hep-th/0505264 LPTENS-05/14 ZMP-HH/05-07 Imperial/TP/050401 Hitchin Functionals in N = 2 Supergravity Mariana Gran˜aa,b, Jan Louisc,d and Daniel Waldrame,f 6 0 aLaboratoire de Physique Th´eorique de l’Ecole Normale Sup´erieure 0 24 rue Lhomond, 75231 Paris Cedex, France 2 [email protected] n a J bCentre de Physique Th´eorique, Ecole Polytechnique 8 91128 Palaiseau Cedex, France 1 4 cII. Institut fu¨r Theoretische Physik der Universit¨at Hamburg v 4 Luruper Chaussee 149, D-22761 Hamburg, Germany 6 [email protected] 2 5 0 dZentrum fu¨r Mathematische Physik, Universit¨at Hamburg, 5 Bundesstrasse 55, D-20146 Hamburg 0 / h t eBlackett Laboratory, Imperial College - p London, SW7 2AZ, U.K. e [email protected] h : v i fThe Institute for Mathematical Sciences, Imperial College X London, SW7 2AZ, U.K. r a ABSTRACT We consider type II string theory in space-time backgrounds which admit eight su- percharges. Such backgrounds are characterized by the existence of a (generically non- integrable) generalized SU(3) SU(3) structure. Wedemonstrate how the corresponding × ten-dimensional supergravity theories can in part be rewritten using generalised O(6,6)- covariant fields, in a form that strongly resembles that of four-dimensional N = 2 super- gravity, and precisely coincides with such after an appropriate Kaluza–Klein reduction. Specifically we demonstrate that the NS sector admits a special K¨ahler geometry with K¨ahler potentials given by the Hitchin functionals. Furthermore we explicitly compute the N = 2 version of the superpotential from the transformation law of the gravitinos, and find its N = 1 counterpart. May 2005 Contents 1 Introduction 2 2 Type II supergravity with SU(3) SU(3) structure 6 × 2.1 N = 2 and SU(3) SU(3) structures . . . . . . . . . . . . . . . . . . . . 6 × 2.1.1 Effective theories and G-structures . . . . . . . . . . . . . . . . . 6 2.1.2 A d = 10 reformulation and generalized structures . . . . . . . . . 8 2.2 Field decompositions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 2.3 Scalar action: kinetic terms . . . . . . . . . . . . . . . . . . . . . . . . . 11 2.3.1 Geometry of SU(3) structures . . . . . . . . . . . . . . . . . . . . 12 2.3.2 Kinetic terms for NS deformations . . . . . . . . . . . . . . . . . 13 2.4 Special K¨ahler manifolds and stable forms . . . . . . . . . . . . . . . . . 16 2.4.1 Review of special K¨ahler geometry . . . . . . . . . . . . . . . . . 16 2.4.2 O(6,6) spinors and SU(3) SU(3) structures . . . . . . . . . . . 17 × 2.4.3 Special K¨ahler structure for stable Spin(6,6) spinors . . . . . . . 20 2.4.4 Special K¨ahler structure for stable J . . . . . . . . . . . . . . . . 25 2.4.5 Special K¨ahler structure for stable ρ . . . . . . . . . . . . . . . . 26 2.4.6 Special K¨ahler structure for metric SU(3) SU(3)-structures . . 28 × 2.5 Supersymmetry transformation of the gravitinos . . . . . . . . . . . . . . 31 2.6 N = 1 superpotentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 3 The N = 2 effective theory in four dimensions 40 3.1 Defining the truncation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 3.2 Reducing the Neveu-Schwarz sector . . . . . . . . . . . . . . . . . . . . . 44 3.3 The reduction of the type IIA RR-sector . . . . . . . . . . . . . . . . . . 46 3.4 The reduction of the type IIB RR-sector . . . . . . . . . . . . . . . . . . 49 3.5 Relation to gauged supergravity . . . . . . . . . . . . . . . . . . . . . . . 51 3.6 Mirror symmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 4 Conclusions 54 1 1 Introduction The interplay between supersymmetry and geometry has been very fruitful in the past. For example, compactifications of ten-dimensional type II supergravities on Calabi–Yau threefolds Y preserve eight supercharges and yield four-dimensional N = 2 (ungauged) supergravities as effective low energy field theories [1, 2, 3, 4]. The spectrum and cou- plings of these N = 2 supergravities are in turn determined by geometrical (and topo- logical) properties of the Calabi–Yau manifolds. Supersymmetry strongly constrains the couplings and thus also constrains the Calabi–Yau geometry. For example, it implies that the moduli space of metric deformations of a Calabi–Yau manifold is the product of two special K¨ahler manifold characterized by two holomorphic prepotentials [1, 5, 6, 7]. The Calabi–Yau moduli space indeed