Horizon area-angular momentum-charge-magnetic fluxes inequalities in 5D Einstein-Maxwell-dilaton gravity Stoytcho Yazadjiev∗ 3 1 0 2 Department of Theoretical Physics, Faculty of Physics, Sofia University n 5 J. Bourchier Blvd., Sofia 1164, Bulgaria a J 8 ] h Abstract t - p In the present paper we consider 5D spacetimes satisfying the Einstein-Maxwell- e h dilaton gravity equations which areU(1)2 axisymmetric but otherwise highly dynami- [ cal. We derive inequalities between the area, the angular momenta, the electric charge 1 andthemagneticfluxesforanysmoothstablyoutermarginally trapped surface. v 8 4 1 Basic notions and setting the problem 5 1 . 1 Thestudyofinequalitiesbetweenthehorizonareaandtheothercharacteristicsofthehorizon 0 has attracted a lot of interest recently. Within the general theory of relativity, lower bounds 3 1 for the area of dynamical horizons in terms of their angular momentum or/and charge were v: givenin[1]–[8], generalizingthesimilarinequalitiesforthestationaryblackholes[9]–[11]. i Theseremarkableinequalitiesarebasedsolelyongeneralassumptionsandtheyholdforany X axisymmetricbutotherwisehighlydynamicalhorizoningeneralrelativity. Foranicereview r a on the subject we refer the reader to [12]. The relationship between the proofs of the area- angular-momentum-chargeinequalitiesforquasilocalblackholesandstationaryblackholes isdiscussedin[13]-[15]. Inequalitiesbetweenthehorizonarea,theangularmomentum,and thecharges werealso studiedin some4D alternativegravitationaltheories[16]. Ageneralizationofthe4Dhorizonarea-angularmomentuminequalitytoD-dimensional vacuum Einstein gravity with U(1)D−3 group of spatial isometries was given in [17]. The purpose of the present work is to derive some inequalities between the horizon area, hori- zon angular momentum, horizon charges and magnetic fluxes in the 5D Einstein-Maxwell- dilaton gravity including as a particular case the 5D Einstein-Maxwell gravity. It should be stressedthatthederivationofthementionedinequalities inthehigherdimensionalEinstein- Maxwelland Einstein-Maxwell-dilatongravityismuchmore difficultand isnotsostraight- forwardas inthehigherdimensionalvacuumgravityeveninspacetimesadmittingU(1)D−3 ∗[email protected] 1 isometry group. The main reason behind this is the lack of nontrivial group of hidden sym- metries for the dimensionally reduced Einstein-Maxwell-dilaton equations in the general case 1[18]. In contrast, the dimensionally reduced vacuum Einstein equations (in space- timeswithU(1)D−3isometrygroup)possessnontrivialgroupofhiddensymmetries,namely SL(D−2,R)andamatrixsigmamodelpresentationispossible. Someofthedifficultiesdue tothepresenceofaMaxwellfieldcanbecircumventedbyfollowingamethodsimilartothat usedin the4D Einstein-Maxwell-dilatongravity[16]as weshowbelow. Let (M,g ,F ,j ) be a 5-dimensional spacetime satisfying the Einstein-Maxwell- ab ab dilatonequations g G =2¶ ¶j j −(cid:209) c(cid:209)j j g −2V(j )g +2e−2ja F F c− abF Fcd , (1) ab a b c ab ab ac b cd 4 (cid:209) e−2ja Fab =0=(cid:209) F , (cid:16) (cid:17) (2) a [a bc] (cid:16) dV((cid:17)j ) a (cid:209) (cid:209) aj = − e−2ja F Fcd, (3) a dj 2 cd where g is the spacetime metric, (cid:209) is its Levi-Civita connection, G = R − 1g R is ab a ab ab 2 ab the Einstein tensor and F is the Maxwell field. The dilaton field is denoted by j , V(j ) is ab its potentialand a is the dilatoncoupling parameter. We assumethat the dilatonpotential is non-negative,V(j ) ≥ 0. The Einstein-Maxwell gravity is recovered by first putting a = 0 andV(j )=0 and thenj =0. As an additional technical assumption we require the spacetime to admit U(1)2 group of spatial isomerties. The commuting Killing fields are denoted by h and h and they are 1 2 normalized to have a period 2p . We also require the Maxwell and the dilaton fields to be invariantundertheflowoftheKillingfields, i.e. £h F =£h j =0. I I LetusfurtherconsideracompactclosedsmoothsubmanifoldB ofdimensiondimB =3 invariantundertheactionofU(1)2. TheinducedmetriconB anditsLevi-Civitaconnection aredenotedbyg andD ,respectively. ThefuturedirectednullnormalstoB willbedenoted ab a by n and l with the normalization condition g(n,l)=−1 and with −l pointing outward. In whatfollowswerequire B to beastablyoutermarginallytrapped surfacewhich means that Q =0and £ Q ≤0with Q beingtheexpansionofn onB. n l n n As a 3-dimensionalcompact manifoldwith an actionofU(1)2, B is topologicallyeither S3, S2×S1 oralens space L(p,q) with p and q being co-prime integers[22, 23]. Moreover, the factor space Bˆ = B/U(1)2 can be identified with the closed interval [−1,+1]. As it was shown in [22, 23], certain linear combinations of the Killing fields h , with integer I coefficients,vanishattheendsofthefactorspace. Inotherwords,thereexistintegervectors a ∈Z2 such that aI h →0 at x=±1, where x is the coordinate parameterizing the factor ± ± I space. Equivalently,theGram matrixdefined by H =g(h ,h ) (4) IJ I J isinvertibleintheinterioroftheinterval [−1,1]andhasone-dimensionalkernelattheinter- valend points,i.e. H aI →0at x=±1. IJ ± Infact,theintegervectorsa determinethetopologyofB. ByaglobalSL(2,Z)redefini- ± tionoftheKillingfields[22,23]wemaypresenta intheform a =(1,0)anda =(p,q) ± + − 1Fortunately,therearesectorsinEinstein-Maxwell-gravitywhicharecompletelyintegrable[19]-[21]. 2 with p and q being coprime integers. The topology of B is then S3 when (p=±1,q =0), S2×S1 when (p=0,q=±1) and thatofalens spaceL(p,q)in theothercases. Proceeding further we consider a small neighborhood O of B. When the neighborhood is sufficiently small it can be foliated by two-parametric copies B(u,r) of B =B(0,0) and parameterized by the so-called null Gauss coordinates defined by a well-known procedure [24]. In Gauss nullcoordinatesthemetricin O can bewrittenin theform g=−2du(dr−r2¡ du−rb dya)+g dyadyb, (5) a ab where n = ¶ , l = ¶ , and the function ¡ and the metric g are invariantly defined on each ¶ u ¶ r B(u,r). Usingthesecoordinatesonecan showthaton B it holds 1 Rg −Dab a− b ab a−2Gabnalb =−2£lQ n ≥0, (6) 2 where it has been taken into account