DUKE-CGTP-00-01 hep-th/0001001 January 2000 Compactification, Geometry and Duality: N = 2. 0 0 Paul S. Aspinwall 0 2 n a Center for Geometry and Theoretical Physics, J 7 Box 90318, 1 Duke University, 2 Durham, NC 27708-0318 v 1 0 0 1 0 0 Abstract 0 / h We review the geometry of the moduli space of N = 2 theories in four dimensions t - from the point of view of superstringcompactification. The cases of a type IIAor type p e IIBstringcompactifiedonaCalabi–Yauthreefoldandtheheteroticstringcompactified h on K3 T2 are each considered in detail. We pay specific attention to the differences : × v betweenN = 2theoriesandN > 2theories. Themodulispacesofvectormultipletsand i X the moduli spaces of hypermultiplets are reviewed. In the case of hypermultiplets this review is limited by the poor state of our current understanding. Some peculiarities r a such as “mixed instantons” and the non-existence of a universal hypermultiplet are discussed. Contents 1 Introduction 2 2 General Structure 3 2.1 Holonomy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 2.2 U-Duality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 2.3 Eight supercharges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 2.4 Type II compactification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 2.5 Heterotic compactification . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 2.5.1 E and its bundle . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 H 2.5.2 S and its bundle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 H 2.6 Who gets corrected? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 3 The Moduli Space of Vector Multiplets 24 3.1 The special K¨ahler geometry of M . . . . . . . . . . . . . . . . . . . . . . . 24 V 3.2 M in the type IIA string . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 V 3.2.1 Before corrections and five dimensions . . . . . . . . . . . . . . . . . 27 3.2.2 Mirror Pairs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 3.2.3 The mirror map . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 3.3 M in the heterotic string . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 V 3.3.1 Supersymmetric abelian gauge theories . . . . . . . . . . . . . . . . . 34 3.3.2 Heterotic/Type IIA duality . . . . . . . . . . . . . . . . . . . . . . . 36 3.3.3 Enhanced gauge symmetry . . . . . . . . . . . . . . . . . . . . . . . . 39 3.3.4 An example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 3.3.5 Quantum corrections to N = 2 gauge theories . . . . . . . . . . . . . 44 3.3.6 Breaking T-Duality . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 4 The Moduli Space of Hypermultiplets 51 4.1 Related Dimensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 4.1.1 N = (1,0) in six dimensions . . . . . . . . . . . . . . . . . . . . . . . 51 4.1.2 N = 4 in three dimensions . . . . . . . . . . . . . . . . . . . . . . . . 54 4.2 Extremal Transitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56 4.2.1 Conifolds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56 4.2.2 Enhanced gauge symmetry . . . . . . . . . . . . . . . . . . . . . . . . 58 4.2.3 Massless Tensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 4.3 The classical limit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 4.3.1 The E E heterotic string . . . . . . . . . . . . . . . . . . . . . . . 62 8 8 × 4.3.2 The Spin(32)/Z heterotic string . . . . . . . . . . . . . . . . . . . . 66 2 4.4 Into the interior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67 4.4.1 The hyperk¨ahler limit . . . . . . . . . . . . . . . . . . . . . . . . . . 67 4.4.2 Mixed instantons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68 4.4.3 Hunting the universal hypermultiplet . . . . . . . . . . . . . . . . . . 69 1 Introduction One of the most basic properties one may study about a class of string compactifications is its moduli space of vacua. If the class is suitably chosen one may find this a challenging subject which probes deeply into our understanding of string theory. In four dimensions it is the N = 2 cases which provide the “Goldilocks” theories to study. As we will see, N = 4 supersymmetry is too constraining and determines the moduli space exactly, leaving no room for interesting corrections from quantum effects. N = 1 supersymmetry is highly unconstrained leaving the possibility that our supposed moduli acquire mass ruining the moduli space completely. N = 2 however is just right — quantum effects are not potent enough to kill the moduli but they can affect the structure of the moduli space. (It is therefore a pity that the real world does not have N = 2 supersymmetry — such a theory is necessarily non-chiral.) The subject of N = 2 compactifications is enormous and we will present here only a rather biased set of highlights. These lectures will be sometimes closely-related to a set of lectures I gave at TASI 3 years ago [1]. Having said that, the focus of these lectures differs from the former and the set of topics covered is not identical. I will however often refer to [1] for details of certain subjects. Related to the problem of finding the moduli space of a class of theories is the following problem in string duality. Consider these four possibilities for obtaining an N = 2 theory in four dimensions: 1. A type IIA string compactified on a Calabi–Yau threefold X. 2. A type IIB string compactified on a Calabi–Yau threefold Y. 3. An E E heterotic string compactified on K3 T2. 