satisfies this property and furthermore one can use geometrical methods together with mirror symmetry to compute both prepotentials exactly [8, 9]. Expanding on earlier work in refs. [10, 11, 12], there has recently been much interest in a generalized class of backgrounds where the Calabi–Yau manifold is replaced by a manifold Y which is no longer Ricci-flat [13]–[44]. One way such generalized compactifi- cations arise is when localized sources (D-branes, orientifold planes) and/or background fluxes are present and the solution of the equations of motion forces the geometry to back-react to the additional background energy density. A certain class of manifolds, called ‘half-flat manifolds’ [45], also appeared as mirror symmetric backgrounds of type II Calabi–Yau compactifications with background fluxes [15, 18, 23]. Within this generalized set-up one is particularly interested in backgrounds which continue to preserve some of the supercharges or more generally where a number of supercurrents exist but the associated supercharges are spontaneously broken. The latter case includes examples which do not satisfy the equations of motion, such as the classical example of a Calabi–Yau manifold with generic background fluxes. In either case, the existence of the supercurrents requires that a set of spinors are globally well defined on the manifold Y which in turn implies that the structure group has to be reduced. In the mathematical literature manifolds with a reduced structure group G are called manifolds with G-structure [46, 47]. Generically G does not coincide with the holonomy group since the spinors are not necessarily covariantly constant with respect to the Levi-Civita connection. The degree to which they fail to be covariantly constant is measured by a quantity known as the intrinsic torsion and can be used to classify the G-structure. From a particle physics point of view preserving the minimal amount of supersym- metry is the most interesting case. On a six-manifold the existence of a single globally defined spinor η requires the reduction of the structure group from Spin(6) to SU(3) and therefore manifolds with SU(3) structure play a special role. They can be characterized by the invariant spinor on the manifold or, more conveniently, by a real two-form J and a complex three-form Ω. Since η is not covariantly constant neither J nor Ω are closed. Instead dJ and dΩ decompose into SU(3) representations. These define the intrinsic torsion and can be used to classify the different SU(3) structures [45]. For Calabi–Yau threefolds J andΩ areclosed, η iscovariantly constant andthe holonomy groupisSU(3). Here, we will focus on type II supergravities which have N = 2 supersymmetry in ten space-time dimensions. Decomposing the spinor representation in ten dimensions under 2 Spin(1,3) Spin(6) and requiring N = 2 supersymmetry in four dimensions implies that × there are two non-vanishing spinors on Y, one for each of the original ten-dimensional spinors. Each defines an SU(3) structure. Locally, the two SU(3) structures define an SU(2) structure, which survives globally as an SU(2) structure if the spinors never become parallel. If the spinors are always parallel we just have a single SU(3) structure. One way to characterize this structure mathematically is in terms of “generalised geometry”, first introduced by Hitchin [48]. One considers the sum of the tangent and cotangent bundle of Y, TY T Y on which there is a natural O(6,6) structure. The ∗ ⊕ two six-dimensional spinors transform under a Spin(6) Spin(6) subgroup defined by × the metric and NS B-field, and being globally defined, imply that the structure group of TY T Y actually reduces to SU(3) SU(3) [34] (see [49] for the original, related ∗ ⊕ × discussion of U(n) U(n) structures). In this formulation, the SU(3) SU(3) structure × × can be defined by a sum of odd forms Φ and a sum of even forms Φ+, each built out − of spinor bilinears [28, 34] (see also [50] for the construction in the case of G G 2 2 × structures). From the point of view of the TY T Y bundle these forms correspond to ∗ ⊕ a pair of Spin(6,6) spinors [51, 52, 48]. Since we are interested in backgrounds with supercurrents but, in general, sponta- neously broken supersymmetry, we do not require the SU(3) structures to be integrable. Enforcing preserved supersymmetry (and the