed using the followingansatz: ds = e √ φ + φ ds + ǫ e √ φ − φ ( dz + A ) + ǫ e − √ φ ( dz + χ dz + A ) , (2.2)where the three-dimensional fields, namely the three-dimensional metric g , the two dila-tons φ and φ , the axion χ , and the two Kaluza-Klein one-form potentials A and A ,do not depend on z and z coordinates. One can also think of this reduction as a twostep process. The first step being the reduction from five to four dimensions over z , andthe second being the reduction from four to three dimensions over z . In each step, thereduction can be performed over either a spacelike or a timelike Killing direction. Thesign ǫ i is +1 when the reduction is performed over a spacelike direction, and − χ , A , and A by F (1) , F , and F respectively. They are defined to be, F (1) = dχ , F = d A + A ∧ dχ , F = d A . The reduction ansatz of the five-dimensional gauge potential A is taken to be A = A (1) + χ dz + χ dz , (2.3)where similarly the three-dimensional gauge potential A (1) and the two axions χ and χ are independent of z and z . The associated field strengths F (2) , F and F are definedto be, F = dχ ,F = dχ − χ dχ , (2.4) F (2) = dA (1) − dχ ∧ ( A − χ A ) − dχ ∧ A . χ , χ , and χ dual to the one forms A (1) , A , and A . This can be done byintroducing the dual one-form field strengths G (1)4 , G (1)5 and G (1)6 for the three axions: e − ~α · ~φ ⋆ F (2) ≡ G (1)4 = dχ + 1 √ χ dχ − χ dχ ) ,ǫ e − ~α · ~φ ⋆ F ≡ G (1)5 = dχ − χ dχ + 13 √ χ ( χ dχ − χ dχ ) , (2.5) ǫ e − ~α · ~φ ⋆ F ≡ G (1)6 = dχ − χ dχ + ( χ χ − χ ) dχ + 13 √ − χ χ + χ )( χ dχ − χ dχ ) . In terms of the new variables, φ , φ , χ , . . . , χ , the Lagrangian becomes L = R ⋆ − ⋆ d~φ ∧ d~φ − ǫ ǫ e ~α · ~φ ⋆ dχ ∧ dχ − ǫ e ~α · ~φ ⋆ dχ ∧ dχ − ǫ e ~α · ~φ ⋆ ( dχ − χ dχ ) ∧ ( dχ − χ dχ ) + 12 ǫ t e ~α · ~φ ⋆ G (1)4 ∧ G (1)4 + 12 ǫ ǫ t e ~α · ~φ ⋆ G (1)5 ∧ G (1)5 + 12 ǫ ǫ t e ~α · ~φ ⋆ G (1)6 ∧ G (1)6 , (2.6)where ǫ t denotes the signature of the three-dimensional metric. It appears in this ex-pression because of the relation ⋆ ⋆ ω (1) = ǫ t ω (1) for any one-form ω (1) . The six doublets ~α , . . . , ~α correspond precisely to the six positive roots of the exceptional Lie algebra g ,given in appendix B. One can note from the Lagrangian (2.6) that for each i the axion χ i is associated to the root α i .To summarize, the three-dimensional theory is determined by a three-dimensional met-ric and a set of eight scalar fields: two dilatons φ and φ and six axions χ , . . . , χ . The non-linear σ -model for G / ˜ K It turns out that the Lagrangian (2.6) can be rewritten as L = R ⋆ L scalar , (2.7)where L scalar is the Lagrangian of a non-linear σ -model for the coset G / ˜ K , with anappropriate subgroup ˜ K depending on the signature ǫ , of the reduced dimensions. Wecan write a coset representative V for the coset G / ˜ K in the Borel gauge by exponen-tiating the Cartan and positive root generators of g ( ) with the dilatons and axions ascoefficients. We can make contact with the reduced Lagrangian (2.6) by choosing the cosetrepresentative to be [38] V = e φ h + φ h e χ e e − χ e + χ e e χ e e χ e − χ e , (2.8)where the representation of g ( ) that we use is given in appendix C. For more grouptheoretic details on the construction and properties of the coset representative V we referthe reader to [38]. Next we define the matrix M as M = S T V T η V S , (2.9)6here η and S are constant matrices whose explicit expressions are given in appendix Cfor the choice ǫ = − , ǫ = +1, to which we specialize from now on. (For other choicesof ǫ ’s, the matrix η needs to be adapted.) The matrix M is symmetric by construction.Under global G transformations it transforms as M → M g = ( S − gS ) T M ( S − gS ) for g ∈ G . (2.10)It can be easily checked by an explicit calculation that the scalar part of the reducedLagrangian (2.6) is also given by L scalar = −
18 Tr (cid:0) ⋆ ( M − dM ) ∧ ( M − dM ) (cid:1) . (2.11)We will see below that the matrix M plays the role of the metric on the Killing fieldsin the standard BZ construction. The vacuum truncation of the matrix M (i.e., setting χ = χ = χ = 0) is block diagonal : M = M − M SL(3) , (2.12)where M SL(3) = − e √ φ φ χ e φ √ − χ + e √ φ φ χ χ e φ √ − e φ √ + φ χ − χ + e √ φ φ χ χ e φ √ − e φ χ + e √ φ ( − e φ χ ) e φ √ φ − e φ √ + φ χ − e φ √ + φ χ − e φ √ + φ χ − e φ √ + φ . (2.13)This M SL(3) is identical to the matrix χ used in [43]. Using a different representation of g ( ) , references [39, 40] gave another coset construction for the coset G / ( SL (2 , R ) × SL (2 , R )) where the matrix M is also symmetric and takes the form (2.12) for vacuumgravity. Further properties of the matrix M are given in appendix C.In most references, and in particular in [38] where a similar discussion is presented, thematrix M is rather defined as V ♯ V , where ♯ denotes the generalized transposition, whichdefines the subgroup on which the coset is taken. That expression is related to the one weuse in this paper in an easy way. First, one can see that in our representation V ♯ = η − V T η. (2.14)As a consequence, the two choices of definitions for M are related in the following way,where we keep the notation M for the expression (2.9) M := V ♯ V = η − ( S T ) − M S − . (2.15) The matrix S above was chosen precisely to ensure that this is the case. This block diagonal form simplyrefers to the fact that the seven dimensional representation of G branches into SL (3 , R ) representationsas = ¯3 + + . .2 From three to two dimensions The reduction to two dimensions is performed by dropping all dependence on the thirdvariable z and using the usual ansatz for the metric g µν = (cid:18) ξ ¯ g mn + ρ B m B n ρ B m ρ B n ρ (cid:19) (2.16)where µ, ν are indices in three dimensions, m, n are two-dimensional indices, and ¯ g mn isthe two-dimensional metric. Two comments are in order: • The two-dimensional Kaluza-Klein vector B m can be dropped under some weakconditions (see [44, 45] in the context of vacuum gravity and [10] in the contextof five-dimensional minimal supergravity). This implies that the two-dimensionalspaces orthogonal to the commuting Killing vector fields are integrable submanifolds. • Any two-dimensional metric can be written locally as a conformal factor times theMinkowski metric. As a consequence, one can take ¯ g mn = δ mn (recall that thethree-dimensional space is Euclidean), while absorbing the conformal factor in ξ .Thus, the ansatz simplifies to the diagonal one g µν = (cid:18) ξ δ mn ρ (cid:19) . (2.17)Taking moreover into account that M − ∂ M = 0, the Lagrangian (2.11) becomes [1] L = 2 ξ − ⋆ dρ ∧ dξ − ρ Tr( ⋆ ( M − dM ) ∧ ( M − dM )) (2.18)where now the Hodge dual ⋆ and the differential d are taken over the two-dimensional flatspace.The equation of motion for the field ρ is given by d ⋆ dρ = 0 (2.19)From (2.19) one deduces that ρ is a harmonic function; therefore, one can choose it tobe one of the coordinates on the two-dimensional space, z = ρ . We take the secondcoordinate z to be z = z such that dz = − ⋆ dρ .In coordinates ( ρ, z ) the equations for the matrix field M can be rewritten as d ( ⋆ρ M − dM ) = 0 . (2.20)We find it useful to rewrite this equation explicitly in terms of the two-dimensional vari-ables as ∂ ρ ( ρ ∂ ρ M M − ) + ∂ z ( ρ ∂ z M M − ) = 0 . (2.21)Defining the matrices U and V as follows, U = ρ ( ∂ ρ M ) M − , V = ρ ( ∂ z M ) M − , (2.22)8he equation of motion (2.21) becomes, ∂ ρ U + ∂ z V = 0 . (2.23)Finally, the equations for ξ can be written as the following system [1] ξ − ∂ ρ ξ = 116 ρ Tr( U − V ) , ξ − ∂ z ξ = 18 ρ Tr(
U V ) . (2.24)The equations for ξ satisfy the integrability condition ∂ ρ ∂ z ξ = ∂ z ∂ ρ ξ as a consequence of(2.23). Therefore, once M is known, the function ξ is determined by a line integral, upto an integration constant. Moreover, while reducing the theory on one timelike and tworotational Killing vectors the ρ and z components of the gauge-field can be taken to bezero [10], A ρ = A z = 0 . (2.25)Equations (2.23) and (2.24) are the equations for a completely integrable two-dimensional sigma model. We will comment on the group theoretic structure of thissigma model in some detail in the following sections. In this section we explore the integrability of the theory. In section 3.1 we derive the BZLax pair for five-dimensional minimal supergravity. This formulation of the Lax pair isequivalent to the one used in [20, 19], but it is better suited to generate new solutionsusing the inverse scattering method. In section 3.2 we discuss the interrelation betweenthe Lax pair introduced in section 3.1 and that of [20, 19]. We also briefly discuss thesolution generating technique of [20, 19].