8 8 × × 4. A Spin(32)/Z heterotic string compactified on K3 T2. 2 × Can we find cases where the resulting 4 dimensional physics is identical for two or more of these possibilities and, if so, how do we match the moduli of these theories to each other? This is a story that began with [2,3] some time ago but many details are poorly-understood to this day. One might suppose that knowing the moduli space of each theory listed above is a prerequisite for solving this problem but actually it is often useful, as we will see, to consider this duality problem at the same time as the moduli space problem. Note that there are other possibilities for producing N = 2 theories in four dimensions such as the type I open string on K3 T2. We will stick with the four listed above in these lectures as they are × quite sufficient for our purposes. As will be discussed shortly this problem breaks up into two pieces. One factor of the moduli space consists of the vector multiplet moduli space and the other factor consists of the hypermultiplet moduli space. By most criteria the moduli space of vector multiplets 2 is well-understood today. This complex K¨ahler space can be modeled exactly in terms of the deformation space of a Calabi–Yau threefold. We will therefore be able to review this subject fairly extensively. In contrast the hypermultiplet moduli space remains a subject of research very much “in progress”. We will only be able to discuss in detail the classical boundaries of these moduli spaces. The interior of these spaces may offer considerable insight into string theory but we will only be able to cover some tantalizing hints of such possibilities. These lectures divide naturally into three sections. In section 2 we discuss generalities about moduli spaces of various numbers of supersymmetries in various numbers of dimen- sions. Although these lectures are intended to focus on the case of N = 2 in four dimensions, there are highly relevant observations that can bemade by considering other possibilities. Of particular note in this section is the rigid structure which emerges with more supersymmetry than the case in question. The heart of these lectures then consists of a discussion of the vector multiplet moduli space in section 3 and then the hypermultiplet moduli space in section 4. It is perhaps worth mentioning again that these lectures do not do justice to this vast subject and should be viewed as a biased account. Topic such as open strings, D-branes and M-theory have been neglected only because of the author’s groundless prejudices. $ The paragraphs starting with a “ ” are technical and can be skipped if the reader does not wish to be embroiled in subtleties. 2 General Structure 2.1 Holonomy We begin this section with a well-known derivation of key properties of moduli spaces based on R-symmetries and holonomy arguments. We should warn the more mathematically- inclined reader that we shall not endeavour to make completely watertight rigorous state- ments in the following. There may be a few pathological special cases which circumvent some of our (possibly implicit) assumptions. Suppose we are given a vector bundle, V, with a connection. We may define the “holon- omy”, Hol(V), of this bundle as the group generated by parallel transport around loops in the base with respect to this connection. (A choice of basepoint is unimportant.) We may also define the restricted holonomy group Hol0(V) to be generated by contractable loops. This notion can be very useful when applied to supersymmetric field theories as noted in [4]. First let us consider the moduli space of a given class of theories. We will consider the moduli space as the base space of a bundle. Note that the moduli space of theories, M, is equipped with a natural metric — that of the sigma-model. The tangent directions in the moduli space are given by the massless scalar fields with completely flat potentials. These massless fields may thus be given vacuum expectation values leading to a deformation of the 3 theory. Let us denote these moduli fields φi, i = 1,... ,dim(M). The low-energy effective action in the uncompactified space-time is then given