equations of motion) would impose inte- grability constrains [28, 34, 44]. The geometric structures used throughout this paper are therefore “almost” (or not necessarily integrable) structures. However, in order to avoid cumbersome wording, we will typically drop the “almost” when referring to them. The description of backgrounds in terms of SU(3) structures and the generalization to SU(3) SU(3) structures has also recently played an important role in topological × string theories. In particular, it has been argued [53, 54, 55] that the target space theory of the A and B model topological strings can be defined in terms of a functional of the structures J or Ω first considered by Hitchin [51, 52, 48]. More generally [56] one must consider the corresponding functional for the Spin(6,6) spinors Φ . Similarly ± it has been possible [29, 32] to generalize the notion of topological strings away from backgroundswithSU(3)-structure(suchasCalabi–Yaumanifolds)tomoregeneralspaces with SU(3) SU(3) structure, again using the spinors Φ . ± × Returning to the physical string, for Calabi–Yau compactifications the N = 2 low en- ergy effective action in four space-time dimensions can be derived by a standard Kaluza– Kleinreductionwhereonlythemassless modescorresponding toharmonicformsonY are kept [1, 2, 3, 4]. This procedure is valid whenever Y is large and the supergravity approx- imation can be used reliably. In the presence of background fluxes the same method has been applied for example in refs. [57]-[68]. One chooses the fluxes to be small, the com- pactification manifold to be large and hence consistently neglects the back-reaction of the geometry. One finds that the kinetic terms are unaltered and the flux parameters appear as gauge couplings and/or mass parameters which turn the supergravity into a gauged or massive supergravity. However, when dealing with manifolds with non-integrable SU(3) structure, this procedure is a bit more tricky since generically it is harder to specify in what sense one is making a small deformation. For instance, turning on H-flux on a Calabi–Yau manifold can map to a change in topology of the mirror manifold. Thus one cannot treat the intrinsic torsion easily as a simple deformation of the supergravity as was done for the fluxes. 3 The goal of this paper is to study type II supergravity in generic backgrounds with SU(3) (the case where the two spinors are always parallel) or, more generally, SU(3) × SU(3) structure. Our motivation is to define a ‘rule’ for deriving the low-energy four- dimensional effective theory and to uncover the role of the torsion in supergravity. As in Calabi–Yau compactifications this might lead to interesting insights into the interplay of geometry and supersymmetry of the effective theory. However, we begin with a more general set-up. We do not immediately confine our interest to the low energy effective action or performing a Kaluza–Klein reduction. This leads us to a reformulation of the fullten-dimensional theory, abandoningmanifest ten-dimensionalLorentzinvariance, but with bosonic fields transforming in Spin(1,3) O(6,6) multiplets. This is similar to and × inspired by the approach pioneered in ref. [69], which considered a related reformulation of eleven-dimensional supergravity. Although we provide no direct evidence, we expect the reformulation has a local Spin(1,3) SU(3) SU(3) symmetry. More specifically, × × we first demand that the tangent space of the ten-dimensional background is a direct sum T1,3 F where T1,3 is a Spin(1,3) bundle while F is a Spin(6) bundle. Then we ⊕ further require that structure group of F reduces admitting an SU(3) structure. In fact we also consider the more general situation where the sum of the tangent plus the cotangent bundle admits SU(3) SU(3) structure. In both cases eight of the original 32 × supercharges are singled out and we can rewrite the ten-dimensional supergravity with 32 supercharges in a form as if it had only eight supercharges. In this framework, the supermultiplet structure and action follow the form of four- dimensional N = 2 supergravity although the theory remains fully ten-dimensional. In particular, concentrating on the bosonic fields which are scalars under Spin(1,3), we defineaspaceof(notnecessarilyintegrable)SU(3) SU(3)structuresandshow, following × refs. [51, 52, 48], that it