The approach of Belinski and Zakharov [17], well known for vacuum gravity in variousdimensions, can be readily generalized to minimal 5d supergravity using the equationsdeduced in section 2.2. Focusing on vacuum gravity in the presence of D − ds = G ¯ µ ¯ ν dx ¯ µ dx ¯ ν + e ν ( dρ + dz ) , (3.1)where x ¯ µ = z , z , z . Furthermore, without loss of generality we can choose coordinatesso that det G = − ρ . (3.2)Recall, from the previous section, that the coordinate z is defined as the harmonic dual of ρ . Integrations of the three Killing directions yields a two-dimensional non-linear sigmamodel that is completely integrable. For vacuum spacetimes containing a black hole, it has been rigorously proven that these coordinatesare globally defined in the domain of outer communications [13]. See [14] for a generalization to the fourdimensional Einstein-Maxwell theory. ∂ ρ ˜ U + ∂ z ˜ V = 0 , (3.3)with ˜ U = ρ ( ∂ ρ G ) G − , ˜ V = ρ ( ∂ z G ) G − , (3.4)and the second group of equations for ν∂ ρ ν = − ρ + 18 ρ Tr( ˜ U − ˜ V ) , ∂ z ν = 14 ρ Tr( ˜ U ˜ V ) . (3.5)These equations (3.3-3.5) are the starting point of the BZ construction. It is clear thatthey are formally identical to equations (2.22-2.24) for 5d minimal supergravity , the maindifference being that in the latter case the matrix M plays the role of the Killing part G ofthe spacetime metric. Therefore, the Lax pair we are seeking can be immediately derivedfrom the BZ Lax pair, D Ψ = ρ V − λ Uλ + ρ Ψ , D Ψ = ρ U + λ Vλ + ρ Ψ , (3.6)where λ is the spacetime dependent spectral parameter, and Ψ( λ, ρ, z ) is the generatingmatrix such that the matrix M can be extracted from Ψ as M ( ρ, z ) = Ψ(0 , ρ, z ) . (3.7) D and D are the commuting differential operators introduced by BZ: D = ∂ z − λ λ + ρ ∂ λ , D = ∂ ρ + 2 λ ρλ + ρ ∂ λ . (3.8)Indeed, since [ D , D ] = 0 the compatibility conditions of (3.6) are given by[ D , D ]Ψ = 1 λ + ρ n λ ( ∂ ρ U + ∂ z V ) + V + ρ ( ∂ z U + ∂ ρ V ) + [ U, V ] o Ψ = 0 . (3.9)We observe that the term in this expression proportional to λ is just the equation of motion(2.21), while the remaining term corresponds to the integrability condition that followsfrom (2.22). This guarantees that the solution of the system (3.6) yields a solution of theoriginal (non-linear) equation (2.21).An important feature when generalizing the BZ construction to 5d minimal super-gravity is the step in the dimensional reduction that consists in dualizing the 3d vectorsinto scalars. Such a dimensional reduction is called an Ehlers reduction, while when thereduction is performed without dualization it is called Matzner-Misner. Without the du-alization, one cannot see the integrability of 5d minimal supergravity in two dimensions.On the other hand, the dualization is not necessary for vacuum gravity, as well as for someother supergravity theories. These subtleties can be better understood in terms of grouptheory and Dynkin diagrams and are discussed in section 7. Explicitly, equation (3.5) is converted into the form (2.24) if one replaces G by M and e ν by ξ / √ ρ . A somewhat different approach for describing integrability of the two-dimensional cosetmodels was taken in [19, 20]. It has the advantage to make the underlying symmetriesmore transparent. Here we summarize salient aspects of this approach and show that theirLax pair is equivalent to the Lax pair given above. Our presentation follows [19, 20]. Seealso [46, 47].In this approach, in order to write the Lax pair, one first defines a new matrix ˆ V ( γ, x )which depends on a spacetime dependent spectral parameter γ and satisfies ˆ V ( γ = 0 , x ) = V ( x ). Here x denotes collectively the coordinates on the two-dimensional flat base space.The Lax pair is now given by d ˆ V ˆ V − = Q + 1 − γ γ P − γ γ ⋆ P , (3.10)where Q = 12 (cid:2) d V V − + τ ( d V V − ) (cid:3) , P = 12 (cid:2) d V V − − τ ( d V V − ) (cid:3) , (3.11) τ is the involution of the Lie algebra g ( ) associated to the coset G / ( SL (2 , R ) × SL (2 , R )), and ⋆ denotes the Hodge dual on the two-dimensional flat base space.The integrability condition for these equations is equivalent to the equations of motion(2.20), provided that the spectral parameter γ obeys the differential equation(1 − γ ) dγ + 2 γ ⋆ dγ = γ (1 + γ ) ρ dρ . (3.12)This equation can be easily solved and the solution is given by1 γ − γ = 2( w − z ) ρ ⇒ γ ± ( w, x ) = 1 ρ h ( z − w ) ± p ρ + ( z − w ) i , (3.13)where w is a constant, the so-called ‘spacetime independent spectral parameter’. Definingthe matrix X ( γ, x ) = V ( x ) − ˆ V ( γ, x ), we find that the Lax pair can be rewritten as dXX − = − γ γ A − γ γ ⋆ A , (3.14)where A = M − dM . The Lax equations (3.10) are the ones used in [19, 20].11 .3 Equivalence between the Lax pairs To show that these equations are equivalent to the BZ Lax pair, first notice that γ − = − µ/ρ and γ + = − ¯ µ/ρ , where µ = p ρ + ( z − w ) − ( z − w ) , ¯ µ = − ρ /µ . (3.15)These correspond to the ‘solitons’ and ‘anti-solitions’ in the context of the BZ constructionfor vacuum gravity. This motivates us to define [19] a new spectral parameter λ ≡ − ρ γ , (3.16)so that now λ corresponds to either a soliton ( λ − = µ ) or an antisoliton ( λ + = ¯ µ ). Noticethat the new spectral parameter satisfies dλ = 2 λ ρλ + ρ (cid:18) dρ − λρ dz (cid:19) . (3.17)We can rewrite the Lax pair for the matrix X , (3.14), in terms of the new spectral param-eter λ . It reads dX X − = λλ + ρ (cid:2) − λ M − dM + ρ ∗ M − dM (cid:3) . (3.18)Defining a generating matrix field Ψ asΨ( λ, x ) = M ( x ) X ( λ, x ) with Ψ(0 , x ) = M ( x ) , (3.19)we note that Ψ satisfies d Ψ Ψ − = ρλ + ρ [ ρ dM + λ ⋆ dM ] M − . (3.20) d Ψ in this equation can be calculated as follows d Ψ( λ, x ) = d Ψ( λ, x ) (cid:12)(cid:12) λ + ∂ Ψ( λ, x ) ∂λ dλ = (cid:20) ∂ ρ Ψ + 2 λ ρλ + ρ ∂ λ Ψ (cid:21) dρ + (cid:20) ∂ z Ψ − λ λ + ρ ∂ λ Ψ (cid:21) dz = ( D Ψ) dρ + ( D Ψ) dz , (3.21)where we have used (3.17) when going from the first line to the second one. We immediatelysee that D and D are precisely the differential operators introduced by BZ, defined in(3.8). Therefore, the ρ and z components of equation (3.20) are given by D Ψ = ρ U + λ Vλ + ρ Ψ , D Ψ = ρ V − λ Uλ + ρ Ψ , (3.22)where U = ρ ∂ ρ M M − and V = ρ ∂ z M M − , which is precisely the BZ Lax pair (3.6).12 .4 Group theoretic structure and solution generation Although the Lax equations are equivalent in the two formalisms, the explicit constructionsof solutions from a given seed solution of the higher dimensional theory are very different.In the rest of this section, we very briefly discuss the group theoretic structure constructionof a new solution from a given seed solution in the formalism of [19, 20], to which we referthe reader for further details.It is useful to start by recalling how the underlying symmetries appear in three dimen-sions, and how they can be used as a solution generating technique. The situation in twodimensions can then be seen as an infinite dimensional generalization of it.The matrix V constructed in three dimensions is an element of the coset G / ˜ K ,where the local isotropy group ˜ K is ˜ K = SL (2 , R ) × SL (2 , R ). In our construction thematrix V is taken to be in the Borel gauge (also known as the triangular gauge). Thethree-dimensional scalar Lagrangian (2.11) is invariant under global G transformations.The transformation of the matrix V under G is not simply by the right multiplicationof the group element g . This is because in general such a multiplication does not preservethe Borel gauge. In order to restore the Borel gauge a local transformation under anelement of ˜ K is needed: V ( x ) → k ( x ) V ( x ) g, (3.23)where g ∈ G and k ( x ) ∈ ˜ K for each x . The Lagrangian is invariant under thesetransformations. Finding the right compensator k ( x ) is in general a very difficult problem.To get round this difficulty, one uses the defining property of elements k ∈ ˜ K , namely k ♯ k = 1, where k ♯ := η − k T η (see equation (2.14)). Thus, one can see that the matrix M = V ♯ V transforms as M → g ♯ M g. Similarly, the matrix M defined in (2.9) transformsas M → ˜ g T M ˜ g , where ˜ g = S − gS . The matrices M or M easily allow us to constructnew solutions starting from a seed solution.Some of these considerations generalize to the theory obtained after reduction to twodimensions, where one now defines a matrix ˆ V ( x, γ ) that depends on a complex parameter γ and is such that ˆ V ( x, γ = 0) = V ( x ). This matrix is therefore an element of the loopextension of G , which we denote G +2(2) and global transformations under elementsof this infinite dimensional group should lead to new solutions of the same theory. Ageneralization of the triangular gauge is obtained by imposing that ˆ V ( x, γ ) is regular at γ = 0, which is convenient as this limit defines V ( x ). As in three dimensions, it is thereforenecessary to perform a compensating transformation, this time under an element ˆ k ( x, γ )such that ˆ k ♯ (cid:16) x, − γ (cid:17) ˆ k ( x, γ ) = . This defines a subgroup of G +2(2) , which we note ˜ K ∞ .In summary, in two dimensions, the transformation of ˆ V ( x, γ ) is of the formˆ V ( x, γ ) → ˆ k ( x, γ ) ˆ V ( x, γ )ˆ g ( w ) , (3.24)where ˆ g ( w ) ∈ G +2(2) and ˆ k ( x, γ ) ∈ ˜ K ∞ for each x . To avoid having to find the rightcompensator k ( x, γ ) and to work only with global transformations, one might want to try The loop extension of a group G is part of the untwisted affine extension of G . The extra elements inthe latter correspond the central element and the derivation of the algebra, which can be seen to have anaction on, respectively, the confomal factor ξ and the size ρ of the internal space [48, 49]. See for example[50] or [51] for more details on the theory of affine groups and algebras. M , which is,ˆ M := ˆ V ♯ (cid:18) x, − γ (cid:19) ˆ V ( x, γ ) . (3.25)A short calculation then reveals that ˆ M only depends on the spacetime independentspectral parameter w , i.e., it is independent of the spacetime coordinates ρ, z :ˆ M = ˆ M ( w ) . (3.26)The solution generating method then consists in multiplying the seed matrix ˆ M ( w ) withan appropriate group element g ( w ) of the loop group G +2(2) so that the new matrix isˆ M ′ ( w ) = g ♯ ( w ) ˆ M ( w ) g ( w ) . (3.27)Now one needs to find the factorization of the new ˆ M ′ ( w ) as in (3.25). Once the matrixˆ V ′ ( γ, x ) is obtained, the matrix V ′ ( x ) can be readily obtained by taking the γ → V should be chosen in the triangular gauge, so that in (3.25) the first factor is analytic ina neighborhood of γ = 0, and the second factor is analytic in the neighborhood of γ = ∞ .Moreover, the overlap of the respective domains of analyticity should be an annulus region[19, 20]. Finding such a factorization is a variant of the famous matrix valued RiemannHilbert problem. Consequently, avoiding the need to find the compensator ˆ k ( x ) is doneat the price of another equally difficult problem. However, when the matrix ˆ M ( w ) ischosen to be just consisting of simple poles in the complex w plane, such a factorizationcan be reduced to an algebraic problem [20]. But even then, one problem remains: how todecide which ˆ M ( w ) to start with, and which g ( w ) to multiply with, in order to generatean interesting solution of the higher dimensional theory?In the approach of [19, 20], although the group theoretic aspects of the correspondinggeneralized Geroch group remain clear, from the practical stand point, the construction ofnew solutions has to-date remained obscure. For this reason, in the next section we gener-alize the BZ construction to the G / ( SL (2 , R ) × SL (2 , R )) coset model. The BZ methodis also a purely algebraic procedure, but it has the advantage that it is more practical fromthe point of constructing new and interesting solutions of the higher dimensional theories.A detailed dictionary between the BZ construction and the approach of [19, 20] is stilllacking. We will discuss this and related points further in section 8. In section 4.1 we generalize the BZ construction to minimal supergravity in terms of the7 × M . In section 4.2 we discuss the action of the BZ transformations on the socalled factor space. 