View metadata, citation and similar papers at core.ac.uk brought to you by CORE
provided by CERN Document Server
DUK-TH-94-68
IASSNS-HEP-94/23
hep-th/9404151
April, 1994
String Theory
on K3 Surfaces
Paul S. Aspinwall,
Scho ol of Natural Sciences,
Institute for Advanced Study,
Princeton, NJ 08540
David R. Morrison
Department of Mathematics,
Box 90320,
Duke University,
Durham, NC 27708-032 0
ABSTRACT
The mo duli space of N =(4,4) string theories with a K3 target space is determined,
establishing in particular that the discrete symmetry group is the full integral orthog-
onal group of an even unimo dular lattice of signature (4,20). The metho d combines an
analysis of the classical theory of K3 mo duli spaces with mirror symmetry. A descrip-
tion of the mo duli space is also presented from the viewp oint of quantum geometry,
and consequences are drawn concerning mirror symmetry for algebraic K3 surfaces.
1 Intro duction
In recentyears, Calabi-Yau manifolds have received great attention in the string literature.
This is mainly b ecause compacti cation on such spaces may b e used to reduce the number
of dimensions in mo dels built from the intrinsically ten dimensional critical sup erstring [1].
The fo cus has largely b een on the case of complex dimension d = 3, since the corresp onding
compacti cation is to four space-time dimensions. (It is also the case that, in many resp ects,
within top ological eld theory on Calabi-Yau target spaces the case d = 3 plays a sp ecial
r^ole [2, 3].) However, one might exp ect interesting prop erties for other values of d as well.
In this pap er we analyze the case d = 2 whose sp ecial features derive from the fact that
the top ology of the target must b e either a torus or a K3 surface. The classi cation of toroidal
target spaces has b een known for some time [4, 5]. The lo cal geometry of the mo duli space
of conformal eld theories with K3 target space top ology was understo o d shortly afterwards
[6] and although some asp ects of the global form of the mo duli space have b een conjectured
and studied [6, 7, 8 ], more precise statements concerning this question have remained elusive.
We will use mirror symmetry to address this question of the form of the mo duli space.
We nd that, given a couple of minor assumptions, combining a -mo del analysis near the
eld theory limit with the study of a mirror symmetry transformation is sucient to give the
precise form of the mo duli space. This pap er is not intended to provide the full mathematical
exp osition of the analysis required | for that the reader is referred to [9]. Here we only give
a brief summary of the metho ds used and conclusions reached by the analysis.
The analysis of the mo duli space of string theories on a K3 surface di ers markedly
from that of the more familiar Calabi-Yau mo duli spaces related to string compacti cations
to four dimensions. Firstly, in studying the case where the target space X is a Calabi-Yau
2;0
threefold, the assumption h = 0 is usually made. Under this assumption, the deformations
of complex structure of X and of the (complexi ed) Kahler form \decouple" to give the
complete mo duli space a lo cal pro duct structure [10] (at least over generic p oints in the
mo duli space). These twotyp es of deformations may then b e studied indep endently.For a
2;0
K3 surface wehaveh = 1 and the ab ove structure is lost. We are thus forced to analyze
deformations of complex structure and deformations of Kahler form together.
In another resp ect however the K3 case is simpler | the N =(2,2) sup erconformal invari-
ance of a string with Calabi-Yau target space is extended to N =(4,4) sup ersymmetry in the
case d = 2 [11]. This is equivalent to the geometric statement that a Calabi-Yau manifold in
complex dimension two admits a hyp erkahler metric. This N =(4,4) structure serves to x
the lo cal form of the mo duli space completely. This can b e contrasted with the case d =3
where the extended chiral algebra [12] has yet to provide much insightinto the classi cation
of string theories.
As in the case d = 3, the lo cal dimension of the mo duli space of conformal eld theories 1
can b e matched to the dimension of the mo duli space of the geometrical ob jects representing
the target space if we regard the \geometrical ob ject" as including, in addition to the target
space metric, the sp eci cation of a B - eld, i.e., an element of the second real (de Rham)
cohomology group of the target. In the case d = 3, the B - eld naturally combined with the
Kahler form to provide a complexi ed Kahler form. In the case of K3 this is clearly imp ossible
| B lives in a 22-dimensional space whereas the dimension of the space of Kahler forms
varies with the complex structure but is at most 20. Despite this fact we will still nd that
a b eautiful structure arises when the B - eld is included in the mo duli space.