admits a special K¨ahler geometry with a K¨ahler potential given by a Hitchin functional. Restricting to the particular case of a single SU(3) structure, we furthermore rewrite the supersymmetry transformation law of the eight gravitinos in a form analogous to the transformation law of the four-dimensional N = 2 gravitinos. This allows us to read off the three ‘Killing prepotentials’ or momentum maps x,x = 1,2,3 P which are the N = 2 equivalent of the superpotential and the D-term. In this ten- dimensional theory they turn out to be determined by the background fluxes and the intrinsic torsion. In the same spirit we can continue the decomposition keeping only four supercharges. In this way we find the most general N = 1 superpotential induced by the fluxes and torsion. We find that this generalized superpotential contains all previously known cases inappropriatelimitswhen either torsion, NSor RRfluxes areset to zero. Forexample, in the torsionless case we recover the Gukov-Taylor-Vafa-Witten superpotential [70, 71, 59]. After having rewritten the ten-dimensional theory in an ‘N = 2 form’ it is straightfor- ward to perform a KK-reduction. We choose the background to be a product M1,3 Y × where M1,3 is a four-dimensional manifold with Minkowskian signature while Y is a compact manifold with SU(3) structure. (The more general case of compactifications with SU(3) SU(3) structure will be discussed elsewhere.) In the KK-reduction one × conventionally keeps the light modes and integrates out the heavy ones. However, back- grounds with a generic Y do not necessarily have a flat Minkowskian ground state and the distinction between heavy and light is not straightforward. Therefore we do not specify the precise form of the truncation, which would depend on the particular choice 4 of background, but instead leave it generic, extracting the set of conditions that such a reduction must satisfy to be self-consistent. The truncation is defined by extracting from the infinite tower of KK-modesonly a finite subset. However, we impose one further condition in that we only keep the two gravitini in the gravitational multiplet but project out all gravitini which reside in their own (massive) spin-3 multiplets. This ensures that 2 the resulting low effective action contains apart form the gravitational multiplet only N = 2 vector, tensor and hypermultiplets. Once the ten-dimensional spectrum is truncated the gauge invariance of the original ten-dimensional theory is no longer automatically maintained. Instead, as we will see, gauge invariance imposes additional constraints on the truncation which also have been observed in [72]. Imposing these constraints, the N = 2 action takes a standard form [73,74]–possibly withmassive tensor multiplets[64,75,76]. Thisenablesustodiscuss in detail the supergravity/geometry correspondence. We find, as expected, that the torsion (aswell asthe fluxes) deformthe N = 2 supergravity andturnit into agaugedor massive supergravity. The gauge charges and mass parameters are directly related to the fluxes and torsion and we derive the precise relationship by computing the supersymmetry transformations of the gravitino. This paper is organized into two main sections. In section 2 we discuss the reformula- tion of the ten-dimensional type II supergravity theory in terms of N = 2–like structures while in section 3 we perform the Kaluza–Klein reduction and compute some of the cou- plings in the low energy effective theory. More specifically in 2.1 we first show that eight linearly realized supercharges require that the theory has a Spin(1,3) SU(3) SU(3) × × structure. In section 2.2 we then show how the ten-dimensional fields decompose into N = 2 multiplets for the case of a single SU(3) structure. After reviewing a few facts about SU(3) structures, we give the part of the action for the deformations of the NS fields in section 2.3. In section 2.4 we show that their kinetic terms form a product of two special K¨ahler geometries in exact analogy with the moduli space of Calabi–Yau compactifications. Furthermore the K¨ahler potential is determined by the sum of two Hitchin functionals both of which can be derived from a universal expression given in terms of a pure Spin(6,6) spinor [51, 48]. In section 2.5 we compute the scalar part of the supersymmetry transformations of the gravitinos and