14 .1 The inverse scattering method for minimal supergravity in 5d Belinski and Zakharov provided a purely algebraic procedure for constructing solutions tothe Einstein vacuum equations with D − ( λ, ρ, z ). Defining a dressing matrix χ ( λ, ρ, z ), one seeks a new solution to (3.6) of the formΨ = χ Ψ . (4.1)Making this substitution into equation (3.6) results in the equations for χ . The matrix χ is further constrained to ensure that the new matrix M ( ρ, z ) = Ψ( λ = 0 , ρ, z ) is real andsymmetric.We are interested in ‘solitonic’ solutions which are characterized by having only simplepoles in the complex λ -plane. From (4.1), it follows that for a general n -soliton transfor-mation, the matrix χ will add n new poles to the seed solution Ψ . Furthermore, we shallrestrict ourselves to cases where the poles are located on the real axis; this determinesuniquely the location of the poles to be [17]˜ µ k = ± p ρ + ( z − ω k ) − ( z − ω k ) , (4.2)where the ω k are real constants. The ‘+’ and ‘ − ’ poles are commonly known as solitonsand antisolitons respectively.A general n -soliton transformation, in addition to the n real constants ω k , is determinedby n arbitrary real seven-dimensional constant vectors m ( k )0 . These are known as the BZvectors, and in the present context, their components control the addition of angularmomentum and charges to a given seed solution.Starting from a seed matrix M , an n -soliton transformation yields a new matrix M , M ab = ( M ) ab − n X k,l =1 ( M ) ac m ( k ) c (Γ − ) kl m ( l ) d ( M ) db ˜ µ k ˜ µ l , (4.3)where the repeated indices are summed. The vectors m ( k ) a are given by, m ( k ) a = m ( k )0 b (cid:2) Ψ − (˜ µ k , ρ, z ) (cid:3) ba , (4.4)and the matrix Γ is defined as, Γ kl = m ( k ) a ( M ) ab m ( l ) b ρ + ˜ µ k ˜ µ l . (4.5)The new matrix M , (4.3), does not satisfy det M = 1, and hence it is not a physicallyacceptable solution . For cases when the matrix M is 2 ×
2, e.g., for four dimensional For minimal supergravity we use a 7 × g ( ) . For other theories (orfor other representations of g ( ) for minimal supergravity) the dimension of these vectors depend on thedimension of the appropriate representation. Since G is a subgroup of SO (3 + , − ), all G matrices have determinant equal to unity. ρ and ˜ µ k ’s [17]. When the matrix M is not 2 × × M .The main idea is the observation that the determinant of the matrix (4.3) does notdepend on the BZ vectors. Thus, one may start with a vacuum static solution ( M , ξ )(hence diagonal matrix M ) and remove some solitons or antisolitons ˜ µ k at z = w k with‘trivial’ vectors, say, with the ones aligned with one of the basis vectors, that is, m ( k )0 b = δ ab for some given a . This amounts to rescaling (cf. (4.3)) the aa component of the matrix M while leaving allother components unchanged:( f M ) aa = − ˜ µ k ρ ( M ) aa , ( f M ) bb = ( M ) bb with b = a . (4.6)Next one re-adds the same solitons and antisolitons to the matrix f M but now with moregeneral vectors m ( k )0 a . Following this procedure, it is ensured by construction that thefinal matrix M and initial matrix M have the same determinant.Finally, the two-dimensional conformal factor ξ for the new solution is obtained fromthe seed solution as ξ = ξ r det Γdet Γ , (4.7)where the matrices Γ and Γ are obtained from (4.5) using M and M respectively. Com-pared to the expression given for e ν in [52] for example, we have an extra square root.This is due to the extra 1 / ξ compared to equations (3.5)for ν . Note that the extra term in the second set of equations, compared to the first one,does not play a role in this calculation (see also footnote 6).There is, however, one serious problem. The new matrix M in (4.3), obtained followingthe procedure described above, is in general not an element of the coset G / ( SL (2 , R ) × SL (2 , R )). A physical solution of minimal supergravity must be represented by a matrixthat is in the coset. We address this problem in section 5.2. In this subsection we discuss the action of the BZ transformations on the so-called factorspace. For the class of solutions we are interested in, the factor space can be identified[8] as the two-dimensional subspace orthogonal to the Killing vectors. We can use ρ and z as coordinates on this space [8, 13]. These coordinates cover the region outside thehorizon(s) and away from the rotational axes. The global structure of the factor space hasbeen discussed in detail in [8], so we shall be brief. Considering spacetimes containing non-extremal horizons , the corresponding factor space can be shown to be a two-dimensional Note that the freedom to rescale these vectors by a constant can be used to set one component of eachof these vectors to unity. See [12] for the generalisation to the extremal case. consists of adding new poles in the complex λ -plane at the location of the solitons. Since any four dimensional black hole spacetimewithin this class can be obtained from Minkowski space, this implies that the action ofthe BZ transformations on the factor space consists of adding new intervals together withthe corresponding vectors . The constants added to the seed solution via the BZ vectorsare related to the angular momentum and the NUT charges of the new solution.In five-dimensional vacuum gravity the action of the BZ transformations can be un-derstood in very much the same way. In practice the BZ method in five dimensionsis implemented following the approach of [52]. This consists of starting with a staticseed solution with (roughly speaking) the desired intervals, followed by removing solitonsand re-adding them using BZ. Unlike in four dimensions where one normally starts withMinkowski space, this procedure does not add new intervals to the boundary of the factorspace. Removing solitons generically renders the seed solution singular; the Killing vectorsno longer vanish in a smooth way on the axes (in fact some may diverge). However, thefactor space still has the same set of intervals as the seed solution. Re-adding the solitonsadds new poles so as to cure the singularities that appeared when removing the solitons .Associated to each interval there will now be a non-trivial vector related to the directionof the solitons.The above considerations are quite general, and do not per se refer to the matterfields of the theory. Therefore we expect that for minimal supergravity the action ofthe BZ dressing method on the factor space is essentially the same as it is for vacuumgravity. Of course, since in minimal supergravity the BZ vectors are seven dimensional,their components in general will be related to the angular momenta, NUT and the othercharges of the final solution. We use this intuition in the practical implementation of theBZ method for minimal supergravity in the following sections. Recall that for four dimensional Minkowski space, the boundary of the factor space is a single infiniteinterval. The final spacetime may still contain singularities that may or may not be removable by a choice ofparameters. Some general results
In this section we establish some general results of interest for the practical implementationof the BZ construction. In section 5.1 we show that the action of the three-dimensionalhidden symmetry transformations on the BZ dressing method is simply the group actionon the BZ vectors. In section 5.2 we discuss the issue of whether or not the BZ dressingmethod preserves the coset.
In this section we explicitly show that the action of the three-dimensional hidden symmetrytransformations on the BZ dressing method is simply the group action on the BZ vectors.Consider the matrices M and M g that are related by an element g of the three-dimensional symmetry group G as follows M g ( x ) = g T M ( x ) g. (5.1)Next consider the matrix M obtained by applying a certain BZ transformation on M .We write this as Ψ ( λ, x ) = χ ( λ, x ) e χ ( λ, x )Ψ ( λ, x ) ≡ χ ( λ, x ) e Ψ ( λ, x ) . (5.2)where Ψ ( λ, x ) and Ψ ( λ, x ) are the generating matrices corresponding respectively to M ( x ) and M ( x ), e χ ( λ, x ) is the dressing matrix that removes solitons from Ψ ( λ, x ) and χ ( λ, x ) is the one that adds solitons to e Ψ ( λ, x ) ≡ e χ ( λ, x )Ψ ( λ, x ). Applying the grouptransformation g to the matrix M leads to another matrix M g : M g ( x ) = g T M ( x ) g. (5.3)We show that by performing the same BZ transformation on M g (with appropriatelymodified BZ vectors to be discussed below) one gets exactly M g . Schematically, thesituation is presented in figure 1.It is useful to first notice that the generating matrices Ψ transform under group ele-ments in the same way as the corresponding M matrices. That is, the generating matricesΨ g , that solve the Lax equations (3.6) and satisfy Ψ g , ( λ = 0 , x ) = M g , ( x ) are simplyΨ g , ( λ, x ) = g T Ψ , ( λ, x ) g . (5.4)Using this and equation (5.2), and inserting = ( g T ) − g T one hasΨ g ( λ, x ) = g T Ψ ( λ, x ) g = (cid:0) g T χ ( λ, x ) ( g T ) − (cid:1) (cid:16) g T e Ψ ( λ, x ) g (cid:17) (5.5) ≡ (cid:0) g T χ ( λ, x ) ( g T ) − (cid:1) e Ψ g . In our case, G = G , but the results of this section apply to more general situations. The precise transformation of the matrix