The M5-brane on K3 × T2 Neil Lambert1 8 0 Department of Mathematics 0 2 King’s College n The Strand a J London 0 1 WC2R 2LS, UK ] h t - p e Abstract h [ 2 We discuss the low energy effective theory of an M5-brane wrapped on v a smooth holomorphic four-cycle of K3 × T2, including the special case of 6 6 T6. In particular we give the lowest order equations of motion and resolve a 1 puzzle concerning the counting of massless modes that was reported in hep- 3 th/9906094. In order to find agreement with black hole entropy and anomaly . 2 inflow arguments we propose that some of the moduli become massive. 1 7 0 : v i X r a [email protected] 1 1 Introduction One of the most exciting achievements in string theory is the remarkable suc- cess in counting microscopic counting of black hole states, starting with the work of [1]. A particularly elegant example of this is provided by considering an M5-brane wrapped on a complex four-cycle of a Calabi-Yau [2]. This yields a black string in five dimensions which can be further reduced to four dimensions by wrapping the string on S1 and including momentum along the S1. A notable feature of this analysis is that, for a generic four-cycle, the M5-brane has a smooth worldvolume and hence the only microscopic infor- mation needed is a knowledge of the worldvolume fields and dynamics of a single M5-brane. There have been several detailed accounts of the M5-brane wrapped on cyclesofagenericCalabi-Yaumanifold, forexamplesee[3,4,5,6,7],andalso the M5-brane on K3 [8]. The case that we are interested in here concerns an M5-brane whose worldvolume has non-trivial one-cycles which occurs when the Calabi-Yau degenerates to K3 ×T2 or T6 (see also [7]). This situation was discussed in [9] where several puzzles arose. In particular the number of massless states was not found to be in accordance with (0,4) supersymmetry andthecounting ofblackholemicrostatesfailed(albeitatsub-leadingorder). To resolve these problems the authors of [9] proposed a novel mechanism whereby some massless modes are charged with respect to the worldvolume gauge fields that arise from reduction of the two-form. The main purpose of this paper is to investigate this proposal. However we find that the correct resolution comes from including additional massless modes which are present when the Calabi-Yau is K3×T2 or T6. The rest of this paper is organized as follows. In section two we present the lowest order equations of motion for an M5-brane which is wrapped on a smooth cycle P in spacetime. In section three we consider in detail the case where spacetime is of the form M = R1,4 × K3 × T2 and M = R1,4 ×T6. We provide a careful counting of the normal bundle moduli and resolve a puzzle concerning (0,4) supersymmetry that was observed in [9]. In section four we consider four-dimensional black hole states that arise by further compactification on S1. We find that, using our analysis, the usual counting of left-moving massless modes to determine black hole entropy does not agree with the supergravity calculations or arguments using anomalies. To resolve this discrepancy we propose that h (P) (4,4) multiplets must 1,0 become massive and hence do not appear in the low energy effective action. This provides an alternative resolution to a second puzzle discussed in [9]. Finally section five contains a brief conclusion. 