by ddx√gG ∂µφi∂ φj +... (1) ij µ Z where G is our desired metric on M. We therefore have a natural torsion-free connection ij on the tangent bundle of M given by the Levi-Civita connection with respect to this metric. Now consider the supersymmetry generators given by spinors Q , A = 1,... ,N, where A as usual N denotes the number of supersymmetries (we suppress the spinor index). These objects are representations of Spin(1,d 1) and are − Real if d = 1,2,3 mod 8 • Complex if d = 0,4 mod 8 • Quaternionic (or symplectic Majorana) if d = 5,6,7 mod 8. • The bundle of supersymmetry generators over M will also have a natural connection related to that on the tangent bundle. The key relation in supersymmetry is the equation γµ Qα,Qβ = δ Pµ, (2) αβ{ A¯ B} A¯B where γ are the usual gamma matrices, P is translation and the bars in this equation are to be interpreted according to whether the spinors are real, complex or quaternionic.1 Because parallel transport must preserve δ in (2) we see immediately that under holonomy Q A¯B A must transform as a fundamental representation of SO(N) if d = 1,2,3 mod 8 • U(N) if d = 0,4 mod 8 • Sp(N) if d = 5,6,7 mod 8, • if the loop around which we transport is contractable. These groups are the “R-symmetries” ofthesupersymmetric fieldtheoriesandgiveHol0 ofthisbundle. Wealso notethatin4M+2 dimensions, for integer M, the supersymmetries are chiral. This means that we consider left and right supersymmetries separately as we will illustrate in some examples below. The massless scalar fields live in supermultiplets. Within each supermultiplet the set of scalar fields will form a particular (possibly trivial) representation of the R-symmetry. We refer to [5] for a detailed account of this. Occasionally the supermultiplet contains only one scalar component and this then transforms trivially under R. So long as this is not the case the holonomy of our tangent bundle is related to the R-symmetry. 1We ignore central charges which are irrelevant for this argument. 4 We may be more precise than this. As we go around a loop in M the scalars within every given supermultiplet will be mixed simultaneously by the R-symmetry. The supermultiplets themselves may also be mixed as a whole into each other by holonomy. This implies that, so long as the scalars transform nontrivially under R, the holonomy of the tangent bundle is factorized with the R-symmetry forming one factor. It is important to note however that we may not mix a scalar from one supermultiplet freely with any scalar from another super- multiplet in a way that violates this factorization. This is incompatible with the detailed supersymmetry transformation laws (as the reader might verify if they are unconvinced). Note in particular that the scalars within two different types of supermultiplets can never mix under holonomy. This is a useful observation given the following due to De Rham (see, for example, [6]) Theorem 1 If a Riemannian manifold is complete, simply connected and if the holonomy of its tangent bundle with respect to the Levi-Civita connection is reducible, then this manifold is a product metrically. Thus if M is simply-connected we see that M factorizes exactly into parts labelled by the type of supermultiplet containing the massless scalars. If M is not simply-connected we may pass to the universal cover and use this theorem again. The general statement is therefore that the moduli space factorizes up to the quotient of a discrete group acting on the product. $ Actually we should treat the word “complete” in the above theorem with a little more care. There are nasty points at finite distance in the moduli space where the manifold structure breaks down. These points alsoleadtoa breakdowninthe factorizationofthe modulispace. Theseextremaltransitionswill be studied more in section 4.2. We should only say that the moduli space factorizes locally away from such points. We may now analyze the structure of each factor of M from the Berger-Simons theorem (for an account of this we refer again to [6]) which states that the manifold must appear as a row in the following list:2 Hol0 dim(M) SO(n) n U(n) 2n SU(n) 2n (3) Sp(1).Sp(n) 4n Sp(n) 4n Spin(7) 8 G 7 2 or be a “symmetric space” (which we will define shortly). Note that the following names are given to some of these holonomies: 2The notation Sp(1).Sp(n) means Sp(1) Sp(n) divided by the diagonal central Z . 