M under global G transformations is (2.10). To avoidnotational clutter we simply use g to mean S − gS in this section. BZ BZ g M M g g ∈ Gg ∈ G M g Figure 1: This figure depicts the action of the hidden symmetry transformations on theBZ construction. Let the matrix M be sent to M g by an element g ∈ G and to M by aBZ transformation. Let M be sent to the matrix M g under the action of g . Then M g canalso be obtained from M g by applying the same BZ transformation but with BZ vectorsappropriately transformed under g .This equation tells us that if one starts with the matrix obtained by removing solitons onΨ ( λ, x ) and transforming it with the element g ∈ G , then the matrix Ψ g is obtained byperforming a BZ transformation using ˆ χ ( λ, x ) ≡ g T χ ( λ, x ) ( g T ) − as the dressing matrix.In the following we study in more detail the new dressing matrix ˆ χ and show that it addsthe same solitons as χ but with BZ vectors transformed appropriately under g .Let us now consider the dressing matrix ˆ χ . It follows from the BZ construction [17]that it obeys ˆ χ = g T χ ( λ, x ) ( g T ) − = 1 + n X k =1 g T R k ( x ) ( g T ) − λ − ˜ µ k ( x ) , (5.6)so we can concentrate on the residue matrices ˆ R k ( x ) ≡ g T R k ( x ) ( g T ) − . From the generalBZ construction [17] we know that the matrices R k are degenerate and factorize as( R k ) ab = n ( k ) a m ( k ) b . (5.7)Note that a, b, . . . are group representation indices. The n ( k ) are naturally column vectorswhereas the m ( k ) take the form of row vectors. Thus,ˆ R k = g T R k ( g T ) − = (cid:16) g T n ( k ) (cid:17) (cid:16) m ( k ) ( g − ) T (cid:17) ≡ ˆ n ( k ) ˆ m ( k ) . (5.8)A short calculation yieldsˆ m ( k ) ≡ m ( k ) ( g − ) T = m ( k )0 e Ψ − (˜ µ k , x )( g − ) T = (cid:16) m ( k )0 g (cid:17) ( e Ψ g ) − (˜ µ k , x ) ≡ ˆ m ( k )0 ( e Ψ g ) − (˜ µ k , x ) , (5.9)where expressions (4.4) and the definition for e Ψ g have been employed in the second andthird equalities, respectively. Regarding the vector ˆ n ( k ) , we now show that its implicitdefinition from the factorization of the residue matrix (5.8) matches the BZ prescription19or the construction of M g from M g . The latter determines ˆ n ( k ) in terms of ˆ m ( k ) and f M g ( x ) ≡ e Ψ g ( λ = 0 , x ). Indeed, recall that [17]ˆ n ( k ) a = X l ˜ µ − l ˆ D kl ˆ N ( l ) a , ˆ N ( l ) a = ˆ m ( l ) c f M g ( x ) ca (5.10)where ˆ D kl is the inverse of the matrix ˆΓ kl introduced in (4.5). Next, observe that thematrix ˆΓ kl that takes us from M g to M g is exactly the same as the matrix Γ kl that takesus from M to M : ˆΓ kl = (cid:0) m ( k ) ( g − ) T (cid:1) (cid:16) g T f M ( x ) g (cid:17) (cid:0) m ( l ) ( g − ) T (cid:1) T ρ + ˜ µ k ˜ µ l = m ( k ) f M ( x )( m ( l ) ) T ρ + ˜ µ k ˜ µ l = Γ kl . (5.11)It follows immediately that ˆ D kl = D kl . Finally, consider the vector ˆ N ( l ) . It is also relatedto its untransformed counterpart in a simple manner:( ˆ N ( l ) ) T = ˆ m ( l ) f M g ( x )= (cid:16) m ( l ) ( g − ) T (cid:17) (cid:16) g T f M ( x ) g (cid:17) = m ( l ) f M ( x ) g = ( N ( l ) ) T g . (5.12)Therefore, ˆ n ( k ) = X l ˜ µ − l ˆ D kl ˆ N ( l ) = g T X l ˜ µ − l D kl N ( l ) = g T n ( k ) , (5.13)in accordance with (5.8).In conclusion, we have established that to generate M g from M g using the inversescattering method all we need to do is to make the replacement m ( k )0 −→ ˆ m ( k )0 = m ( k )0 g , and (5.14) e Ψ ( λ, x ) −→ e Ψ g ( λ, x ) = g T e Ψ ( λ, x ) g , (5.15)in the calculation that generates M from M . Thus, the BZ construction of M g from M g and of M from M are essentially the same. In particular, this allows us to easily obtainthe Cvetiˇc-Youm (CY) solution by applying the BZ construction on Reissner-Nordstr¨om(RN). The latter is obtained from Schwarzschild by a G transformation [39, 38], soall we need to know is how to derive Myers-Perry (MP) from Schwarzschild using inversescattering, and this is accomplished in section 6.1. In section 6.2 we also explicitly obtainthe Cvetiˇc-Youm (CY) solution from Reissner-Nordstr¨om (RN).The result of this section supports the idea that the inverse scattering method isincapable of producing charging transformations since BZ transformations are essentiallyinsensitive to the presence of charges generated by the three-dimensional hidden symmetrytransformations. 20 .2 Do we stay in the coset after the BZ transformations? In this subsection we address the issue of whether or not the BZ transformations preservethe coset. We start by discussing a toy model where similar problems arise but can beeasily resolved. Then we discuss the case of minimal supergravity where the resolutionseems to require to impose an extra non-linear constraint (equation (5.22) below) on theparameters of the solution.Suppose for a moment that instead of minimal supergravity we were working withsome other theory with the following coset model structure: SL (3 , R ) SO (2 , × SL (3 , R ) SO (2 , . (5.16)The higher dimensional origin of this coset model is not important for the point we wantto make . The coset representative matrix M for this theory will have the form M = M M ! , (5.17)where M and M are two independent matrices representing two SL (3 , R ) SO (2 , cosetrepresentatives. Multisoliton solutions of this theory will be characterized by two integers( n , n ), referring to the number of solitons associated to each SL (3 , R ) SO (2 , coset model. Foreach soliton in each SL (3 , R ) there is a freedom to add three free parameters, namely,the BZ vector associated to the soliton. Therefore, for ( n , n ) solitons we can add amaximum of 3( n + n ) free parameters and still maintain the block diagonal form of thecoset representative. This number is obviously less than 6( n + n ), which is the numberof free parameters one might have excepted very naively from the fact that the matrix M is six dimensional. This theory has an additional symmetry, which we call “double image”symmetry: every solution possesses a companion solution that is obtained by interchangingthe field content of the two SL (3 , R ) /SO (2 ,
1) coset models. Two solutions related by thissymmetry may admit very different interpretations.The situation in minimal supergravity is somewhat similar. Recall that the sevendimensional representation of G branches into SL (3 , R ) representations as = ¯ + + . (5.18)The and ¯ represent two appearances of vacuum gravity in the matrix M : M = M − M SL(3) . (5.19) Although a higher dimensional origin of the SL (3 , R ) SO (2 , × SL (3 , R ) SO (2 , coset model is not immediately obvious,appropriate truncation of heterotic theory does give rise to SL (2 , R ) SO (2) × SL (2 , R ) SO (2) coset model, see e.g., [53]. Asimilar structure also arises in certain truncations of five-dimensional Einstein-Maxwell theory [54]. Note that there is a freedom to rescale each of these vectors by a constant and therefore, without lossof generality, we can always set one of the components of each vector to unity. For the sake of simplicityof the argument, we are not concerned with the normalization of the BZ vectors in this discussion. M SL(3) and M − both represent exactly the same SL (3 ,R ) SO (2 , coset model. We call this situation a “mirror image”. Therefore, in order to maintain theform (5.19), when we add n -solitons to the matrix M SL(3) and 3 n free parameters throughBZ vectors, we necessarily have to add n -antisolitons to the matrix M − . The BZ vectorsassociated to the antisolitons have to be related to the solitonic BZ vectors in such a waythat the final matrix M has precisely the block diagonal form (5.19). It is intriguing thatin all the examples we have looked at this can be achieved by relating the antisolitonic BZvectors to the solitonic BZ vectors in a simple way. Though, a priori, it is not clear if itcan be achieved for a general solution of vacuum gravity.In the general case, the matrix M representing a solution of minimal supergravity hasthe symmetrical block diagonal structure (C.5) [39, 40] M = A U BU T ˜ S V T B T V C , (5.20)where A and C are symmetric 3 × B is a 3 × U and V are 3-componentcolumn matrices, and ˜ S a scalar. The inverse matrix is given by (C.6) [39, 40] M − = C − V B T − V T ˜ S − U T B − U A . (5.21)We observe that • Any solution of minimal supergravity should admit a limit to a solution of vacuumgravity, and since vacuum gravity is encoded two times in the 7 × M , onceas M SL(3) and second time as M − , it is natural to think that the field content ofminimal supergravity is also ‘encoded two times’ in the matrix M . • This observation is further strengthened by the fact that the inverse of the matrix M is simply given by a reshuffling of the rows and columns (up to certain minussigns) of the matrix M .Based on these observations, we expect that the soliton transformations on matrix M inthe general case are always accompanied with antisoliton transformations.In practice, when we apply solitonic and antisolitonic transformations with general BZvectors we move out of the coset. (We discuss an explicit example where this problemarises in the next section.) This is a serious problem, because a physical solution ofminimal supergravity must be represented by a matrix that is in the coset. Since the cosetconstruction is based on the exceptional Lie group G one cannot write a simple matrixcriterion for a 7 × M obtained through BZ construction to be in the coset are the following1. The matrix M := V ♯ V = η − ( S T ) − M S − should preserve the three form c ( a ) abc constructed in appendix D in the sense that X a,b,c =1 c ( a ) abc M ae M bf M cg = c ( a ) efg , (5.22)22. The matrix M should be symmetric.Roughly speaking, the first condition ensures that the final matrix M is in G and the second condition, given the first condition, ensures that it is in the coset G / ( SL (2 , R ) × SL (2 , R )). The BZ construction guarantees the second condition, butit does not guarantee the first one. We believe that through certain relationships betweensolitonic and antisolitonic BZ vectors one might ensure that the first condition is alsosatisfied. However, how to find these precise conditions in the general case is not clear tous. Therefore, at this stage, one should regard equation (5.22) as an extra non-linear con-straint that every physical solution has to satisfy. Further exploration of these questionsis left for the future. As an illustration of our formalism, we obtain the doubly spinning five-dimensional Myers-Perry black hole by applying solitonic transformations on the Schwarzschild black hole(section 6.1). We also derive the Cvetiˇc-Youm black hole by applying solitonic transfor-mations on the Reissner-Nordstr¨om black hole (section 6.2).