At the top ological level, mirror symmetry might app ear rather trivial for a K3 surface
since the mirror of a K3 surface (equipp ed with metric and B - eld) is another suchK3
surface. However, the geometric data of the mirror K3 surface thus obtained is not, in
general, isomorphic to the original. Thus the mirror map acts as a non-trivial automorphism
on the \classical" mo duli space of K3 surfaces (i.e., the -mo del mo duli space of metrics and
B - elds). The mirror map in this context was rst studied long ago in [13] and some of the
observations in that pap er lead to some of the metho ds used here.
Although the purp ose of this pap er is to concentrate on the case where the target space
is a K3 surface, it should b e noted that much of what follows applies to any target space
2;0
with h = 1. The only steps in the following argument which will not b e directly applicable
to the determination of the mo duli space of this more general case are the explicit form of
the mo duli space of Einstein metrics, and the particular self-mirror K3 theory used.
In section 2 we will present the outline of the construction of the mo duli space of N =(4,4)
and N =(2,2) theories. In section 3 we discuss the interpretation of the mo duli space in terms
of the space of total cohomology of K3. In section 4, some asp ects of mirror symmetry on
algebraic K3 surfaces are discussed.
2 The Mo duli Space
Let us rst x some notations. nG=H denotes the double coset space resulting from dividing
the group G by the right-action of H and left-action of . (G and H will always b e continuous
a;b
Lie groups and will b e discrete.) denotes the unique (up to isomorphism) even self-
dual lattice in (a + b)-dimensional space with signature (a; b), when ab 6= 0. Basis vectors
may b e taken such that the inner pro duct has the following form:
a;b
: h; i ( E ) ( E ) ::: H H :::; (1)
=
8 8
| {z }
n times 2
where b a =8n (one uses E in place of E when b a<0), E is the Cartan matrix of
8 8 8
the Lie algebra E , and H is the hyp erb olic plane:
8
!
0 1
H = : (2)
1 0
We will use X to denote a sp eci c smo oth K3 surface. One can show that the intersection
2 3;19
pairing on H (X; Z) gives it the structure of a lattice isomorphic to (see for example
a;b
[14]). Let R bea(a+b)-dimensional space with an inner pro duct of signature (a; b).
a;b a;b
a;b a;b
O (a; b) is the orthogonal group on R and O ( ) is the subgroup preserving R .
Let
a;b
T = O (a; b)= ( O (a) O (b)) : (3)
a;b
a;b
We can identify T as the set of space-like a-planes in R (i.e., a-planes on which the
a;b
inner pro duct is p ositive-de nite). T can also b e regarded as one of the two connected
comp onents of the set of oriented space-like a-planes. Under this latter interpretation, we
can write
a;b +
T = O (a; b)= (SO(a) O(b)) ; (4)
+
where O (a; b) is the index 2 subgroup of O (a; b) with the same connected comp onents as
SO(a) O(b).
It was shown in [6] using arguments from sup ergravity [15] that the required mo duli space,
4;20
M, of conformal eld theories on a K3 surface is lo cally of the form T . This is probably
b est understo o d from the argument presented in [16], whichwenow review. The N =(4,4)
sup erconformal algebra contains ane SU (2) algebras in b oth the left and right sectors. This
symmetry acts on the marginal op erators spanning the tangent spaces of M. The existence
of such a symmetry restricts the form of the holonomy group of the Zamolo dchikov metric of
M. The restriction is so severe, in fact, that it then follows from Berger's classi cation [17]
of holonomy groups that either M is lo cally isomorphic to a quaternionic symmetric space,
or that the holonomy is reducible. Moreover, the non- atness of the Zamolo dchikov metric
is enough to rule out the reducible holonomy case. A simple analysis of any conformal eld
1
theory giving rise to a K3 target space tells us that M has real dimension 80 [6, 11 ]. This,
together with the known classi cation of quaternionic symmetric spaces (cf. [18]), completes
the pro of.