determine the ten-dimensional analog of the Killing prepotential x. By an appropriate further reduction we compute P the N = 1 superpotential in section 2.6. In section 3 we perform the KK-truncation. We first definethe‘rules’forthereductionin3.1. Weproject outall3 ¯3representations and ⊕ then impose local p-form gauge invariance on the remaining spectrum. In 3.2 we discuss the reduction of the common NS-sector and show that the resulting K¨ahler potentials precisely coincide with the analogous K¨ahler potentials of Calabi–Yau manifolds. In 3.3 and 3.4 we perform the reduction of the RR-sector for type IIA and type IIB. In 3.5 we check that the ‘proper’ N = 2 Killing prepotential x obtained by truncation from P its higher-dimensional ‘father’ agrees with the generic form dictated by N = 2 gauged supergravity. Finally in 3.6 we briefly discuss mirror symmetry for these generalized compactifications and 4 contains our conclusions. 5 2 Type II supergravity with SU(3) SU(3) structure × The goal in this section is to understand some of the details of how we can reformulate the ten-dimensional type II supergravity theory in terms structures analogous to N = 2 four-dimensional supergravity. In doing so we lose manifest Spin(1,9) Lorentz symmetry, and instead arrange the fields in Spin(1,3) O(6,6) multiplets. We will concentrate on × thescalar fieldpartofthetheory, thatismultiplets which containfields which aresinglets under Spin(1,3). In an N = 2 language these are the vector, tensor and hypermultiplets. In particular, we will show that there are special K¨ahler geometries on the spaces of scalar fields describing their kinetic terms. Furthermore, we will show how the ten- dimensional theory gauges a set of isometries on these spaces, described by a set of Killing prepotentials again just as in four-dimensional N = 2 supergravity. We find that all these objects can be written in a simple way in terms of generalised geometrical structures, invariant under O(6,6) transformations. In particular, the K¨ahler potential of the special K¨ahler geometry is given by the Hitchin functional. Let us start by discussing the relation between rewriting the theory in terms of eight linearly realized supercharges (N = 2) and the existence of generalised SU(3) SU(3) × structures. 2.1 N = 2 and SU(3) SU(3) structures × 2.1.1 Effective theories and G-structures One motivation for this paper is to consider the general low-energy gauged supergravity theory that arises when type II string theory (or rather type II supergravity) is compact- ified on the space-time background M1,9 = M1,3 Y . (2.1) × Here M1,3 is the four-dimensional, physical space-time while Y is a six-dimensional com- pact manifold.1 The product structure of the space-time background (2.1) implies a de- composition of the Lorentz group Spin(1,9) Spin(1,3) Spin(6) and an associated de- compositionofthespinorrepresentation16 ⊃Spin(1,9)a×ccordingto16 (2,4) (¯2,¯4). ∈ → ⊕ We are interested in the situation where the effective theory on M1,3 has the minimal N = 2 supersymmetry. In other words, we need to single out eight particular type II supersymmetries which descend to the effective theory. For type IIA we start with two supersymmetry parameters of opposite ten-dimensional chirality. Using a standard decomposition of the ten-dimensional gamma matrices ΓM = (Γµ,Γm) as Γµ = γµ 1 , µ = 0,1,2,3 , Γm = γ γm , m = 1,...,6 , (2.2) 5 ⊗ ⊗ where γ = iγ0γ1γ2γ3, we can write 5 ε1 = ε1 η1 +ε1 η1 , IIA + ⊗ + − ⊗ − (2.3) ε2 = ε2 η2 +ε2 η2 , IIA + ⊗ ⊗ + − − 1In this paper we do not consider the possibility of a warped background but leave the study of this class of compactification to a separate publication. 6 whereε1,2 = (ε1,2)c andη1,2 = (η1,2)c. (Hereasusualηc = Dη ,whereD istheintertwiner + + ∗ giving −γm = D 1γmD−. We also have η¯ = η A, where γm = AγmA 1. By a slight ∗ − † † − − abuse of notation we use plus and minus to indicate both four-dimensional and six- dimensional chiralities.) For type IIB both spinors have negative chirality resulting in the decomposition εA = εA ηA +εA ηA , A = 1,2 . (2.4) IIB + ⊗ ⊗ + − − In each case we have a pair of spinors εA in M1,3 parameterizing the N = 2 supersym- + metries. In addition, we have two spinors ηA of Y fixing precisely which of the ten- + dimensional supersymmetries descend to four dimensions. Note that generically