The five-dimensional Myers-Perry solution has been constructed from 5d Schwarzschild-Tangherlini using the BZ method in [52] in the context of vacuum gravity. However, we arenow considering five-dimensional minimal supergravity and so this calculation provides agood testing ground for our framework.For 5d Schwarzschild the metric in the parametrization (3.1) takes the form G Schw¯ µ ¯ ν = diag n G Schw tt , G Schw φφ , G Schw ψψ o = diag n − µ µ , µ , ρ µ o , (cid:0) e ν (cid:1) Schw = µ ( ρ + µ µ )( ρ + µ )( ρ + µ ) . (6.1)In terms of the SL (3 , R ) matrix of [43], the Killing metric part translates into M SL(3) = diag ( G Schw tt , G Schw ψψ , G Schw tt G Schw ψψ ) = diag (cid:26) − µ µ , µ , − µ (cid:27) . Since this is a vacuum solution we know from (2.12) that M Schw = diag n M − , , M SL(3) o = diag (cid:26) − µ µ , µ , − µ , , − µ µ , µ , − µ (cid:27) . It is known that the five-dimensional doubly spinning Myers-Perry solution can be obtainedfrom Schwarzschild-Tangherlini by a two-soliton transformation [52]. In the context of5d minimal supergravity this, in practice, amounts to a four-soliton transformation on M Schw . This is because each transformation on the M SL(3) block must be accompanied bya corresponding transformation on the M − block.To obtain the seed matrix we first remove an antisoliton at position z = w with BZvector ( ~ , , ~e ) and, accordingly, remove a soliton at the same position but with BZ vector23 ~e , ,~ ~e ≡ (1 , ,
0) for convenience. The second pair consists ofremoving a soliton (respectively an antisoliton) at position z = w with the same BZvectors as the former antisoliton (respectively a soliton). This whole procedure can beimplemented by multiplying M Schw by H Schw = diag (cid:26) µ µ , , , , µ µ , , (cid:27) . (6.2)In addition, we rescale the matrix thus obtained by ( µ µ ) − to avoid the appearance ofdivergences during the BZ construction. Hence, our seed matrix is f M Schw ≡ ( µ µ ) − H Schw M Schw = diag (cid:26) − µ , µ µ , − µ , µ µ , − µ , µ , − µ µ (cid:27) , and the associated solution of the Lax pair (3.6), e Ψ Schw ( λ, ρ, z ), is obtained by shifting µ k → µ k − λ , where k = 1 , • add an antisoliton at w with m (1)0 = (cid:16) ~ , , A , B , (cid:17) , • add a soliton at w with ¯ m (1)0 = (cid:16) ¯ A , ¯ B , , ,~ (cid:17) , • add a soliton at w with m (2)0 = (cid:16) ~ , , A , , C (cid:17) , • add an antisoliton at w with ¯ m (2)0 = (cid:16) ¯ A , , ¯ C , ,~ (cid:17) .The new matrix, M MP , is then obtained from (4.3). It is now convenient to use the freedomof translating the z coordinate so that the (anti-)solitons are placed at w = α = − w .The mere fact that the first 3 × M must be the inverse of the last 3 × C = 4 αC ¯ A A , ¯ B = B α ¯ A A . (6.3)If we do not impose these conditions then the matrix M MP is not an element of thecoset G / ( SL (2 , R ) × SL (2 , R )). The conditions (6.3) are precisely the solution of thenon-linear constraint (5.22) in this example. With these conditions the new matrix M MP depends only on four parameters, which we take to be { A , B , A , C } . Moreover, ifwe take B , C → M MP we then extract the five scalars φ , φ , χ , χ and χ that are turned on for this solution using (2.13).To construct the conformal factor for the dressed solution we first note that e ν = e √ φ + φ ξ . (6.4)from (2.2) and (2.17). Applying the transformation (4.7) to the conformal factor forthe Schwarzschild solution given in (6.1) we obtain the conformal factor for the dressed24olution. To compare it with the Myers-Perry one we relate the components of the BZvectors to the angular momentum parameters as B = l A r αβ , C = l A √ αβ , (6.5)where we have defined β ≡ α + p l l + 4 α for convenience. As mentioned in section 3.1,there is still a multiplicative constant of integration which is undetermined by the equationsof motion for the conformal factor but it can be fixed by the requirement of asymptoticflatness. Expressed in prolate spherical coordinates (see appendix A) the conformal factorfor Myers-Perry becomes (cid:0) e ν (cid:1) MP = 4 αx + ( l − l ) y + 2 m α ( x − y ) , (6.6)where 2 m ≡ l + l + 2 p l l + 4 α . To recover the full five-dimensional metric one then needs to follow the inverse routeof the procedure outlined in section 2.1. Namely, one dualizes the new scalars χ and χ back into one-forms and then reconstructs the metric using (2.2). In order to identifythe new solution as Myers-Perry, one needs to perform a linear transformation mixingthe coordinates { t, φ, ψ } . The latter is easily determined by imposing that the angularcoordinates match asymptotically with those for flat spacetime. This transformation reads t −→ t − l (cid:0) l − l − α + β (cid:1) l − l φ + l l − l ψ ,φ −→ − l l √ αβ φ + r β α ψ ,ψ −→ r β α φ − l l √ αβ ψ . (6.7)The metric thus obtained is the Myers-Perry solution expressed in prolate spherical co-ordinates. This is displayed in appendix A, with the understanding that the unchargedsolution has δ = 0. ˇc -Youm solution from 5d Reissner-Nordstr¨om Using the derivation of the previous subsection and the results of section 5.1 we knowhow to construct the Cvetiˇc-Youm solution by applying a solitonic transformation onthe Reissner-Nordstr¨om solution. In this subsection, we present the main steps of thiscalculation.One first needs to construct the matrix M for the Reissner-Nordstr¨om solution. Thisis most easily done as [38] M RN = g T M Schw g (6.8)for g = S − e √ k δ S where the matrix S and the generator k are defined in appendicesB and C while δ is the charge parameter appearing in the Reissner-Nordstr¨om solution25given in appendix A if one takes l = l = 0). The corresponding generating matrix issimilarly obtained as Ψ RN = g T Ψ Schw g. (6.9)As discussed in section 5.1, the seed solution is obtained by removing from Ψ RN the samesolitons and antisolitons as in the vacuum case but with BZ vectors that are now non-trivial: they are equal to ( ~ , , ~e ) g and ( ~e , ,~ g . This is equivalent to taking the seedto be e Ψ RN = g T e Ψ Schw g and f M RN = g T f M Schw g . (6.10)To perform the dressing we add the same solitons and antisolitons as in the vacuumcase but with BZ vectors taken to be b m (1)0 = m (1)0 g , b ¯ m (1)0 = ¯ m (1)0 g, (6.11) b m (2)0 = m (2)0 g , b ¯ m (2)0 = ¯ m (2)0 g. (6.12)In this case, we already know all five conditions to impose on the eight BZ parameters. Itis reassuring that the new matrix M CY is in the coset and that the new conformal factormatches with the Cvetiˇc-Youm solution.From M CY one can reconstruct the five-dimensional metric and gauge field. As for theMyers-Perry case, the identification of the solution requires a linear change of coordinates: t −→ t + c φ + c ψ ,φ −→ − l l √ αβ φ + r β α ψ ,ψ −→ r β α φ − l l √ αβ ψ . (6.13)where c = − l (cid:0) l − l − α + β (cid:1) c − l (cid:0) − l + l − α + β (cid:1) s ( l − l ) ,c = l c − l s ( l − l ) , and s = sinh δ and c = cosh δ . It is easy to see that this change of coordinates simplifiesto (6.7) when δ = 0. The metric and the gauge field thus obtained are the ones forCvetiˇc-Youm expressed in prolate coordinates, which is presented in section A. In this section we argue that to see the integrability of five-dimensional minimal super-gravity one necessarily has to perform an Ehlers reduction to two dimensions.In sections 2 and 3.1 we described two ways of reducing a theory to two dimensions.The first one, performed in section 2 for minimal five-dimensional supergravity, is in two26teps. First one reduces to three dimensions and dualizes vectors into scalars. Second,one finally reduces to two dimensions. This two-step way of doing the reduction to twodimensions is called an Ehlers reduction. The second method, described in section 3.1 forvacuum gravity in five dimensions, consists in reducing directly to two dimensions, withoutdualizations. This way of doing the reduction is called a Matzner-Misner reduction.The dualization of vectors into scalars in three dimensions can make some symmetriestransparent. These are the so-called hidden symmetries. For a series of theories [2], thehidden symmetry group ( G ) is larger than (or equal to) the symmetry group obtained viadimensional reduction without dualization. In all these cases the dimensionally reducedsupergravity Lagrangian can be written in three dimensions, after dualization, as gravityplus a non-linear sigma model with the symmetry group G . Upon further reduction to twodimensions this hidden symmetry group G gets enlarged to its untwisted affine extension,generally denoted as G (1) . In all known examples, this untwisted affine extension of thegroup G is the symmetry group of the dimensionally reduced theory in two dimensions.One way to show this is through the Lax pair in the Ehlers reduction, as explained insection 3.4. Thus, the untwisted affine extension of the group G naturally leads to anintegrable two-dimensional sigma model based on the group G . However, in some cases, theMatzner-Misner reduction also allows us to see this symmetry, and thus the integrabilityof the theory. This is the case, for example, for vacuum gravity. This is most easilyunderstood by considering the problem from a slightly different perspective, namely interms of Dynkin diagrams.Schematically one can also view the affine extension of G appearing in the dimensionalreduction to two dimensions in the following way (without invoking the Lax pair): • Perform an Ehlers reduction to two dimensions: one then obtains a system of scalarsΦ that transform non-linearly under G , which is the symmetry group in three di-mensions. • Perform a Matzner-Misner reduction to two dimensions: one obtains another systemof scalars ¯Φ, related to Φ by non-local field redefinitions involving dualizations, andtransforming in a certain way under ¯ G . • Studying the intertwining of G with ¯ G , that is, studying how G acts on ¯Φ and ¯ G acts on Φ, one finds that they form an infinite-dimensional group which is the affineextension of G . This is how, for example, the infinite dimensional Geroch group [55]of four dimensional general relativity was discovered (although the affine structureof the group was not clear at the time of [55]).In the case of five-dimensional vacuum gravity, the Ehlers group G and the Matzner-Misner group ¯ G are both SL (3 , R ). These two SL (3 , R ) intertwine to give the affine group SL (3 , R ) + , which is the symmetry of vacuum five-dimensional gravity when reduced totwo dimensions. The Dynkin diagram of SL (3 , R ) + is shown in figure 2. On the Dynkindiagram the Ehlers SL (3 , R ) corresponds to the two white dots, while the Matzner-Misner SL (3 , R ) is made out of the black dot and one of the white dots. In this case, as G and¯ G are the same groups, taking the Lax pair based on any of the two leads to the sameinfinite dimensional symmetry. The integrability can accordingly be seen from both pointof views. This is the reason why we are able to perform a BZ construction based on27igure 2: Dynkin diagram of the untwisted affine extension of sl ( , R ).the Killing sector of the metric (Matzner-Misner) as well as based on the matrix M SL(3) (Ehlers).For five-dimensional minimal supergravity, the Ehlers group G is G . On the otherhand, the Matzner-Misner group is SL (3 , R ). Through a Matzner-Misner reduction, thescalars coming from the five-dimensional metric will form an SL (3 , R ) /SO (2 ,