1
One way to do this count is as follows: the space of complex structures has complex dimension 20 (real
dimension 40), the space of Kahler forms has real dimension 20, and the space of B - elds has real dimension
2
22. However, for each Ricci- at metric on the K3 surface there is an S of complex structures, so the overall
dimension count is40+20+22 2 = 80. 3
Let us now maketwo assumptions ab out the form of M, b oth of whichwe consider to b e
2
quite reasonable. Firstly we assume M is geo desically complete and secondly we assume M
to b e Hausdor . While some non-Hausdor mo duli spaces have app eared in string theory
[19], this happ ened in the context of a target space of inde nite signature. There is no reason
to b elieve that something as unpleasant as this should happ en in the case of a K3 surface.
It then follows that
4;20
M nT ; (5)
=
4;20
for some group acting discretely on T .Thus to complete the description of M,we only
need to nd .
We b egin our determination of by analyzing the classical form of the mo duli space of
4;20
K3 surfaces. Using the techniques of [20 ], one may decomp ose T as
4;20 3;19 3;19
T R T R ; (6)
=
+
where R is the half-line of p ositive real numb ers whichwe parameterize by . The decom-
+
4;20
?
p osition dep ends on a choice of null vector v 2 R , and the space v =Rv provides the
3;19
3;19 3;19
R on which O (3; 19) acts and on which T is based. The space T is known to b e
isomorphic to the Teichmuller space for Einstein metrics of volume one on a K3 surface (and
their orbifold limits) [21 , 22 , 23 ]. It is natural then to identify as giving the size of the K3
3;19
surface and R as the mo duli space of B - elds on the target space. These identi cations
may b e established by lo oking at the metrics on the ab ove mentioned spaces. The Zamolo d-
4;20
chikov metric on T is known to b e the left-invariant metric. From [24] this induces a
metric on each of the terms on the right hand of (6) where this should now b e viewed as
a\warp ed pro duct", i.e., the metric do es not resp ect the pro duct structure. One maynow
show (using [23]) that the metrics induced are those given precisely by the extended Weyl-
Petersson metric in the sense of [25]. Such an isometric identi cation allows us to identify
4;20
every p ointinT with an Einstein metric on a K3 surface (or orbifold metric) together
with a B - eld. This can b e taken as another version of the statement that the non-linear
-mo del is exactly conformally invariant on a K3 surface with Ricci- at metric [26, 27 , 28 ].
It also shows that the Zamolo dchikov metric and Weyl-Petersson metric coincide exactly on
the mo duli space | a fact which in general N =(2,2) theories holds only to leading order in
the large radius limit [29 ].
The mo duli space of smo oth Einstein metrics of volume one on a K3 surface X is deter-
0
3;19
mined in [23] to b e nT Z, where = Di (X )= Di (X ) is the group of comp onents
0 0
of the di eomorphism group of X , and Z is the space of orbifold metrics. (It is generally
b elieved that such orbifolds should b e included when considering a string target space [30 ];
2
As we will p oint out b elow, wemust include orbifold p oints in our mo duli space in order to ensure this
geo desic completeness. 4
3;19
if we include them, we get the geo desically complete space nT .) The discrete group
0 0
is determined in [31, 32 , 33 ] to coincide with the group
+ 3;19 3;19 +
O ( ):= O( ) \ O (3; 19); (7)
3;19 3;19
which has index 2 in O ( ). The \missing" Z in O ( )may b e generated by I acting
2
3;19
on R .
We also know that the non-linear -mo del on a Calabi-Yau manifold is invariant under
3;19
2
all translations B ! B + v , where v 2 H (X; Z). Thus we should divide the space R of
3;19
B - elds by additive translations by . (As an abstract group, these additive translations
22
simply form a Z .) Furthermore one may consider complex conjugation of the target space.
If one considers such a complex conjugation for a -mo del one sees that the transformation
B ! B is required in addition to the conjugation of the complex structure of the target.
Such a transformation may b e represented by I and thus generates the missing Z from
2
3;19 3;19
ab ove. Therefore, in terms of the decomp osition (6) one sees a group O ( ) n (which
3;19
is the full space-group of ) of identi cations that should b e made on the right-hand side
4;20
and thus on T .Thus we obtain
3;19 3;19