that Q n =0 on B. Here Rg and Da are the Ricci scalar curvatureandLevi-Civitaconnectionwithrespecttothemetricg onB. Takingintoaccount ab thatthedilatonpotentialisnonnegativethisinequalitycan berewrittenintheform 1 Rg −Dab a− b ab a−2G˜abnalb =2V(j )−2£lQ n ≥0, (7) 2 where G˜ =G +2V(j )g . ab ab ab Making use of (7), for every axisymmetric function f (i.e. every function f invariant undertheisometrygroup)wehave 1 0≤ −Dab a− b ab a+Rg −2G˜abnalb f2dS ZB 2 (cid:18) (cid:19) 1 = 2fb aDaf − b ab af2+Rg f2−2G˜abnalbf2 dS, (8) ZB 2 (cid:18) (cid:19) where dS is the surface element on B. Now we consider the unit tangent vector Na on B whichis orthogonalto h . Withits help and takingintoaccount that g ab−NaNb D f =0, I b we find b b a = g ab−NaNb b b +(Nab )2 and 2fb aD f = 2(fNab ) NbD f which a a b a a (cid:0) a (cid:1)b gives (cid:0) (cid:1) (cid:0) (cid:1) 1 1 0≤ 2(fNab a) NbDbf − (Nab a)2 f2− g ab−NaNb b ab bf2+Rg f2−2G˜abnalbf2 dS. (9) ZB 2 2 (cid:20) (cid:21) (cid:16) (cid:17) (cid:16) (cid:17) Finally,takingintoaccountthat 1 2 2(fNab ) NbD f − (Nab )2 f2 ≤2 NbD f (10) a b a b 2 (cid:16) (cid:17) (cid:16) (cid:17) 3 and that NbD f 2 = g ab−NaNb D fD f +NaNbD fD f =g abD fD f =D fDaf we b a b a b a b a obtaintheimportantinequality (cid:0) (cid:1) (cid:0) (cid:1) 1 0≤ 2DafDaf − g ab−NaNb b ab bf2+Rg f2−2G˜abnalbf2 dS. (11) ZB 2 (cid:20) (cid:21) (cid:16) (cid:17) In order to extract the constructive information from this inequality we should perform a dimensionalreduction and express the inequalityas an inequalityon the factor space Bˆ = B/U2(1) = [−1,1]. The dimensional reduction can be performed along the lines of [18]. So we shall give here only some basic steps and results without going into detail. As a first stepitisveryconvenienttopresenttheKillingfieldsh inadaptedcoordinates,i.e. h = ¶ I I ¶f I wherethecoordinates f I are 2p -periodic. Then theinduced metric g onB takestheform ab dx2 g = +H df Idf J, (12) C2h IJ whereC>0isaconstantandh=det(H ). Theabsenceofconicalsingularitiesrequiresthe IJ followingconditiontobesatisfied h H aI aJ lim C2 IJ ± ± =1. (13) x→±1 1−x2 1−x2 Thearea A ofB can beeasily foundfrom (12)andtheresultis A =8p 2C−1. (14) Thereforethecondition(13)can berewrittenin theform A h H aI aJ 1/4 h H aI aJ 1/4 = lim IJ + + lim IJ − − . (15) 8p 2 x→1 1−x2 1−x2 x→−1 1−x2 1−x2 (cid:18) (cid:19) (cid:18) (cid:19) Since the factor space O/U(1)2 is simply connected we can introduce electromagnetic potentialsF and Y invariantundertheisometrygroup anddefined by I dF I =ih F, dY =e−2ja ih ih ⋆F. (16) I 2 1 TheMaxwell2-formcan thenbewrittenin theform F =HIJh ∧dF +h−1e2ja ⋆(dY ∧h ∧h ). (17) I J 1 2 Using the field equations one can show that there exist potentials c invariant under the I isometrygroup suchthatthetwist w =⋆(h ∧h ∧dh ) satisfies I 1 2 I 4 w =dc +2F dY −2Y dF . (18) I I I I By direct computation of the twist using the metric (5) one finds that on B it holds b = I ih b =Ci¶ w I orinexplicitform I ¶ x b =¶ c +2F ¶ Y −2¶Y F . (19) I x I I x x I Alsoonecan showthaton