1) cosetmodel. These scalars will be coupled in a non-linear way with the degrees of freedomcoming from the reduction of the five-dimensional gauge field A . On the Dynkin diagramof G +2(2) given in figure 3, the Ehlers group G is given by the two white dots, whilethe Matzner-Misner SL (3 , R ) is made of the black dot and the white dot connected toit. In this case, as ¯ G is not the same as G but only a subgroup, taking the Lax pairFigure 3: Dynkin diagram of the untwisted affine extension of g ( ) .based on ¯ G will not encompass the whole symmetry of the theory. (It is not even clear ifsuch a Lax pair construction can be carried out.) As a consequence, it is necessary to dothe Ehlers reduction in order to see the integrability of five-dimensional supergravity inthe presence of three commuting Killing vectors. In principle, with appropriate non-localfield redefinitions one should be able to combine the SL (3 , R ) fields in the Matzner-Misnerreduction with the degrees of freedom of the dimensionally reduced gauge field A and seean integrable sigma model based on the affine extension of g ( ) . We however believe thatsuch field redefinitions would precisely correspond to an Ehlers reduction. Furthermore,we note that all theories based on classical Lie algebras, namely the A r , B r , C r and D r series of Lie algebras, give rise to the same integrable sigma model structure upon doingthe reduction either the Ehlers way or the Matzner-Misner way. In particular, it is truefor vacuum gravity in any dimension and for heterotic string theory [56]. One aim of this paper was to explore the integrability of five-dimensional minimal su-pergravity in the presence of three commuting Killing vectors with potential applications28n mind. Although it is well known in the supergravity literature [19, 20, 21] that thedimensionally reduced theory is completely integrable, the integrability structure has notbeen used as a solution generating technique. In this paper we took steps in this direc-tion. We derived the BZ Lax pair for minimal supergravity, using a 7 × M , which is well suited for many applications. We also generalizedthe well-known BZ construction; however, we are unable to provide precise conditions (tobe imposed perhaps on the BZ vectors) that will ensure by construction that the dressedmatrix is also in the coset. Since the coset construction for this theory is based on theexceptional Lie group G , checking whether a final matrix is in the coset is itself anon-trivial task. We provided an algebraic way for checking if the final matrix M is in thecoset. Our method is based on the invariant three-form for our representation of the Liealgebra g ( ) . An explicit construction of this three-form is given in appendix D.In section 3.3, following [19] we clarified the relationship between the BZ Lax pairand the Lax pair of [20, 19]. However, the precise relationship between the BZ inverse-scattering construction and the group theoretic solution generating approach of [20, 19] isfar from clear. We believe that understanding the precise relationship between these twotechniques will also give some insights into the problem of ensuring in the BZ dressingmethod that the final matrix M is in the coset by construction. A detailed exploration ofthese ideas is left for the future.As an illustration of the usefulness of our formalism, we obtained the doubly spin-ning five-dimensional Myers-Perry black hole using the 7 × M as a four solitontransformation on the (four solitons removed) Schwarzschild black hole. We also derivedthe Cvetiˇc-Youm black hole solution as a four-soliton transformation on the four-solitonremoved Reissner-Nordstr¨om solution. The second calculation was achieved by first study-ing in detail the action of the three-dimensional hidden symmetry transformations on theBZ construction.In section 7 we argued that to see the integrability of five-dimensional minimal super-gravity one necessarily has to perform an Ehlers reduction. A direct dimensional reductionto two dimensions does not allow us to see the integrability of the theory in an easy way.This situation is in contrast with the situation in vacuum gravity [17] and with many othertheories, e.g., heterotic theory [56], where the integrability can be seen either way.The next step would be to understand the physics of dipole charges in our formalism.This is especially interesting because the known dipole black ring [24], having only onenon-zero angular momentum parameter, is not the most general dipole ring. A dipoleblack ring with two independent angular momenta is expected to exist. Our formalismprovides a framework where a construction of doubly spinning dipole black ring can inprinciple be carried out, unlike in the approach of [54] where such a construction is notpossible in principle. In order to construct the doubly spinning dipole ring, one could firsttry to understand the dipole charge in our coset formalism and then add rotation on thetwo-sphere using solitonic transformations. Once this is achieved adding electric charge tothis solution would be relatively easy; i.e., it can be done, say, using the charging transfor-mations of [39, 38]. This line of investigation may very well let us discover the conjecturedfive parameter family of non-supersymmetric black rings in minimal supergravity.29 cknowledgements We thank Joan Camps, Sophie de Buyl, Roberto Emparan, Marc Henneaux, Axel Klein-schmidt, Josef Lindman H¨ornlund, James Lucietti, Jnanadeva Maharana, Hermann Nico-lai, Jakob Palmkvist, and Simon Ross for interesting discussions. We are particularlygrateful to Jakob Palmkvist for his patient explanations of octonions. PF is supported byan STFC rolling grant. EJ was a FRS-FNRS bursar. EJ and AV were supported by IISN -Belgium (conventions 4.4511.06 and 4.4514.08) and by the Belgian Federal Science PolicyOffice through the Interuniversity Attraction Pole P6/11. JVR acknowledges financialsupport from
Funda¸c˜ao para a Ciˆencia e Tecnologia (FCT)-Portugal through fellowshipSFRH/BPD/47332/2008. PF and JVR would like to thank the IISN and the TheoreticalPhysics group at ULB for hospitality during the final stages of this work.
A Cveti ˇc -Youm solution in canonical coordinates A four parameter family of charged rotating black holes in five-dimensional minimal super-gravity was obtained by Cvetiˇc and Youm in [57] by applying boosts and string dualitiesto the (neutral, rotating) Myers-Perry black hole in five dimensions. The four parametersspecifying the solution can be chosen to be the mass M , two angular momenta J φ,ψ , andelectric charge Q E . It is convenient to present the solution in terms of the quantities m, l , l , and δ related to the above through [58] M = (cid:18) π G (cid:19) m (cid:0) s (cid:1) , (A.1) J φ = (cid:18) π G (cid:19) m (cid:0) l c − l s (cid:1) , (A.2) J ψ = (cid:18) π G (cid:19) m (cid:0) l c − l s (cid:1) , (A.3) Q E = 14 π Z S ∞ (cid:18) ⋆F − F ∧ A √ (cid:19) = − √ πmcs . (A.4)where for notational convenience we use s := sinh δ , and c = cosh δ .The spacetime has three commuting Killing directions which are denoted by t , φ and ψ ,the first coordinate being time-like whereas the last two represent angular directions. Theremaining coordinates parametrize the radial direction and another angle, respectively r and θ . In these coordinates the metric and the gauge field read as follows ds = g tt dt +2 g tφ dtdφ +2 g tψ dtdψ + g φφ dφ + g ψψ dψ +2 g φψ dφψ + g rr dr + g θθ dθ , (A.5)and the gauge field as A = A t dt + A φ dφ + A ψ dψ , (A.6)30ith g tt = − Σ(Σ − m )(Σ + 2 ms ) ,g tφ = − m sin θ (cid:2) Σ (cid:8) l c − l s (cid:9) + 2 ml s (cid:3) (Σ + 2 ms ) ,g tψ = − m cos θ (cid:2) Σ (cid:8) l c − l s (cid:9) + 2 ml s (cid:3) (Σ + 2 ms ) ,g φφ = sin θ (Σ + 2 ms ) (cid:2) ( r + 2 ms + l )(Σ + 2 ms ) + 2 m sin θ (cid:8) Σ( l c − l s ) + 4 ml l c s − ms ( l c + l s ) − ml s (cid:9)(cid:3) ,g ψψ = cos θ (Σ + 2 ms ) (cid:2) ( r + 2 ms + l )(Σ + 2 ms ) + 2 m cos θ (cid:8) Σ( l c − l s ) + 4 ml l c s − ms ( l c + l s ) − ml s (cid:9)(cid:3) ,g φψ = 2 m cos θ sin θ (cid:2) l l (cid:8) Σ − ms (cid:9) + 2 m ( l + l ) s c − ml l s (cid:3) (Σ + 2 ms ) ,g rr = r (Σ + 2 ms )( r + l )( r + l ) − mr ,g θθ = Σ + 2 ms ,A t = 2 √ msc (Σ + 2 ms ) ,A φ = − A t ( l c − l s ) sin θ ,A ψ = − A t ( l c − l s ) cos θ . (A.7)For convenience we have definedΣ( r, θ ) ≡ r + l cos θ + l sin θ . (A.8)Setting δ = 0 reproduces the five-dimensional MP black hole.To make contact with the inverse scattering method we need to express the metric (A.5)in canonical form, ds = G ¯ µ ¯ ν dx ¯ µ dx ¯ ν + e ν (cid:0) dρ + dz (cid:1) , with ρ = p | det G | , (A.9)where the three-dimensional Killing part of the metric, G ¯ µ ¯ ν , and ν depend only on thecoordinates ( ρ, z ). It turns out that the determinant of the Killing part of (A.5) is exactly the same as for the five-dimensional MP solution, namelydet g (3) = − (cid:2) ( r + l )( r + l ) − mr (cid:3) sin θ cos θ . (A.10)Thus, we immediately obtain ρ = r sin 2 θ s r (cid:18) l r (cid:19) (cid:18) l r (cid:19) − m . (A.11)31o determine z = z ( r, θ ) one just assumes separability and requires (A.5) to take theform (A.9). Once again this yields the same expression as for 5-dimensional MP: z = (cid:18) r − m − l − l (cid:19) cos 2 θ . (A.12)The conformal factor ν ( r, θ ) is given by e ν = Σ + 2 ms (cid:2) ( r + l )( r + l ) − mr (cid:3) cos θ + (cid:0) r + l + l − m (cid:1) sin θ . (A.13)Of course, we need to express ν in terms of the canonical coordinates { ρ, z } . To this end, itis useful to introduce the prolate spherical coordinates, x ∈ [1 , ∞ ) and y ∈ [ − , x = 2 r + l + l − m α , y = cos 2 θ , (A.14)where α = 14 q (2 m − l − l ) − l l . (A.15)In terms of these coordinates the conformal factor simplifies, e ν = 4 αx + ( l − l ) y + 2 m (1 + 2 s )8 α ( x − y ) . (A.16)The last term, proportional to s , is the only difference relative to the five-dimensionalMP case.The components of the metric G ¯ µ ¯ ν and the gauge field A ¯ µ may be easily expressed interms of the prolate spherical coordinates, as well. One finds G tt = − (cid:0) αx + ( l − l ) y (cid:1) − m (cid:8) αx + ( l − l ) y + 2 m (1 + 2 s ) (cid:9) ,G tφ = − m (1 − y ) (cid:2)(cid:8) αx + ( l − l ) y + 2 m (cid:9) l c − (cid:8) αx + ( l − l ) y − m (cid:9) l s (cid:3)(cid:8) αx + ( l − l ) y + 2 m (1 + 2 s ) (cid:9) ,G tψ = − m (1 + y ) (cid:2)(cid:8) αx + ( l − l ) y + 2 m (cid:9) l c − (cid:8) αx + ( l − l ) y − m (cid:9) l s (cid:3)(cid:8) αx + ( l − l ) y + 2 m (1 + 2 s ) (cid:9) ,G φφ = (1 − y ) (cid:8) αx + l − l + 2 m (1 + 2 s ) (cid:9) m (1 − y ) (cid:8) αx + ( l − l ) y + 2 m (1 + 2 s ) (cid:9) (cid:2)(cid:8) αx + ( l − l ) y + 2 m (cid:9) (cid:0) l c − l s (cid:1) +8 ml l c s − ms (cid:0) l c + l s (cid:1) − ml s (cid:3) , ψψ = (1 + y ) (cid:8) αx − l + l + 2 m (1 + 2 s ) (cid:9) m (1 + y ) (cid:8) αx + ( l − l ) y + 2 m (1 + 2 s ) (cid:9) (cid:2)(cid:8) αx + ( l − l ) y + 2 m (cid:9) (cid:0) l c − l s (cid:1) +8 ml l c s − ms (cid:0) l c + l s (cid:1) − ml s (cid:3) ,G φψ = 2 m (1 − y ) (cid:2) l l (cid:8) αx + ( l − l ) y + 2 m (1 − s ) (cid:9) + 2 m ( l + l ) s c − ml l s (cid:3)(cid:8) αx + ( l − l ) y + 2 m (1 + 2 s ) (cid:9) ,A t = 4 √ mcs (cid:8) αx + ( l − l ) y + 2 m (1 + 2 s ) (cid:9) ,A φ = − A t ( l c − l s ) (1 − y )2 ,A ψ = − A t ( l c − l s ) (1 + y )2 . The above components may be translated into canonical coordinates by using thefollowing relations [59]: ρ = α p ( x − − y ) , z = αxy , (A.17) x = R + + R − α , y = R + − R − α , (A.18)with R ± = p ρ + ( z ± α ) . This completes the task of expressing the CY metric (A.5) in canonical form (A.9).