these can be different for the two ten-dimensional supersymmetry parameters εA. For such a reduction to work, the spinors ηA must be globally defined and nowhere + vanishing on Y and hence the structure group of the tangent space of Y has to reduce. Consider one such global spinor. It has to transform as a singlet under the structure group. Decomposing under SU(3) Spin(6), the complex spinor representation splits ⊂ as 4 = 3 1. Thus if the structure group is contained in SU(3) we indeed get a spinor ⊕ singlet. Manifolds with this property are known as ‘manifolds with SU(3) structure’ in themathematical literature[45]. Since wedo not require thebackground tobesupersym- metric, only that the four-dimensional effective action has a set of N = 2 supercurrents, there are generically no differential conditions on the spinors ηA. In the mathematical + literature this means we have an “almost” or not necessarily integrable SU(3) structure. From eqs. (2.3) and (2.4) we see that in general we have a pair ηA of such spinors, + each of which defines an SU(3) structure. In summary d = 4, N = 2 effective theory Y admits a pair of SU(3) structures . (2.5) ⇔ Locallythetwospinorsη1 andη2 spanatwo-dimensionalsubspaceofthefour-dimensional + + space of positive chirality Spin(6) = SU(4) spinors. This space is invariant under ∼ SU(2) SU(4) rotations, under which both spinors are singlets. Thus locally the pres- ⊂ ence of two SU(3) structures actually implies that we have an SU(2) structure. However, globally there can be points where the spinors become parallel, and hence at these points no SU(2) structure is defined. The extreme case where the two spinors are parallel every- where is allowed, and in this case the two SU(3) structures coincide, leading to a single SU(3) structure. As we discuss in more detail below, a special case of a supersymmetric compactifi- cation is where Y is a Calabi–Yau manifold and η1 = η2. In this case, in deriving the + + low-energy effective theory, one keeps only the massless modes and disregards all heavy Kaluza–Klein modes (together with all heavy string states). However, for compactifica- tions on generic manifolds with a pair of SU(3) structures the distinction between heavy and light modes is not straightforward. It appears that we have to define a ‘rule’ for the reduction to decide which modes we keep in the four-dimensional effective action and which modes we truncate away. In fact, as we now discuss, we can actually start by doing something more general, where we do not truncate the theory at all. 7 2.1.2 A d = 10 reformulation and generalized structures The previous discussion was based on the assumption that we had a product mani- fold (2.1). However it is not really necessary to make such an assumption. In general, if we break the local Spin(1,9)-invariance one can always rewrite the full d = 10 type II supergravity theory as though it were a theory with only eight supercharges. The struc- ture of the theory is then analogous to N = 2 in four dimensions, but no Kaluza–Klein expansion is made and instead we work in ten space-time dimensions keeping all the de- grees of freedom. A similar reorganization of eleven-dimensional supergravity was done in ref. [69] in order to understand the origin of hidden symmetries in lower dimensional supergravities. More precisely, we require only that the ten-dimensional manifold M1,9 admits a pair of SU(3) structures. This means that the tangent space TM1,9 splits as TM1,9 = T1,3 F , (2.6) ⊕ where T1,3 is a real SO(1,3) vector bundle and F is a SO(6) vector bundle which admits a pair of SU(3) structures. In other words we have two different decompositions of the complex vector bundle FC = F C, that is ⊗ FC = E1 E¯1 = E2 E¯2 , (2.7) ⊕ ⊕ where each EA is a complex SU(3) vector bundle corresponding to the SU(3) structure defined by ηA. Equivalently, recall that the original type II theory is formulated on + a supermanifold M1,916+16 of bosonic dimension (1,9), with a manifest local SO(1,9) | invariance and with the Grassmann variables transforming as a pair of 16-dimensional spinor representations. The requirement that we have a pair of SU(3) structures means there is a sub-supermanifold N1,94+4 M1,916+16 (2.8) | | ⊂ still with bosonic dimension (1,9), but now with only eight Grassmann variables trans- forming as spinors of Spin(1,3) and