B wehave G˜ nalb =D j Daj +2e−2ja HIJD F DaF +h−1e2ja D Y DaY . (20) ab a a I J a Usingtheexplicitform(12)ofthemetricinduced on B, by direct computationwefind Rg =C2h −¶ 2xh+1h−2(¶ xh)2−1Tr H−1¶ xH 2 . (21) h 4 4 (cid:20) (cid:21) (cid:0) (cid:1) Finally,choosing 1−x2 1/2 f = , (22) h (cid:18) (cid:19) substituting(19),(20)and(21)into(11)andtakingintoaccountthatdS=C−1dx(cid:213) df I,we I obtain 1 (1−x2) 1Tr H−1¶ H 2+1h−2(¶ h)2+ x x Z 8 8 −1 (cid:26) (cid:20) 1 (cid:0) (cid:1) h−1HIJ(¶ c +2F ¶ Y −2¶Y F )(¶ c +2F ¶ Y −2¶Y F ) x I I x x I x J J x x J 4 1 +e−2ja HIJ¶ F ¶ F +e2ja h−1(¶ Y )2+(¶ j )2 − dx≤0. (23) x I x J x x 1−x2 (cid:27) (cid:3) Nowwecan introducethestrictlypositivedefinitemetric G givenby AB G dXAdXB = 1Tr H−1dH 2+1h−2(dh)2+ AB 8 8 1h−1HIJ(dc +2F (cid:0)dY −2Y (cid:1)dF )(dc +2F dY −2Y dF )+ I I I J J J 4 e−2ja HIJdF dF +e2ja h−1(dY )2+(dj )2 (24) I J onthe9-dimensionalmanifoldN ={(H (I≤J),c ,F ,Y ,f )∈R9;h>0}. Intermsofthis IJ I I metrictheinequality(23)takes theform 5 1 dXAdXB 1 I [XA]= (1−x2)G − dx≤0. (25) ∗ Z AB dx dx 1−x2 −1 (cid:20) (cid:21) Inordertotransformthisinequalityintoaninequalityfor thearea,weusecondition(15) whichcombinedwith(25)gives A ≥8p 2eI[XA], (26) where 1 h H aI(x)aJ(x) I[XA]=I [XA]+ xln IJ |x=1 (27) ∗ 4 1−x2 1−x2 x=−1 (cid:20) (cid:21) with aI(x) defined by aI(x) = 1(1+x)aI + 1(1−x)aI . We should note that there is an 2 + 2 − ambiguityindefining thefunctionalI[XA]. Forexample,wecan define itby 1 h H aI(x)aJ(x) I[XA]=aI [XA]+ xln IJ |x=1 , (28) ∗ 4 1−x2 1−x2 x=−1 (cid:20) (cid:21) where a is an arbitrary positive number. This ambiguity, however, does not affect the final resultssinceI [XA]=0 as weshowbelow. ∗ 2 Minimizer existence lemma In orderto put a lowerbound on the area we should find the minimumofthe function I[XA] withappropriate boundaryconditionsifthe minimumexists. Below we showthat in certain casestheminimumexists. Thenaturalclassoffunctionsfortheminimizingproblemisgiven by s = −ln h ∈C¥ [−1,1], ln HIJaIaJ ∈C¥ [−1,1], (c ,F ,Y ,j ) ∈C¥ [−1,1] with 1−x2 1−x2 I I boundary co(cid:16)ndition(cid:17)s s (±1) = s ±, ((cid:16)c (±1),(cid:17)F (±1),Y (±1),j (±1)) = (c ±,F ±,Y ±,j ±). I I I I Since the electromagnetic potentials and the twist potential are defined up to a constant, withoutlossofgeneralitywecan choose c + =−c −, F + =−F −, Y + =−Y −. (29) I I I I Lemma 1. For dilaton coupling parameter satisfying 0 ≤ g 2 ≤ 8, there exists a unique 3 smoothminimizerofthefunctionalI[XA]with theprescribed boundaryconditions. Proof. Let usconsiderthetruncated functional x2 dXAdXB 1 I [XA][x ,x ]= (1−x2)G − dx (30) ∗ 2 1 Z AB dx dx 1−x2 x 1 (cid:20) (cid:21) 6 with−1<x <x <1. Byintroducinganewvariablet = 1ln 1+x thetruncatedfunctional 1 2 2 1−x takestheform (cid:0) (cid:1) t2 dXAdXB I [XA][t ,t ]= G −1 dt, (31) ∗ 2 1 AB Z dt dt t 1 (cid:20) (cid:21) which is just a modified version of the geodesic functional in the Riemannian space (N,G ). Consequently the critical points of the functional are geodesics in N. It was AB shown in [18] that for 