B Generalities on G For the purpose of reference, here we collect some basic facts about the Lie algebra g andits split real form g ( ) . For further details we refer the reader to standard references, suchas [60]. See also [38].The algebra g is the smallest of the exceptional Lie algebras. It has rank 2 and itsdimension is 14. Its Dynkin diagram is presented in figure 4. Each node of this diagramFigure 4: Dynkin diagram of G .corresponds to a triple of Chevalley generators { H a , E a , F a } , a = 1 ,
2. The H a ’s span theCartan subalgebra h of g . The E a ’s are the generators associated to the two simple roots ~α and ~α of g . These generators satisfy the Chevalley relations[ H , E ] = 2 E , [ H , E ] = − E , [ E , F ] = H , [ H , E ] = − E , [ H , E ] = 2 E , [ E , F ] = H . (B.1)33he simple roots belong to the dual h ⋆ of h . By taking multiple commutators of E a ’s, andusing Serre relations, one obtains a set of four more positive generators E k , k = 3 , . . . , E = [ E , E ] , E = [ E , E ] , E = [ E , E ] , E = [ E , E ] . (B.2)In the basis h = 1 √ H , h = H + 2 H ,e = E , e = 1 √ E , e = 1 √ E ,e = 1 √ E , e = 16 E , e = 16 E ,f = F , f = 1 √ F , f = 1 √ F ,f = 1 √ F , f = 16 F , f = 16 F . (B.3)the positive roots take the following values: ~α = ( −√ , ~α = ( √ , ~α = ( − √ ,
1) = ~α + ~α , ~α = ( √ ,
1) = ~α + 2 ~α , ~α = ( √ ,
1) = ~α + 3 ~α , ~α = (0 ,
2) = 2 ~α + 3 ~α .We are interested in compactifying five-dimensional minimal supergravity over onetimelike and one spacelike Killing direction. When one first compactifies along a directionof signature ǫ and then along a direction of signature ǫ —where ǫ , take values +1 or − τ is given as (for a more detailed discussion see [38]): τ ( h ) = − h , τ ( h ) = − h ,τ ( e ) = − ǫ ǫ f , τ ( e ) = − ǫ f , τ ( e ) = − ǫ f ,τ ( e ) = − ǫ ǫ f , τ ( e ) = − ǫ f , τ ( e ) = − ǫ f . (B.4)The subalgebra of elements fixed under τ is not compact. It consists of all the elementsof the form { e i + τ ( e i ) } , that is, k = e + f , k = e + f , k = e − f ,k = e + f , k = e − f , k = e + f . (B.5)These elements generate the sl (2 , R ) ⊕ sl (2 , R ) algebra. C Representation of g ( ) and coset representative In this appendix we give a representation of g ( ) and a construction of symmetric cosetrepresentative M for the coset G / ( SL (2 , R ) × SL (2 , R )). The coset construction given34elow is largely based on the one used in [38], but it differs in one important aspect,namely that the final matrix M given below is symmetric — which is not the case in [38].In [38] the matrix M is symmetric under generalized transposition, but not under theusual transpose. The symmetric matrix given below is better suited for inverse scatteringconstructions. The representation of g ( ) we use is identical to the one used in [38].For completeness and for ease of reference here we present relevant details. We start bydefining the Chevalley generators: E = , E = ,F = , F = ,H = − − , H = − − − . Using these generators we can easily write the representation matrices for the rest of thegenerators. Matrices in the basis (B.3) can then be readily obtained.We write a coset representative V for the coset G / ( SL (2 , R ) × SL (2 , R )) in theBorel gauge by exponentiating the Cartan and positive root generators of g ( ) with thedilatons and axions as coefficients. We can make contact with the reduced Lagrangian(2.6) by choosing the coset representative to be [38] V = e φ h + φ h e χ e e − χ e + χ e e χ e e χ e − χ e . (C.1)35ext we introduce the following two matrices η = − − − − , S = √ −√ √ − √ √
00 0 √ , (C.2)and define the coset representative matrix M as M = S T V T η V S . (C.3)The matrix M is symmetric by construction. It can be easily checked by an explicitcalculation that the scalar part of the reduced Lagrangian (2.6) is given by L scalar = −
18 Tr (cid:0) ⋆ ( M − dM ) ∧ ( M − dM ) (cid:1) . (C.4)The matrix M has the symmetrical block structure [39, 40] M = A U BU T ˜ S V T B T V C , (C.5)where A and C are symmetrical 3 × B is a 3 × U and V are 3-component column matrices, and ˜ S a scalar. It also follows that the inverse matrix isgiven by [39, 40] M − = ( S T mS ) M ( S T mS ) = C − V B T − V T ˜ S − U T B − U A , (C.6)where m is the SO (3 + , − ) metric (D.14) for our representation. Explicit expressions forthe matrices A, B, C, U, V, ˜ S can be readily written. Since these expressions are exceedinglylong, we do not present them here. However, we note thatdet A = det C. (C.7) D Invariant 3-form and the basis of octonions for our rep-resentation Recall that one definition of the complex Lie algebra g is as the algebra of endomor-phisms of a seven dimensional vector space V preserving a general three-form on V . In This appendix is done in collaboration with Josef Lindman H¨ornlund and Jakob Palmkvist. c ( a ) abc for our repre-sentation . This invariant three-form provides a way to check if a 7 × M is inthe group G in our representation. If a matrix M ab preserves this three-form, i.e., if itsatisfies X a,b,c =1 c ( a ) abc M ae M bf M cg = c ( a ) efg (D.1)then it is in the group G .There are several ways to construct this three form. A systematic way of doing this for agiven representation of g is presented in e.g., [61]. However, in this section we do not followthe approach of [61]; instead, we provide the octonion description of our representation.The octonion structure constants ˜ c ( a ) abc are directly related to the invariant 3-form c ( a ) abc .The octonion description has the advantage of being more intuitive; and it also naturallyallows us to introduce the metric m ab on the seven dimensional vector space V .We start by recalling some basic facts about octonions. The octonions form a realeight dimensional division algebra O with a basis { e , e , e , . . . , e } where e spans thereal numbers. The seven imaginary units { e , e , . . . , e } anti-commute and square to − a = b , the product e a e b is given in a standard basis, e.g., in the basis given in [62],by e = e e = e e = e e , e = e e = e e = e e , e = e e = e e = e e , e = e e = e e = e e , e = e e = e e = e e , e = e e = e e = e e , e = e e = e e = e e . (D.2)The complex Lie algebra g is the derivation algebra of the octonions O . A basis of thederivation algebra of O is given as e L e L − e L e L , e L e L − e L e L , e L e L − e L e L , e L e L − e L e L , e L e L − e L e L , e L e L − e L e L , e L e L − e L e L , e L e L − e L e L , e L e L − e L e L , e L e L − e L e L , e L e L − e L e L , e L e L − e L e L , e L e L − e L e L , e L e L − e L e L , (D.3)where for any octonion e we define e L as the endomorphism of O which acts on the left as e L : o → eo for any o ∈ O . (D.4) The reason why we choose the notation c ( a ) abc for the invariant three-form will become clear in thefollowing. O . To obtain theChevalley generators of g , one takes the following linear combinations of the derivationelements (D.3) E = 14 [ i ( e L e L − e L e L ) − e L e L + e L e L ] ,F = 14 [ i ( e L e L − e L e L ) + e L e L − e L e L ] ,E = 14 [ i (2 e L e L − e L e L − e L e L ) − e L e L + e L e L + e L e L ] ,F = 14 [ i (2 e L e L − e L e L − e L e L ) + 2 e L e L − e L e L − e L e L ] ,H = 12 i ( e L e L − e L e L ) ,H = 12 i (2 e L e L − e L e L − e L e L ) . (D.5)Since all elements of the derivation algebra act trivially on the real numbers, we consideronly the action of the elements (D.3) on the seven dimensional subspace Im O spannedby the imaginary units { e , e , . . . , e } . This provides the unique (up to a change of basis)seven dimensional representation of g , the smallest non-trivial representation.In the basis { a , a , a , a , a , a , a , a } (D.6)where a = e a = i e + e a = e + i e a = i e − e a = − i e a = − i e − e a = − e + 2 i e a = − i e + 2 e (D.7)the 7 × . Aneasy way to see this is to write the matrix A that takes us from one basis to another.Columns of the matrix A are the a -basis vectors expressed in the e -basis, A = − i − i − i i i i − i − − . (D.8) Note the use of i ’s in the definitions (D.7). This is because we are working with the split real form of g : g ( ) . Alternatively, one can use a basis of split octonions instead of (D.2). Then the entries in thecorresponding matrix A , (D.8), would be all real. g ( ) are precisely A − T i A (D.9)where T i are the matrices obtained from (D.5). The octonion structure constants in the e -basis e a e b = X c =1 ˜ c ( e ) abc e c (D.10)and in the a -basis a a a b = X c =1 ˜ c ( a ) abc a c . (D.11)are related by the basis change˜ c ( a ) mnp = X a,b,c =1 A am A bn ˜ c ( e ) abc ( A − ) pc . (D.12)To construct the invariant three form we multiply the structure constants with the SO (3 + , − ) metric c ( a ) abc = X e =1 m ce ˜ c ( a ) abe . (D.13)The SO (3 + , − ) metric is obtained using the matrix Am = A T A = − − − . (D.14) References [1] P. Breitenlohner, D. Maison and G. W. Gibbons, “Four-Dimensional Black Holesfrom Kaluza-Klein Theories,” Commun. Math. Phys. , 295 (1988).[2] E. Cremmer, B. Julia, H. Lu and C. N. Pope, “Higher-dimensional origin of D = 3coset symmetries,” arXiv:hep-th/9909099.[3] D. C. Robinson, “Uniqueness of the Kerr black hole,” Phys. Rev. Lett. (1975) 905.