2 2 Lowest Order Equations of Motion Covariant equations of motion of the M5-brane were first derived in [10]. We will not need give the full non-linear form of these equations, however it will be enlightening to give the lowest order equations (in terms of a derivative expansion). We will work in static gaugewhere the six coordinates xµ, µ,ν = 0,1,2,...,5, of the M5-brane worldvolume are identified with the first six coordinates of spacetime. The massless fields consist of 5 scalars XA, A,B = 6,7,8,...,10, a two-form B and a Fermion ψ which satisfies Γ ψ = −ψ. µν 012345 Here we use a full 32-component spinor of SO(1,10). It will be sufficient to work at the lowest order in the fields. XA represents the coordinates of the M5-braneinthetransversespaceandinparticularXA = 0correspondstothe M5-branewrappedonacalibratedsubmanifold. WeuseM,N = 0,1,2,...,10 to denote all eleven coordinates. We use an underline to denote tangent space indices. We will use a hat to denote eleven-dimensional quantities, the spacetime is denoted by M and the M5-brane worldvolume by W. First recall the case where the worldvolume W admits a chiral Killing spinor ǫ; D ǫ = 0, Γ ǫ = ǫ. For example if M = R1,4 × K3 × T2 and µ 012345 W = R1,1×K3. To lowest order in fluctuations, the equations of motion are just that of a free theory on a curved background D2XA = 0 iΓµD ψ = 0 (1) µ 1 H = ǫ Hρστ, µνλ µνλρστ 3! where H = 3∂ B and ǫ is totally antisymmetric with ǫ012345 = 1. µνλ [µ νλ] µνλρστ These equations are invariant under the supersymmetry transformations δXA = iǫ¯ΓAψ δB = iǫ¯Γ ψ (2) µν µν 1 δψ = ∂ XAΓµΓ ǫ+ ΓµνλH ǫ. µ A µνλ 2·3! Next we consider the case where the spacetime M admits a chiral covari- antly constant spinor ǫˆ, Dˆ ˆǫ = 0, Γ ǫˆ= ǫˆ but where this does not de- M 012345 scend toaKilling spinoronW. ForexamplewecantakeM = R1,4×K3×T2 but with W = R1,1×Σ×T2 where Σ is a 2-cycle in K3. We choose a vielbein frame such that, at least locally, e ν 0 eˆ N = µ . (3) M (cid:18)e ν e B(cid:19) A A 3 Therefore, in the static gauge that we are considering, the induced metric on the M5-brane is simply g = gˆ (XA = 0). We may further choose µν µν ωˆ νB(XA = 0) = 0 and ωˆ νλ(XA = 0) = ω νλ, where ω νλ is the spin connec- µ µ µ µ tion that one would calculate from the vielbein e ν. Finally we also see that µ Γˆ = eˆ νΓ = Γ is the same γ-matrix that one would calculate simply using µ µ ν µ the worldvolume metric g . µν This allows us reinterpret the bulk Killing spinor condition on the world- volume as 0 = Dˆ ǫ µ 1 1 = ∂ ǫ+ ωˆ νλΓ + ωˆ ABΓ (4) µ 4 µ νλ 4 µ AB = D ǫ+A ǫ, µ µ where ǫ = ǫˆ(XA = 0), ω AB = ωˆ AB(XA = 0) and A = 1ω ABΓ . µ µ µ 4 µ AB We find that, at lowest order in the fields XA, B and ψ, the following µν symmetries close on-shell into translations, gauge transformations and local tangent frame rotations δXA = iǫ¯ΓAψ δB = iǫ¯Γ ψ (5) µν µν 1 δψ = ∇ XAΓµΓ ǫ+ ΓµνλH ǫ, µ A µνλ 2·3! where ∇ XA = ∂ XA +ω AXB. The Fermion equation of motion that is µ µ µB required to close the algebra is Γµ∇ ψ = 0, (6) µ where ∇ ψ = D ψ +A ψ. µ µ µ What are the remaining equations of motion? The B-field has a self-dual field strength H = dB and hence one finds d ⋆ H = 0. This condition is preserved by the supersymmetries (5). Taking a supersymmetry variation of the Fermion equation of motion (6) leads the condition 1 0 = Γ ∇2XAǫ+ Γ ΓµνF AXBǫ, (7) A 2 A µνB where F A = ∂ ω A −∂ ω A +ω Cω A −ω Cω A µνB µ νB ν µB µB µC µB µC = Rˆ A(XA = 0). (8) µνB 4 To proceed we assume there is a relation of the form 1 Γ ΓµνF Aǫ = MA Γ ǫ, (9) 2 A µνB B A in which case the equation of motion for XA, along with the other fields, is ∇2XA +MA XB = 0 B iΓµ∇ ψ = 0 (10) µ 1 H = ǫ Hρστ. µνλ µνλρστ 3! We need to confirm that these equations are supersymmetric. To this end we note that the Fermion equation of motion (6) implies 1 1 ∇2ψ − Rψ + ΓµνF CDΓ ψ = 0. (11) 4 8 µν CD To connect with (7) we multiply this on the left by ǫ¯ΓA to find 1 1 ǫ¯ΓA∇2ψ − Rǫ¯ΓAψ + ¯ǫΓAΓµνF CDΓ ψ = 0. (12) 4 8 µν CD Next we note that since ωˆ νA = 0 we have that Rˆ λA = 0 and therefore the µ µν Killing spinor integrability condition [Dˆ ,Dˆ ]ǫ = 1Rˆ MNΓ ǫ = 0 implies µ ν 4 µν MN R λρΓ ǫ = −F CDΓ ǫ, (13) µν λρ µν CD and hence 1 ǫ¯R = ¯ǫΓ ΓµνF CD. (14) 2 CD µν Using this we see that (12) implies 1 ǫ¯ΓA∇2ψ + ¯ǫΓBΓµνF AXBψ = 0. (15) 2 µνB The XA equation can be compared to (15) by noting that δ∇2XA = iǫ¯ΓA∇2ψ, (16) and one sees that the equations (10) are preserved by supersymmetry. Let us consider for example the case where W is non-trivially embedded in eight dimensions, so that only F 67 6= 0. We see from (9) and (14) that µν the only non-vanishing components of MA are B M6 = M7 = R. (17) 6 7 Thus we find the scalar equations are ∇2X6 +RX6 = 0 , ∇2X7 +RX7 = 0 , (18) ∇2XA = 0 , A = 8,9,10. 5 3 Counting Moduli In the previous section we determined the lowest order equation of motion for an M5-brane wrapped on a general calibrated submanifold W of M. As a result we saw that the Fermions and scalar fields couple minimally to the gauge field associated to the structure group of the normal bundle and some scalars develop a mass term from the curvature. However the three-form remains closed and self-dual (at the linearized level). In this section we wish to perform a precise counting of the massless degrees of freedom for an M5- brane wrapped on a four-cycle P ⊂ M, i.e. W = R1,1 ×P, in a spacetime of the form M = R1,4×K where K is some compact Calabi-Yau space that contains P. This has been discussed in great detail in [2, 9] and we will largely follow their discussion. Thesimplest fieldtoconsider isthedimensionalreductionofthetwo-form gauge field. As a consequence of the self-duality condition one finds b+(P) 2 − right moving scalars and b (P) left moving scalars. For the compact K¨ahler 2 manifoldsthatweconsiderhereb+(P) = 2h (P)+1andb−(P) = h (P)−1. 2 2,0 2 1,1 If h (P) is non-vanishing then there will be 2h (P) Abelian gauge fields in 1,0 1,0 the two-dimensional effective theory. However these are non-dynamical we will not need them here. Next we consider reduction of the scalars XA. In total there are five. Three of these, X8,X9,X10 simply parameterize the location in the non- compact transverse space. These always give3 leftand3right moving scalars in two dimensions. The remaining two scalars are in fact sections of the normal bundle of P inside K. As such the number of such zero modes is hard to calculate. Let us denote the number of normal bundle moduli by N(P,K). These are left-right symmetric and we will discuss them in more detail shortly. As for the Fermions it is well known (see [11]) that spinors on a K¨ahler manifold P can be realized as (0,p)-forms on P. To see this one first consider complex coordinates for P so that {Γa,Γb} = {Γa¯,Γ¯b} = 0 and {Γa,Γ¯b} = 2ga¯b with a,b = z,w. In particular we consider a spinor ground state |0i which is annihilated by the holomorphic γ-matrices; Γa|0i = 0. We can then construct a general spinor by 1 |ψ >= ω|0i+Γa¯ω |0i+ Γa¯¯bω |0i. (19) a¯ 2 a¯¯b By construction ω is totally anti-symmetric and hence represents a a¯1...a¯p (0,p)-form on P. Furthermore if we choose the complex γ-matrices Γz = Γ2 + iΓ4 and Γw = Γ3 + iΓ5 then one sees that Γ |0i = |0i and more 2345 6 generally Γ |ω i = (−1)p|ω i, (20) 2345 p p where |ω i = 1ω Γa¯1..a¯p|0i. Since the Fermions on the M5-brane satisfy p p! a¯1...a¯p Γ ψ = −ψ we see that |ω i leads to right