0 ≤ a 2 ≤ 8 the Riemannian space (N,G ) is simply connected, 3 AB geodesically complete and with negative sectional curvature. Therefore, for fixed points XA(t ) and Xa(t ) thereexista uniqueminimizinggeodesicconnectingthese points. There- 1 2 fore the global minimizer of I [XA][t ,t ] exists and is unique for 0 ≤ a 2 ≤ 8. Since ∗ 2 1 3 (N,G ) is geodesically complete the global minimizer of I [XA][t ,t ] can be extended AB ∗ 2 1 to a global minimizer of I [XA]. Indeed, let us take x (e )= −1+e and x (e ) = 1−e (i.e. ∗ 1 2 t (e )=−t (e )= 1ln e )withe beingasmallpositivenumberandconsiderthetruncated 1 2 2 2−e functional (cid:0) (cid:1) Ie [XA]= x2(e ) (1−x2)G dXAdXB − 1 dx (32) ∗ Zx1(e ) (cid:20) AB dx dx 1−x2(cid:21) with boundary conditions XA(x (e )) and XA(x (e )). Consider now the unique minimizing 1 2 geodesicG e in N between thepointsXA(x1(e ))and XA(x2(e )). Then wehave Ie [XA]≥Ie [XA]|G (33) ∗ ∗ e where the right hand side of the above inequality is evaluated on the geodesic G e . Taking intoaccountthat l 2e =GABddXtAddXtB isconstanton thegeodesicG e , weobtain I∗e [XA]|G e =Zt1t(2e()e )(cid:20)GABddXtAddXtB −1(cid:21)dt = l 2e −1 (t2(e )−t1(e )). (34) (cid:0) (cid:1) Our next step is to evaluate l 2 and this can be done by evaluating G dXAdXB at the e AB dt dt boundary points which are in a small neighborhood of the poles x =±1. For this purpose wefirst write l 2 in theform e (1−x2)2 dH 2 (1−x2)2 dh 2 l 2 = Tr H−1 + h−2 (35) e 8 dx 8 dx (cid:18) (cid:19) (cid:18) (cid:19) (1−x2)2 dc dY dF dc dY dF + h−1HIJ I +2F −2Y I J +2F −2Y J I J 4 dx dx dx dx dx dx (cid:18) (cid:19)(cid:18) (cid:19) dF dF dY 2 dj 2 +(1−x2)2e−2ja HIJ I J +(1−x2)2e2ja h−1 +(1−x2)2 . dx dx dx dx (cid:18) (cid:19) (cid:18) (cid:19) 7 Withintheclass offunctionsweconsider,wehave (1−x2)2 dh 2 1 h−2 = +O(e ) (36) 8 dx 2 (cid:18) (cid:19) inasmallneighborhoodofthepoles. InordertoestimatethetermassociatedwithH wetakeintoaccountthatH−1dH satisfies dx itsowncharacteristicequation,namely Tr H−1dH 2 =h−2 dh 2−2h−1detdH. Hence we dx dx dx find (cid:0) (cid:1) (cid:0) (cid:1) (1−x2)2 dH 2 1 Tr H−1 = +O(e ). (37) 8 dx 2 (cid:18) (cid:19) Proceeding further we notice that ¶ /c¶ are Killing fields for the metric G and conse- I AB quentlywe havethefollowingconstantsofmotionon thegeodesicsG e 1 dc dY dF 1−x2 dc dY dF h−1HIJ I +2F I −2Y I = h−1HIJ I +2F I −2Y I =cIe . 2 dt dt dt 2 dx dx dx (cid:18) (cid:19) (cid:18) (cid:19) (38) Henceweobtain (1−x2)2 dc dY dF dc dY dF h−1HIJ I +2F −2Y I J +2F −2Y J = I J 4 dx dx dx dx dx dx (cid:18) (cid:19)(cid:18) (cid:19) hH cIcJ =O(e ). (39) IJ e e Fortheremainingterms,itis easy toseethat theybehaveas dF dF (1−x2)2e−2ja HIJ I J =O(e ), (40) dx dx dY 2 (1−x2)2e2ja h−1 =O(e ), (41) dx (cid:18) (cid:19) dj 2 (1−x2)2 =O(e 2). (42) dx (cid:18) (cid:19) Summarizingtheresultsso far, weconcludethatthebehaviorof l 2 forsmalle is e l 2 =1+O(e ). (43) e Thereforewehave 8 limIe [XA]|G =0 (44) e →0 ∗ e which,inviewof(33), gives I [XA]= limIe [XA]≥0. (45) ∗ e →0 ∗ Therefore, there exists a unique