[4] P. O. Mazur, “Proof Of Uniqueness Of The Kerr-Newman Black Hole Solution,” J.Phys. A (1982) 3173.[5] P. O. Mazur, “Black Hole Uniqueness From A Hidden Symmetry Of Einstein’s Grav-ity,” Gen. Rel. Grav. (1984) 211. 396] G. Bunting, “Proof of the uniqueness conjecture for black holes.” PhD Thesis, De-partment of Mathematics, University of New England, Armidale, N.S.W. 1983.[7] B. Carter, “Bunting Identity and Mazur Identity for Non-linear Elliptic SystemsIncluding the Black Hole Equilibrium Problem,” Commun. Math. Phys. (1985)563.[8] S. Hollands and S. Yazadjiev, “Uniqueness theorem for 5-dimensional black holes withtwo axial Killing fields,” Commun. Math. Phys. , 749 (2008) [arXiv:0707.2775 [gr-qc]].[9] S. Hollands and S. Yazadjiev, “A Uniqueness theorem for 5-dimensional Einstein-Maxwell black holes,” Class. Quant. Grav. , 095010 (2008) [arXiv:0711.1722 [gr-qc]].[10] S. Tomizawa, Y. Yasui and A. Ishibashi, “A uniqueness theorem for charged rotatingblack holes in five-dimensional minimal supergravity,” Phys. Rev. D , 124023 (2009)[arXiv:0901.4724 [hep-th]].[11] A. J. Amsel, G. T. Horowitz, D. Marolf and M. M. Roberts, “Uniqueness of ExtremalKerr and Kerr-Newman Black Holes,” arXiv:0906.2367 [gr-qc].[12] P. Figueras and J. Lucietti, “On the uniqueness of extremal vacuum black holes,”arXiv:0906.5565 [hep-th].[13] P. T. Chrusciel and J. Lopes Costa, “On uniqueness of stationary vacuum blackholes,” arXiv:0806.0016 [gr-qc].[14] J. L. Costa, “On the classification of stationary electro-vacuum black holes,”arXiv:0912.0834 [gr-qc].[15] S. Tomizawa, Y. Yasui and A. Ishibashi, “A uniqueness theorem for charged dipolerings in five-dimensional minimal supergravity,” arXiv:0911.4309 [hep-th].[16] J. Armas and T. Harmark, “Uniqueness Theorem for Black Hole Space-Times withMultiple Disconnected Horizons,” arXiv:0911.4654 [hep-th].[17] V. A. Belinski and V. E. Zakharov, “Integration Of The Einstein Equations By TheInverse Scattering Problem Technique And The Calculation Of The Exact SolitonSolutions,” Sov. Phys. JETP , 985 (1978) [Zh. Eksp. Teor. Fiz. , 1953 (1978)].V. A. Belinski, V. E. Sakharov, “Stationary Gravitational Solitons with Axial Sym-metry,” Sov. Phys. JETP , 521 (1978).[19] P. Breitenlohner and D. Maison, “On The Geroch Group,” Annales Poincare Phys.Theor. , 215 (1987). 4020] H. Nicolai, “Two-dimensional gravities and supergravities as integrable system,” inRecent Aspects of Quantum Fields, H. Mitter and H. Gausterer (eds.), Springer-Verlag, Berlin (1991).[21] H. Nicolai, D. Korotkin and H. Samtleben, “Integrable classical and quantum gravity,”arXiv:hep-th/9612065.[22] G. A. Alekseev, “N soliton solutions of the Einstein-Maxwell equations,” JETP Lett. , 377,G. Alekseev and V. Belinski, “Einstein-Maxwell Solitons,” AIP Conf. Proc. ,333-385, COSMOLOGY AND GRAVITATION: XIII Brazilian School on Cosmologyand Gravitation (XIII BSCG), and references therein.[23] R. Emparan and H. S. Reall, “A rotating black ring in five dimensions,” Phys. Rev.Lett. , 101101 (2002) [arXiv:hep-th/0110260].[24] R. Emparan, “Rotating circular strings, and infinite non-uniqueness of black rings,”JHEP , 064 (2004) [arXiv:hep-th/0402149].[25] H. Elvang, R. Emparan, D. Mateos and H. S. Reall, “A supersymmetric black ring,”Phys. Rev. Lett. , 211302 (2004) [arXiv:hep-th/0407065].[26] R. Emparan and H. S. Reall, “Black rings,” Class. Quant. Grav. , R169 (2006)[arXiv:hep-th/0608012].[27] R. Emparan and H. S. Reall, “Black Holes in Higher Dimensions,” Living Rev. Rel. , 6 (2008) [arXiv:0801.3471 [hep-th]].[28] H. Elvang, R. Emparan and P. Figueras, “Non-supersymmetric black rings as ther-mally excited supertubes,” JHEP , 031 (2005) [arXiv:hep-th/0412130].[29] A. A. Pomeransky and R. A. Sen’kov, “Black ring with two angular momenta,”arXiv:hep-th/0612005.[30] S. D. Mathur, “The fuzzball proposal for black holes: An elementary review,” Fortsch.Phys. , 793 (2005) [arXiv:hep-th/0502050].I. Bena and N. P. Warner, “Black holes, black rings and their microstates,” Lect.Notes Phys. , 1 (2008) [arXiv:hep-th/0701216].K. Skenderis and M. Taylor, “The fuzzball proposal for black holes,” Phys. Rept. , 117 (2008) [arXiv:0804.0552 [hep-th]].V. Balasubramanian, J. de Boer, S. El-Showk and I. Messamah, “Black Holes asEffective Geometries,” Class. Quant. Grav. , 214004 (2008) [arXiv:0811.0263 [hep-th]].[31] J. de Boer, S. El-Showk, I. Messamah and D. Van den Bleeken, “A bound on theentropy of supergravity?,” arXiv:0906.0011 [hep-th].[32] V. Jejjala, O. Madden, S. F. Ross and G. Titchener, “Non-supersymmetric smoothgeometries and D1-D5-P bound states,” Phys. Rev. D , 124030 (2005) [arXiv:hep-th/0504181]. 4133] S. Giusto, S. F. Ross and A. Saxena, “Non-supersymmetric microstates of the D1-D5-KK system,” JHEP , 065 (2007) [arXiv:0708.3845 [hep-th]].[34] J. H. Al-Alawi and S. F. Ross, “Spectral Flow of the Non-Supersymmetric Microstatesof the D1-D5-KK System,” arXiv:0908.0417 [hep-th].[35] I. Bena, S. Giusto, C. Ruef and N. P. Warner, “A (Running) Bolt for New Reasons,”arXiv:0909.2559 [hep-th].[36] N. Bobev and C. Ruef, “The Nuts and Bolts of Einstein-Maxwell Solutions,”arXiv:0912.0010 [hep-th].[37] J. P. Gauntlett, J. B. Gutowski, C. M. Hull, S. Pakis and H. S. Reall, “All supersym-metric solutions of minimal supergravity in five dimensions,” Class. Quant. Grav. ,4587 (2003) [arXiv:hep-th/0209114].[38] G. Compere, S. de Buyl, E. Jamsin and A. Virmani, “G2 Dualities in D=5 Supergrav-ity and Black Strings,” Class. Quant. Grav.
12 (2009) arXiv:0903.1645 [hep-th].[39] A. Bouchareb, G. Clement, C. M. Chen, D. V. Gal’tsov, N. G. Scherbluk and T. Wolf,“ G generating technique for minimal D=5 supergravity and black rings,” Phys. Rev.D , 104032 (2007) [Erratum-ibid. D , 029901 (2008)] [arXiv:0708.2361 [hep-th]].[40] G. Clement, “The symmetries of five-dimensional minimal supergravity reduced tothree dimensions,” J. Math. Phys. , 042503 (2008) [Erratum-ibid. , 079901(2008)] [arXiv:0710.1192 [gr-qc]].[41] P. Breitenlohner and D. Maison, “On nonlinear sigma-models arising in (super) grav-ity,” Commun. Math. Phys. , 785 (2000) [arXiv:gr-qc/9806002].[42] S. Mizoguchi and N. Ohta, “More on the similarity between D = 5 simple supergravityand M theory,” Phys. Lett. B , 123 (1998) [arXiv:hep-th/9807111].[43] S. Giusto and A. Saxena, “Stationary axisymmetric solutions of five dimensionalgravity,” Class. Quant. Grav. , 4269 (2007) [arXiv:0705.4484 [hep-th]].[44] R. M. Wald, “General Relativity,” Chicago, Usa: Univ. Pr. ( 1984) 491p [45] R. Emparan and H. S. Reall, “Generalized Weyl solutions,” Phys. Rev. D , 084025(2002) [arXiv:hep-th/0110258].[46] H. Lu, M. J. Perry and C. N. Pope, “Infinite-Dimensional Symmetries of Two-Dimensional Coset Models,” arXiv:0711.0400 [hep-th].H. Lu, M. J. Perry and C. N. Pope, “Infinite-Dimensional Symmetries of Two-Dimensional Coset Models Coupled to Gravity,” Nucl. Phys. B , 656 (2009)[arXiv:0712.0615 [hep-th]].[47] J. H. Schwarz, “Classical symmetries of some two-dimensional models,” Nucl. Phys.B , 137 (1995) [arXiv:hep-th/9503078].4248] D. Bernard and B. Julia, “Twisted self-duality of dimensionally reduced gravity andvertex operators”, Nucl. Phys. B547 ,427 (1999) [arXiv:hep-th/9712254].[49] A. Kleinschmidt and H. Nicolai, “Gradient representations and affine structures”,Class. Quant. Grav. ,4457 (2005) [arXiv:hep-th/0506238].[50] V. G. Kac, “Infinite dimensional Lie algebras,” Cambridge, UK: Univ. Pr. (1990)400 p [51] J. Fuchs and C. Schweigert, “Symmetries, Lie Algebras And Representations: AGraduate Course For Physicists,”
Cambridge, UK: Univ. Pr. (1997) 438 p [52] A. A. Pomeransky, “Complete integrability of higher-dimensional Einstein equationswith additional symmetry, and rotating black holes,” Phys. Rev. D , 044004 (2006)[arXiv:hep-th/0507250].[53] I. Bakas, “O(2,2) transformations and the string Geroch group,” Nucl. Phys. B ,374 (1994) [arXiv:hep-th/9402016].[54] S. S. Yazadjiev, “Completely integrable sector in 5D Einstein-Maxwell gravity andderivation of the dipole black ring solutions,” Phys. Rev. D , 104007 (2006)[arXiv:hep-th/0602116].[55] R. P. Geroch, “A Method for generating new solutions of Einstein’s equation. 2,” J.Math. Phys. , 394 (1972).[56] A. Sen, “Duality Symmetry Group Of Two-Dimensional Heterotic String Theory,”Nucl. Phys. B , 62 (1995) [arXiv:hep-th/9503057].[57] M. Cvetic and D. Youm, “General Rotating Five Dimensional Black Holes ofToroidally Compactified Heterotic String,” Nucl. Phys. B , 118 (1996) [arXiv:hep-th/9603100].[58] M. Cvetic and F. Larsen, “General rotating black holes in string theory: Greybodyfactors and event horizons,” Phys. Rev. D , 4994 (1997) [arXiv:hep-th/9705192].[59] T. Harmark, “Stationary and axisymmetric solutions of higher-dimensional generalPhys. Rev. D , 124002 (2004) [arXiv:hep-th/0408141].[60] J. E. Humphreys, “Introduction to Lie algebras and representation theory,” (3rdprint., rev.), New York, USA: Springer (1980).[61] William Fulton and Joe Harris, “Representation Theory: A First Course,” GraduateTexts in Mathematics , 1991 Springer, New York, section 22.3.[62] J. C. Baez, “The Octonions,” Bull. Amer. Math. Soc. (2002), 145-205; Errata inBull. Amer. Math. Soc.42