and left moving Fermions in 012345 p two dimensions if p is even or odd respectively. To find massless two-dimensional modes we assume that |0i is Killing with respect to ∇ defined above. In this case one see that solutions to (Γa∇ + Γa¯∇ )ψ = 0 correspond to ∂¯ ω = 0 and gba¯1∂ ω = 0, a a¯ [¯b a¯1...a¯p] [b a¯1...a¯p] i.e. ω ∈ H(0,p)(P). Thus one finds that number of massless left and right p moving two-dimensional Fermions is NL = 4h (P), NR = 4(h (P)+h (P)). (21) F 1,0 F 0,0 2,0 Here the factor of 4 comes from the fact the spinor ‘groundstate’ |0i can be thought of as having 32 real components but is subject to the three con- straints: Γz|0i = Γw|0i = 0 and Γ |0i = −|0i. Thus |0i has four real 012345 independent components. Let us summarize our counting so far. We find NL = 2+h (P)+N(P,K) B 1,1 NR = 4+2h (P)+N(P,K) B 2,0 NL = 4h (P) (22) F 1,0 NR = 4h (P)+4, F 2,0 where we have assumed that h (P) = 1. Since the wrapped M5-brane 0,0 preserves (at least) (0,4) supersymmetry the right-movers must have Bose- Fermi degeneracy. This immediately allows us to determine the number of normal moduli to be N(P,K) = 2h (P), (23) 2,0 and hence the massless spectrum is NL = 2h (P)+h (P)+2 B 2,0 1,1 NR = 4h (P)+4 B 2,0 NL = 4h (P) (24) F 1,0 NR = 4h (P)+4. F 2,0 Note that this also ensures that the number of right-moving modes is a multiple of 4, as also required by (0,4) supersymmetry. We would like to emphasis that this formula should apply whenever it make sense to talk of a 7 classical M-brane that is wrapped on a smooth complex submanifold of any smooth Calabi-Yau (including K3×T2 and T6). This formula should be contrasted with the result N(P,K) = 2h (P)−2h (P), (25) 2,0 1,0 first obtained in [2] for an ample four-cycle P in a generic Calabi-Yau and extended to K3 × T3 and T6 in [9]. We see that there is agreement for a generic Calabi-Yau where h (P) = 0. However, as pointed out in [9], the 1,0 formula (25) contradicts supersymmetry when h (P) 6= 0. In the rest of 1,0 this section we will argue that (23) is the correct counting and identify the missing modes that are absent from (25). We start with a brief review of the calculation in [2]. This starts from the observation that a 4-cycle P ⊂ K is defined by the zeros of a section of a line bundle over K. The Poincar´e dual two-form to P, which we denote by [P], determines the Chern class of the line bundle. Thus counting the number of deformations of P corresponds to counting the (real) dimension of the dimension of the space of line bundles. However one must take into account the fact that if P is described by zeros of a section s then the zeros of λs describe the same P for any λ ∈ C⋆. Thus one needs the real dimension of the projective space of line bundles. In this way one determines N(P,K) through N(P,K) = 2dim(H0(K,L))−2 = 2 (−1)idim(Hi(P,L))−2 Xi = 2 e[P]Td(K)−2 (26) ZK 1 1 = [P]3 + [P]∧c (K)−2. 2 3 Z 6 Z K K Here thesecond linefollowsfromtheKodairavanishing theorem; Hi(K,L) = ∅ for i > 0, and the third line from a Riemann-Roch index formula. For an account of these theorems see [12, 13]. Next one can use the formula (see [2, 9]) 1 1 h (P) = [P]3 + [P]∧c (K)+h (P)−1, (27) 2,0 2 1,0 6 ZK 12 ZK to obtain (25). So what is missing from this calculation (26)? A central assumption of [2] is that P is an ample cycle. Technically this means that the Poincar´e dual two-form [P] lies inside the K¨ahler cone, i.e. it defines a positive volume for 8 all complex 2-,4- and 6-cycles in K. More intuitively an ample cycle P of a manifold K is one that is sufficiently generic so that the set of all