IPMU??-????

Supersymmetric partition functions and higher

genus GW invariants

Hans Jockers†1 , Albrecht Klemm† , Mauricio Romo∗2

∗Kavli Institute for the Physics and the Mathematics of the Universe
The University of Tokyo, 5-1-5 Kashiwanoha, Kashiwa 277-8583, Japan
† Bonn

Abstract

1
jockers-at-uni-bonn.de
2
mauricio.romo-at-ipmu.jp
Contents
1 Review of BCOV equations 2

2 Genus 0 data from supersymmetric partition functions 3

3 The topological limit 4

4 GLSM for the GN Calabi-Yau 5

5 Flop transition 6

6 Genus 1 GW invariants 6

7 Fundamental period and mirror map 7

8 B-branes near a large volume point on the GN model 8

1 Review of BCOV equations

Two-dimensional N = 2 theories with at least one nonanomalous U (1)R symmetry posses
topological subsectors given by Q-exact operators. Here we will focus on superconformal
theories with ĉ = 3. There are two inequivalent choices for Q plus their hermitian conjugates.
Given a choice of Q, we have a chiral ring spanned by the chiral operators {φa }. Chiral
operators form a ring in Q-cohomology:
c
φa φb = Cab φc + Q(. . .) (1.1)

Moreover, they can be uniquely mapped to a vacuum state φa → |ai. A completely equivalent
story is valid for Q† -cohomology i.e. antichiral operators {φ̄a }. By tt∗ and topological fusion,
one can define two distinct metrics in parameter space:

ηba = ha|bi gbā = hā|bi (1.2)

η is called the topological metric and g the tt∗ metric. These operators can be classified w.r.t.
their left and right U (1) charges. For ĉ = 3 SCFTs, the allowed values are (0, 0), . . . , (3, 3).
The operator corresponding to the distinguished vacuum without any insertion is (0, 0) and
the (1, 1) operators corresponds to the marginal directions. The other operators are the duals
w.r.t. η, of the formers. We will denote the (1, 1) operators by a subindex i, j, etc. They can
be used to deform the theory without losing superconformality, by adding to the action the
terms: Z
i
t d2 zd2 θφi + h.c. (1.3)

the partition function is Rnormalized such that taking derivatives w.r.t. the parameters t̄i
brings down an operator d2 zd2 θφ̄i . Under this choice of coordinates:

∂ i |ai = 0 (1.4)

2
this choice of coordinates and normalization of the partition function is called holomorphic
gauge (is the same as flat/special coordinates?). The Yukawa couplings are defined by the
following topological correlator on S 2 :
l
Cijk = h0|φi φj φk |0i = Cij ηlk (1.5)

denote by M the space spanned by the coordinates (t, t̄). Then, |0i spans a holomorphic
line bundle L → M. The manifold M enjoys a local special Kähler structure with a Weil-
Petersson metric given by the correlator of (1, 1) chirals on S 2 :
Z
(2) (2) (2)
Gij̄ = hφi (1)φ̄j (0)iS 2 φi := d2 θφi (1.6)

so, locally,
Gij̄ = ∂i ∂ j K K = − logh0̄|0i (1.7)
Its relation with the tt∗ metric is given by:

gij̄ = e−K Gij̄ (1.8)

Choosing the basis for the vacua given by {|0i, |ii, |īi, |0̄i}, it spans the following vector bundle
over M:
L ⊕ (T ⊗ L) ⊕ (T ⊗ L) ⊕ L T = T (1,0)∗ M (1.9)
there exists a holomorphic connection over this bundle, compatible with gij̄ , is given by (the
rest of the components vanish):

Akij = g kl̄ ∂i gl̄j Ak̄īj̄ = g k̄l ∂ī glj̄ A0i0 = −∂i K A0̄ī0̄ = −∂ī K (1.10)

the connection A just defined, is equivalent to the Levi-Civita of the Weil-Petersson metric
plus the connection −∂i K on L. We will denote it by

Di = ∂i + Γi − ∂i K Γkij = Gkl̄ ∂i Gl̄j (1.11)

Now, we have all the ingredients to write the BCOV equations:

1  χ
∂ i Fj1 = Cjkl Cīkl + 1 − Gj ī (1.12)
2 24
g−1
!
g 1 jk g−1
X
r g−r
∂ i F = Cī Dj Dk F + Dj F Dk F (g > 1) (1.13)
2
r=1

2 Genus 0 data from supersymmetric partition functions

Our first step into determining higher genus GW invariants is to express the genus 0 data in
terms of the supersymmetric partition functions of GLSMs. From the conjecture

ZS 2 = e−K (2.1)

we can easily write the Weil-Petersson and tt∗ metrics:

∂i ZS 2 ∂j̄ ZS 2 − ZS 2 ∂i ∂j̄ ZS 2 ∂i ZS 2 ∂j̄ ZS 2 − ZS 2 ∂i ∂j̄ ZS 2
Gij̄ = gij̄ = (2.2)
ZS2 2 ZS 2