2 × 5 Manifold Real Dimension A I SL(n,R)/SO(n) 1(n 1)(n+2) 2 − A II SU∗(2n)/Sp(n) (n 1)(2n+1) − A III SU(p,q)/S(U(p) U(q)) 2pq × BD I SO (p,q)/(SO(p) SO(q)) pq 0 × D III SO∗(2n)/U(n) n(n 1) − C I Sp(n,R)/U(n) n(n+1) C II Sp(p,q)/(Sp(p) Sp(q)) 4pq × E I E /Sp(4) 42 6(6) E II E /(SU(6) SU(2)) 40 6(2) × E III E /(SO(10) U(1)) 32 6(−14) × E IV E /F 26 6(−26) 4 E V E /SU(8) 70 7(7) E VI E /(SO(12) SU(2)) 64 7(−5) × E VII E /(E U(1)) 54 7(−25) 6 × E VIII E /SO(16) 128 8(8) E IX E /(E SU(2)) 112 8(−24) 7 × F I F /(Sp(3) SU(2)) 28 4(4) × F II F /SO(9) 16 4(−20) G I G /(SU(2) SU(2)) 8 2(2) × Table 1: Symmetric Spaces K¨ahler if Hol U(n), • ⊂ Ricci-flat K¨ahler if Hol SU(n), • ⊂ Quaternionic K¨ahler if Hol Sp(1).Sp(n), • ⊂ Hyperk¨ahler if Hol Sp(n). • ⊂ A symmetric space is a Riemannian manifold which admits a “parity” Z -symmetry 2 about every point. This parity symmetry acts as 1 in every direction on the tangent space. − All symmetric spaces are of the form G/H for groups G and H, where the holonomy is given by H. They have been classified by E. Cartan and we list all the noncompact forms in table 1.3 The noncompact forms are the ones relevant to moduli spaces. 3WehavebeensloppyaboutthepreciseglobalformofthegroupH. Aslistedoneoftenneedstoquotient by a finite group to get the correct answer. For example in entry “E V”, SU(8) is not a subgroup of E 7(7) 6 A key point to note here is that the symmetric spaces are rigid — they have no de- formations of the metric which would preserve the holonomy. The same is not true for the non-symmetric spaces listed in (3). Thus if the holonomy is of a type which forces a symmetric space as the only possibility we will refer to this as a rigid case. Let us consider a few examples. N = (1,1) in 6 dimensions (i.e., one left-moving supersymmetry and one right-moving • supersymmetry). This implies that the R-symmetry is Sp(1) Sp(1) = SO(4) (up × to irrelevant discrete groups). Analysis of the supermultiplets shows that matter su- permultiplets have 4 scalars transforming as a 4 of SO(4). If we only had one such supermultiplet we could say nothing about the moduli space as a generic Riemannian manifold of 4 dimensions has holonomy SO(4). If we have a generic number, n, of supermultiplets and assuming the moduli space doesn’t factorize unnaturally then a quick look at the list above shows that the only possibility is the symmetric space. SO (4,n)/(SO(4) SO(n)). Thus this case is rigid. The gravity supermultiplet has a 0 × single scalar giving an additional factor of R to the moduli space. N = (2,0) in 6 dimensions. This implies that the R-symmetry is Sp(2) = SO(5) • (up to irrelevant discrete groups). Analysis of the supermultiplets shows that matter supermultiplets have 5 scalars transforming as a 5 of SO(5). If we have a generic number, n, of supermultiplets and assuming the moduli space doesn’t factorize un- naturally then the list above shows that the only possibility is the symmetric space SO (5,n)/(SO(5) SO(n)). Thus this case is rigid. There are no further moduli. 0 × N = 4 in 4 dimensions. This implies that the R-symmetry is U(4) = SO(6) U(1) (up • × toirrelevant discrete groups). Analysis ofthesupermultiplets shows thatmatter super- multiplets have 6 scalars transforming as a 6 of SO(6). If we have a generic number, n, of supermultiplets and assuming the moduli space doesn’t factorize unnaturally then the only possibility for this factor is the symmetric space SO (6,n)/(SO(6) SO(n)). 0 × Thus this case is rigid. The gravity multiplet contains a complex scalar transforming under the U(1) factor of the holonomy. By holonomy arguments this contributes a complex K¨ahler factor to the moduli space. Closer analysis of this supergravity shows that this factor is actually SL(2,R)/U(1). This last example demonstrates an important point. Analysis of the R-symmetry may be sufficient to imply that we have a rigid moduli space but sometimes the moduli space is rigid even when the holonomy may imply otherwise. A more detailed analysis of the supergravity action is required in some cases to show that we indeed have a symmetric space. The rule of thumb is as follows: If we have maximal (32 supercharges, e.g. N = 8 in four dimensions) — the correct form of H should be SU(8)/Z . The notation SO (p,q) refers to the part of the Lie group 2 0 connected to the identity. 7 or half-maximal (16 supercharges, e.g. N = 4 in four dimensions) supersymmetry then, and usually only then, is the moduli space rigid. Note that there are a few strange examples such as [3] where the moduli space is rigid even when there are fewer than 16 supercharges. 