
Suppose an isomorphic flop arises at a boundary facet of the K¨ahler cone Kof X. Then, there is an
involution, which, relative to a suitable basis (Di) of divisor classes, can be described by an h×hmatrix
˜
M(satisfying ˜
M2=h×h). If two divisors D=kiDiand D0=k0iDiare related by this involution,
that is, k0=˜
Mk, then it turns out that their associated linear systems have the same dimension [6], so
h0(OX(D)) = h0(OX(D0)). Isomorphic flops can arise across more than one facet of the K¨ahler cone
boundary. In this case, we have multiple involutions and corresponding matrices ˜
M1,..., ˜
Mkwith
˜
M2
1=··· =˜
M2
k=h×h, generating groups ˜
G. Such groups generated by reflections were introduced
by Coxeter [7] and further studied by Tits and Vinberg [8]. Especially the latter studies reflections
along the walls of polyhedral cones, and shows that these correspond to Coxeter groups. This structure
gives rise to infinite sequences of isomorphic flops generated by repeatedly reflecting the K¨ahler cone
along a flop wall and its reflection images under ˜
Mi. The union Kext =∪αKαof their K¨ahler cones is
referred to as the extended K¨ahler cone and is mirror-dual to the complex structure moduli space of
the mirror of X[2]. It turns out, the zeroth cohomology of line bundles on Xis invariant under the
entire group ˜
Gand this fact can be immensely helpful for deriving formulae for cohomology [6]. In the
context of infinite flop chains acting on divisors, the authors of [9] recently studied Euclidean D3 branes
and noted that theta functions also appear in the non-perturbative superpotential of Type IIB.
In the present paper we are interested in the implications of this symmetry for Gopakumar-Vafa (GV)
invariants [10, 11] and the instanton prepotential for K¨ahler moduli. To this end, we introduce a (dual)
basis (Ci) of curve classes and represent arbitrary classes Cby h-dimensional integer vectors dsuch
that C=diCi. Their GV invariants are denoted by nd. On these curve classes, the involutions act via
the matrices Ma=˜
MT
aand the entire group via the dual Gof ˜
G, generated by the matrices Ma. Our
main observation is that classes dwhich do not flop at any of the facets of the (possibly infinite sequence
of) CYs Xα, the GV invariants are unchanged under the action of G, so nd=ngd for all g∈G. This
implies that a part of the instanton prepotential for the K¨ahler moduli (specifically, the part associated
to non-flopping curve classes) is ˜
G-invariant and can be expressed in terms of ˜
G-invariant functions
ψG
d(T) = X
g∈G
e2πi(gd)·T=X
˜g∈˜
G
e2πid·(˜gT )⇒ψG
d(˜gT ) = ψG
d(T)∀˜g∈˜
G , (1.1)
where T=χ+it are the complexifications of the K¨ahler parameters ti. As far as we are aware, these
functions, invariant under certain representations of Coxeter groups, have not been introduced and
studied in this context. We will show that, for certain special cases, depending on the underlying CY
manifold X, they can be expressed in terms of Jacobi theta functions whose appearance can be traced
to an elliptic fibration structure of X. For explicit examples, we will work with complete intersection
CYs in products of projective spaces (CICYs) [12].
Note added: On the day we submitted our paper to arxiv, a revised version of [13] appeared on the
arxiv, which included a discussion of Coxeter groups in the context of the Hulek-Verrill manifold [14].
The plan of the paper is as follows. In Section 2, we start with a simple warm-up example, a CY manifold
with h1,1(X) = 2 and only a single flop boundary, leading to a finite symmetry ˜
G=h˜
M1i∼
=Z2. The
insight from this example is used in Section 3 to study CY manifolds with h1,1(X) = 2 and two flop
boundaries, with the associated groups ˜
G=h˜
M1,˜
M2iisomorphic to universal Coxeter groups with two
generators. In Section 4 we generalize the discussion to manifolds with h1,1(X)>2 but still with two
isomorphic flop boundaries and group ˜
G=h˜
M1,˜
M2i. As we will show, this case exhibits new features,
compared to the h1,1(X) = 2 case, and, in particular, we find that the ˜
Ginvariant functions ψG
dcan
sometimes be expressed in terms of Jacobi theta functions. The general case with arbitrary h1,1(X)
and arbitrary number of flop boundaries, and its relation to Coxeter groups, is discussed in Section 5.
3