3
In the following we will assume we are working in flat/special coordinates. The map between
GLSM FI-theta parameters and flat coordinates can be read from Z2 after properly normaliz-
ing the constant term, proportional to ζ(3) or directly computed by the hemisphere partition
function ZD2 provided we find the boundary conditions that give us the regular period and
the ”log(z) period” (NOTE: This is quite simple in practice, for our examples at
least. Could be a nice addition/concistency check?).
For g > 1 we will also need the value of the Yukawa couplings Cijk . In order to get this from
ZS 2 , recall that, in a symplectic basis for the vacuum bundle:
 
I
e−K = i X FI − X I F I (2.3)
then
e−K
= 2i= 2F + (t − t̄)i Fi


0 2
(2.4)
i|X |
so  −K 
∂3 e
Cijk = i j k = Fijk (2.5)
∂t ∂t ∂ t̄ i|X 0 |2
We also need the propagators S ij , S i , S, all sections of L−2 . S ij can be obtained by solving
the algebraic equations:
S ij Cjkl = δji Kk + δki Kl − Γikl + fkli
(2.6)
i are meromorphic functions, that can be chosen with the only condition that we
where fkl
can solve the equations for S ij (since, for dimM > 1, this is an overdetermined system of
equations). The other propagators can be obtained by integrating the equations:
Giī ∂ ī S j = S ij
Giī ∂ ī S = S i (2.7)

3 The topological limit

Here we review the topological limit of the the free energies F g (t, t̄). The topological limit, also
called the holomorphic limit, is obtained by fixing the special coordinates ti to an arbitrary
value and rescaling t̄i → s̄t̄i , then taking the limit s̄ → ∞. Consider the Kähler potential
I
e−K = X FI − X I F I (3.1)
then, the leading order in s̄ is given by |Π0 |2 s̄3 g(t̄),
so, we define:

K top = − ln(Π0 ) (3.2)
another relevant quantity is the WP metric. Its leading order in s̄ is given by, in special
coordinates
1
Gij̄ = 2 Lij̄ (t̄) + O(s̄−3 ) (3.3)
s̄
then, in algebraic coordinates:
1 dtl dt̄k
Gαβ̄ = Ll k̄ ( t̄) + O(s̄−3 ) (3.4)
s̄2 dz α dz̄ β
so, we can define the limit of the following relevant quantities, in algebraic coordinates:
   −1  
 ∂t  ∂t ∂t
ln det G top = ln det
 Γα top =
 ∂α (3.5)
∂z ∂z ∂z

4
4 GLSM for the GN Calabi-Yau
Here we review some geometric properties of the GN Calabi-Yau and its GLSM presentation.
The matter content of the GN model in the PAX presentation (G = U (1) × U (2)) is given by:

φ1,...,8 x1,...,4 p1,...,4

U (1) 1 0 −1
(4.1)
U (2) 0  
U (1)V 2ε1 2ε2 2 − 2(ε1 + ε2 )
and, the GLSM potential of this model is given by
X
W = pαi Aaij φa xj α = tr(pA(φ)x) (4.2)
i,j,a

and the D-term equations are given by:

8
X
|φa |2 − tr(p† p) = ζ0
a=1
pp† − x† x = ζ1 12 (4.3)

This model has three large volume points. We will concentrate on the one located deep in
the phase:

Phase I : {ζ0 + 2ζ1 > 0} ∩ {ζ1 > 0} (4.4)

In this phase only φ and p get VEVs, and the IR dynamics is determined by the NLSM with
target:

XAT = {(φ, p) ∈ P7 × G(2, 4)|pA(φ) = 0} (4.5)

therefore, the R-charges in the IR must be given by the limit in the ε’s parameters that fix
the R-charges of these fields to 0 i.e. ε1 → 0+ , ε2 → 1. Since h1,1 (XAT ) = 2, the generators of
H 2 (XAT , Z) are identified with the FI parameters: the restriction of σ = c1 (Q) from G(2, 4)
is identified with ζ1 and the hyperplane class H of P7 with ζ0 + 2ζ1 . Then, is convenient
to define the following coordinates, which will be identified with the algebraic coordinates in
this phase:

z = e−2πζ0 +iθ0 +2(−2πζ1 +iθ1 ) w = e−2πζ1 +iθ1 (4.6)

in terms of w and z, the principal components of the discriminant are given by

∆1 = (1 − w)4 − 2(1 + 6w + w2 )z + z 2
∆2 = −(1 − w)8 + 4(1 − w)4 (1 − 34w + w2 )z − 2(3 + 372w + 1298w2 + 372w3 + 3w4 )z 2
+4(1 − 34w + w2 )z 3 − z 4 (4.7)

5
5 Flop transition
for the GN Calabi-Yau, the phases with ζ0  0 and ζ0  0 are related by a flop transition. In
the interphase, the curves with homology class η gets contracted. In the phase {ζ1 < 0, ζ0 > 0}
we have a CY X and η can be identified with a curve in P7 . Its dual divisor is identified with
Dη = H − 2σ1 in this phase and so Dη3 = 28. In the phase {ζ1 < 0, 2ζ1 + ζ0 < 0} we have a
CY X 0 and the dual divisor to η is identified to be −H + 2σ1 . Therefore, by the formula by
Wilson:
X
Dη (X 0 )3 = Dη (X)3 − d3 nd (η) (5.1)
d>0

where nd (η) are the GW invariants on the homology class η, we get that all GW invariants
in the class q0k (k > 1)) vanish. By symmetry, we get the same for the GW invariants in the
class (q0 q12 )k (k > 1).