global minimizer of the functional I [XA]. Since the ∗ functionals I[XA] and I [XA] differ in boundary terms the global minimizerof I [XA] is also ∗ ∗ aglobalminimizerof I[XA]. Thiscompletestheproof. It shouldbenoted thatfrom (25)and (45)immediatelyfollowsthat I [XA]=0. ∗ The extremalstationary nearhorizon geometry is in fact defined by thesamevariational problem with the same boundary conditions and by the same class of functions. Therefore, as an directconsequenceoftheprovenlemmaweobtainthefollowing Corollary. Foreverydilatoncouplingparametera intherange0≤a 2 ≤ 8 theareaA ofB 3 satisfiestheinequality A ≥A , (46) ENHG where A is the area associated with the extremal stationary near horizon geometry of ENHG Einstein-Maxwell-dilaton gravity with V(j ) = 0, for the corresponding a . The equality is saturated only for the area associated with extremal stationary near horizon geometry with V(j )=0. 3 Horizon area-angular momenta-charge-magnetic fluxes inequality for critical dilaton coupling parameter For the critical coupling a 2 = 8 the Riemannian space (N,G ) is an SL(4,R)/O(4) sym- 3 AB metric space [16] and therefore, there exists a matrix M such that the metric G can be AB writtenintheform G dXAdXB = 1Tr M−1dM 2, (47) AB 8 where M is positivedefinite and M ∈SL(4,R). (cid:0)Finding th(cid:1)e explicitform of the matrix M is atedioustaskand herewepresent onlythefinal result. Thematrix M is givenby E 0 N 0 E S N NS M = 2×2 2×2 = , (48) ST E 0 Y 0 E STN STNS+Y 2×2 2×2 (cid:18) (cid:19)(cid:18) (cid:19)(cid:18) (cid:19) (cid:18) (cid:19) where E is the unit 2×2 matrix and S, N and Y are 2×2 matrices which have the 2×2 followingexplicitform: 2F 2F 1 2 S= , (49) c +2F Y c +2F Y 1 1 2 2 (cid:18) (cid:19) 9 N =e 23j h−1 e−4 23j h+4Y 2 −2Y , (50) q q−2Y 1 ! 2j Y =e 3 H. (51) q In termsofthematrix M, theEuler-Lagrangeequationsare d dM (1−x2)M−1 =0. (52) dx dx (cid:20) (cid:21) Henceweobtain dM (1−x2)M−1 =2A, (53) dx where A isaconstantmatrixwith TrA=0, sincedetM =1. Integratingfurtherwefind 1+x M =M exp ln A (54) 0 1−x (cid:18) (cid:19) withM beingaconstantmatrixwiththesamepropertiesas M andsatisfyingATM =M A. 0 0 0 As a positive definite matrix, M can be written in the form M = BBT for some constant 0 0 matrix B with |detB| = 1 and this presentation is up to an orthogonal matrix O, i.e it is invariant under the transformation B −→ BO. This freedom can be used to diagonalize the symmetricmatrix BTABT−1. So wecan takeBTABT−1 =diag(l ,l ,l ,l )and weobtain 1 2 3 4 l 1+x 1 0 0 0 1−x l M =B (cid:0) 0(cid:1) 11+−xx 2 0 0 BT. (55) l 0 0 1+x 3 0 (cid:0) (cid:1) 1−x l 0 0 0 1+x 4 (cid:0) (cid:1) 1−x The eigenvalues l can be found by comparing the s(cid:0)ingul(cid:1)ar behavior of the left and the i righthandsideof(55)atx→±1. TakingintoaccountthatonlythematrixN inM issingular at x→±1, we find that l =1,l =−1,l =l =0. Even more, if we write the matrix B 1 2 3 4 inblockform B R 1 B= , (56) L B 2 (cid:18) (cid:19) where B , B ,R and L are2×2 matrices,from thesingularbehaviorat x→±1 wefind 1 2 10