singlets of one or the other of the SU(3) groups. It is natural to reformulate the d = 10 supergravity in this sub-superspace. As such, although the theory is still defined in ten-dimensions (though without manifest SO(1,9) invariance), it will have structures analogous to those of d = 4, N = 2 supergravity, such as special K¨ahler moduli spaces and Killing prepotentials. Let us now turn to a second key point. The pair of SU(3) structures are actually most naturally described as a single “generalized structure”, a notion first introduced by Hitchin [48]. One starts by considering the sum of the tangent and cotangent bundles TY T Y, or more generally F F . If v F and ξ F , one can see that there is a ∗ ∗ ∗ ⊕ ⊕ ∈ ∈ natural O(6,6) metric on this space, defined by (v +ξ,v+ξ) = ξ(v) ξ vm , (2.9) m ≡ which makes no reference to any additional structure (such as a metric) on F. Note that the metric is invariant under the diffeomorphism group GL(6,R) O(6,6) acting on F ⊂ and F . The choice of metric g and NS two-form B can be shown to correspond to fixing ∗ an O(6) O(6) substructure. Given this substructure, the two spinors ηA transform × + 8 separately under the two different Spin(6) groups and the pair of SU(3) structures is actually equivalent to a (not necessarily integrable) SU(3) SU(3) structure on F F , ∗ × ⊕ as discussed in ref. [34, 50]. In summary, we have argued that N = 2-like reformulation of F F∗ admits a (not necessarily (2.10) ⇔ ⊕ type II integrable) SU(3) SU(3) structure × We expect that this SU(3) SU(3) structure is actually promoted to a local sym- × metry of the reformulated theory, in analogy with [69]. Suppose, for instance we had compactified on a torus Y = T6. It is then a familiar result that the low-energy theory has a local O(6) O(6) symmetry and a global O(6,6) symmetry, concomitant with × the fact that the string theory has a T-duality symmetry [77]. For instance the scalar degrees of freedom coming from the internal metric and B-field arrange themselves into a O(6,6)/O(6) O(6) coset. The two Spin(6) groups act separately on each spinor ηA. On × + T6 any pair of constant spinors (η1,η2) parameterizes a pair of preserved supersymme- + + tries in four dimensions and hence compactification gives an N = 8 effective theory. If we isolate a single pair, this can be reformulated as an N = 2 theory. The local O(6) O(6) × symmetry should then reduce to those symmetries that leave the pair invariant, namely a local SU(3) SU(3) symmetry. Thus, generically we expect that the effective theory on × N1,94+4 will have a local Spin(1,3) SU(3) SU(3) symmetry. In what follows we will | × × however concentrate on the analog of the scalar sector of the theory and do not provide any direct evidence for this local symmetry. In order to simplify the discussion we will frequently specialize to the case where the SU(3) SU(3) structure defines a global SU(3) structure (rather than some local SU(2) × structure). In other words we assume η1 = η2. This is mostly for convenience and + + it also allows us to connect with the existing literature on compactifications on spaces with SU(3)-structure. We stress nonetheless that from a supergravity perspective the naturalframeworkforN = 2theoriesandtruncationsisactuallyagenericSU(3) SU(3) × structure. 2.2 Field decompositions Let us first look at the decomposition of the fields of type II supergravities in the sub- supermanifold N1,94+4. Let us use the same notion as the previous section even though | we no longer necessarily have a product manifold. A µ,ν,... index denotes the repre- sentation of a field as a tensor of T1,3 while a m,n,... index denotes the representation as a tensor of F. In addition, we specialize to the case of a global SU(3) structure. This means we can futher decompose the F-tensors into into SU(3) representations. The common NS sector contains the metric g , an antisymmetric tensor B and MN MN the dilaton φ. Their decomposition into SU(3) representation is displayed in table 2.1. The notation ab denotes a field in the SU(3) representation a and with four-dimensional spin b, with T denoting an antisymmetric tensor or pseudo-scalar. The representations are real except for 6 and 3 and their conjugates. The RR sector of type IIA supergravity features a vector C and a three-form C ; M MNP their decompositions are given in table 2.2. 9