2.2 U-Duality In this section we will focus on global properties of the rigid moduli spaces. The analysis of the moduli spaces so far is not quite complete. The problem is that the moduli space need not be a manifold. There may be singular points corresponding to the theories with special properties. In the rigid case however the fact that the moduli space is symmetric wherever it is not singular is a very powerful constraint. Let us suppose first that we have an orbifold point. That is a region in the moduli space which looks locally like a manifold divided by a discrete group fixing some point x. Away from the fixed point set the moduli space is symmetric and thus “homogeneous”. That is, there exist a transitive set of translation symmetries. Assuming geodesic completeness of the moduli space, these translations may be used to extend the local orbifold property to a global one. That is, the moduli space is globally of the form of a manifold divided by a discrete group [7]. This homogeneous structure of the moduli space may also be used to rule out other possibilities of singularities which occur at finite distance. Consider beginning at a smooth point in moduli space and approaching a singularity. The homogeneous structure implies that nothing about the local structure of the moduli space may change as you approach the singularity — everything happens suddenly as you hit the singularity. This rules out every other type of “reasonable” singularity that one may try to put in the moduli space. To be completely rigorous would require us to make precise technical definitions about the allowed geometry of the moduli space. Instead we shall just assert here that any type of singularity at finite distance that one might think of (such as a conifold) would ruin the homogeneous nature of the moduli space and so is not allowed in the rigid case. We therefore arrive at the conclusion that the only allowed global form of a rigid moduli space is of a symmetric space divided by a discrete group. Thisimpliesthatanyanalysisofthemodulispaceofstringtheoriesinthecaseofmaximal or half-maximal supersymmetry comes down to question of this discrete group. This group is precisely the group known as S-duality, T-duality or U-duality depending on the context. Many examples of such dualities were discussed in [1] and we refer the reader there for details as well as references. For example the general rule is that a space locally of the form SO (p,q)/(SO(p) SO(q)) becomes 0 × O(Υ ) O(p,q)/(O(p) O(q)), (4) p,q \ × where Υ is some lattice (often even and unimodular) of signature (p,q) and O(Υ ) is its p,q p,q discrete group of isometries. 8 Indeed the only interesting question one may ask about the moduli space in the rigid case is what exactly this discrete groupis! Any quantum corrections to the localstructure are not allowedduetorigidity. Itisnotthereforesurprising thatS,TandU-dualityaresoubiquitous when studying theories with a good deal of supersymmetry. As we will we see however, the picture becomes quite different when the supersymmetry is less that half-maximal. One final word of warning here. We have not been clear about what we mean by a “class” of string theories. If we determine the moduli space of some kind of string compactified on some kind of space, up to topology, then our moduli space may have numerous disconnected components. In this case the above results apply to each component separately. This reducibility often happens when we have half-maximal supersymmetry. 2.3 Eight supercharges Now we turn our attention to theories with quarter-maximal supersymmetry, or a total of 8 supercharges. Here we will also specify how one might obtain such a theory from string theory. If we compactify a ten-dimensional supersymmetric theory on R1,d−1 M, where M is × some compact manifold, then holonomy arguments may be again used to determine the number of unbroken supersymmetries in R1,d−1. This time it is the holonomy of the compact space M rather than the moduli space which we analyze. The basic idea is roughly that a symmetry in ten dimensions will be broken by the holonomy of (a suitable bundle on) M to the centralizer of this holonomy group. That is, a symmetry in uncompactified space is bro- ken if it can be transformed by parallel transport around a loop in the internal compactified dimensions.4 We begin with N = (1,0) in six dimensions. We may obtain this by compactifying a heterotic string theory on a four-dimensional manifold with holonomy SU(2). The only such manifold is a K3 surface. We refer to [1] for an explanation of these points. The R-symmetry in this case is Sp(1). An analysis of supermultiplets show that scalars may occur in either of two types: 1. The Hypermultiplet contains 4 scalars which we may view as a quaternion. The holon- omy Sp(1) may then be viewed as multiplication on the left by another quaternion of unit norm. 2. The Tensor multiplet contains a single scalar. Thus holonomy tells us nothing inter- esting. Note that the vector supermultiplet contains no scalars. The moduli space of such theories will locally factorize into a moduli space of hypermultiplets, which will be quaternionic K¨ahler and a moduli space of tensor multiplets. 4While this seems a very reasonable statement it is probably not rigorous. Breaking the gauge group of the heterotic string in this way does not always lead to the correct global form. 9