normal vectors to P spans the entire tangent space of K. AkeyassumptionoftheKodairavanishingtheoremisthatthelinebundle L is positive and hence (26) counts the dimension of the space of positive line bundles. While every ample four-cycle in K defines a positive line bundle there are zero modes which do not correspond to positive line bundles. In particular consider translations of P along any of the S1 factors in K. These S1 factors are trivial and describing the location of an M5-brane in S1 simply corresponds to specifying a value of the coordinate for that S1. As such the location is simply a section of a trivial U(1) line bundle over P and this extends to a trivial U(1) bundle over K. These deformations are not counted in (25) since the associated line bundle is trivial. There are 2h (K) such 1,0 translations and, using the Lefschetz hyperplane theorem (valid for ample four-cycles), we have that h (K) = h (P). Therefore we find an extra 1,0 1,0 2h (P) normal modes that arise from translations along the S1 factors of 1,0 K. Including these modes in (26) gives (23). An alternative description of these translational modes is to note that the S1 factors are orbits of a U(1) Killing isometry that acts on K. An ample cycle breaks the symmetries corresponding to translations along the S1 factors and hence there must be 2h (K) = 2h (P) Goldstone modes. 1,0 1,0 There are also smooth but non-ample four-cycles for which h (P) 6= h (K) 1,0 1,0 and the index theorem does not apply. In these cases one also finds that the cycle breaksfewer U(1)isometries andasaresult hasfewer Goldstonemodes. We will explicitly see in the examples below that nevertheless (23) is valid for all smooth four-cycles, as required by supersymmetry. 3.1 Three Examples To illustrate this discussion let us consider some explicit examples for K = K3 ×T2. We will consider three choices for P: P = K3, P = Σ× T2 and P = K3 + Σ × T2, where Σ is a two-cycle in K3. For a useful account of various facts about K3 see [14]. Following this we will also discuss the case where K = T6. First we consider the case where P = K3 which was first studied in detail in [8]. The Hodge diamond of K3 is 1 0 0 K3 : 1 20 1. (28) 0 0 1 9 In this case it is clear that N(K3,K3 ×T2) = 2 since K is simply a direct product K = K3 ×T2. Hence the normal bundle to P = K3 is trivial and there is no obstruction to moving the K3 around inside K. Since the Killing spinor on K3 is chiral, reduction on K3 × T2 leads to a two-dimensional theory with (0,8) supersymmetry. Looking at the field content we find the massless modes NL = 24, NR = 8 , NL = 0, NR = 8, (29) B B F F which is the same as the worldsheet action for the Heterotic string on T3. Next we consider the case where P = Σ×T2. Let us suppose that Σ is ample in K3. In complex dimension two the Lefschetz hyperplane theorem does not imply that h (Σ) = h (K3) = 0 and hence h (Σ) = g need not 1,0 1,0 1,0 be zero. Assuming Σ is connected the Hodge diamond of Σ×T2 is 1 1+g 1+g Σ×T2 : g 2+2g g. (30) 1+g 1+g 1 To determine N(Σ × T2,K3 × T2) we note that N(Σ × T2,K3 × T2) = N(Σ,K3). Since K3 does not have any S1 factors we may use a similar calculation as in (26), suitably adapted to 2 complex dimensions. We find that 2 dim(H0(K3,L)) = (−1)idim(Hi(K3,L)) Xi=0 = e[Σ]Td(K3) (31) Z K3 1 1 = c (K3)+ [Σ]2 2 Z 12 2 K3 1 = 2+ [Σ]2. 2 Z K3 Thus proceeding as before and taking account of the projective equivalence we find N(Σ,K3) = 2(dim(H0(K3,L))−1) = 2+ [Σ]2. (32) Z K3 Continuing we observe that [Σ] is the Poincare dual to Σ and hence we find [Σ]2 = [Σ] Z Z K3 Σ 10