6 Genus 1 GW invariants
Recall the BCOV equation for Fj1 :

1  χ
∂ i Fj1 = Cjkl Cīkl + 1 + Gj ī (6.1)
2 24
In order to integrate this equation, first, we use
jk
C ī = ∂ ī S jk (6.2)

then we can integrate the anti-holomorphic derivative to:

1  χ
Fj1 = Cjkl S kl + 1 − Kj + fj1 (6.3)
2 24
here, fj1 are meromorphic functions which we have to fix by boundary conditions. Next, we
use that, from tt∗ and special geometry:

S ij Cijl = (n + 1)Kl − Γiil + fl = (n + 1)Kl − det G−1 ∂l (det G) + fl (6.4)

the meromorphic functions fl can be combined with fl1 , therefore, we get:

1 χ 1
F1 = 3+n− K − log det G + f 1 (6.5)
2 12 2
the function f 1 encodes the holomorphic ambiguity and can be chosen as
X
f1 = sa ln ∆a (6.6)
a

where the holomorphic functions ∆a are the different components of the discriminant locus,
including the limiting points, i.e. ∆α = z α . Notice that we can write F 1 entirely in terms of
ZS 2 as
χ/12−3+n Q
!
ZS 2 sa
1 1 a ∆a
F = ln (6.7)
2 det(∂t ZS 2 ∂ t̄ ZS 2 − ZS 2 ∂ t̄ ∂t ZS 2 )

6
For computing the instanton numbers, is more convenient to take the topological limit. Now,
we specialize to the GN CY, so, we set χ = −64 and n = 2. We get that, in the topological
limit the genus one free energy i given by (up to an additive constant):
 0 1
dt0 dt1
  
1 1
 1 64 1 dt dt
F = F top = − 5+ ln(Π0 )− ln − +r0 ln z+r1 ln w+r2 ln ∆1 +r3 ln ∆2
2 12 2 ∂z ∂w ∂w ∂z
(6.8)
when expressed in terms of the special coordinates t0 , t1 , F 1 takes the form:
!
1 c2 · J X n0η
F = + + nη Li1 (q η )
1
(6.9)
24 12
η∈H2 (X,Z),η6=0

where J is the complexified Kähler class, n0η are the genus zero GW invariants and n1η the
genus one GW invariants. To fix the coefficients ra , we use the asymptotics, given by
c2 · J 7
= 2πi(t0 + t1 ) (6.10)
24 3
and also we can use the fact that the flop transition imposes that the numbers n1s,0 and n1s,2s
must vanish (here we use the notation n1s0 ,s1 where s0 and s1 are the classes of the curves C1
and Γ − C0 , respectively). therefore the coefficients of q0s and q0s q12s takes the form:
14 s 14 s 2s
q q q (6.11)
3s 0 3s 0 1
all these conditions fix:
17 1
r0 = r1 = − r2 = r3 = − (6.12)
6 12

7 Fundamental period and mirror map

The following periods were derived with the hemisphere partition function.
The fundamental period is given by:
∞ X
∞ X
k1
X Γ(k1 − k + k0 + 1)4 Γ(k + k0 + 1)4
Π0 = 2 z k0 wk1
Γ(1 + k0 )8 Γ(k + 1)4 Γ(1 + k1 − k)4
k0 =0 k1 =0 k=0
 
1
× (k1 − 2k) (ψ(k1 − k + k0 + 1) − ψ(k1 − k + 1) − ψ(k + k0 + 1) + ψ(k + 1)) + (7.1)
2

We also have periods proportional to log(−w) and log(z):

∞ X
∞ X
k1
(1)
X Γ(k1 − k + k0 + 1)4 Γ(k + k0 + 1)4
Π1 = Π0 log(−w) + 2 z k0 wk1
Γ(1 + k0 )8 Γ(k + 1)4 Γ(1 + k1 − k)4
k0 =1 k1 =0 k=0
h k1
 
× k− ψ (1) (1 + k1 − k + k0 ) − ψ (1) (1 + k + k0 ) − ψ (1) (1 + k1 − k) + ψ (1) (k + 1)
2
+ ψ(1 + k1 − k + k0 ) (2(2k − k1 ) (ψ(1 + k1 − k + k0 ) − 2ψ(1 + k1 − k)) − 1)
 
1
+ ψ(1 + k1 − k) (2(2k − k1 )ψ(1 + k1 − k) + 1) + 2ψ(1 + k) + (k1 − 2k)ψ(1 + k)
2

7
 i
+ ψ(1 + k + k0 ) (2(−2k + k1 ) (ψ(1 + k + k0 ) − 2ψ(1 + k)) − 1) (7.2)

∞ X
∞ X
k1
(0)
X Γ(k1 − k + k0 + 1)4 Γ(k + k0 + 1)4
Π1 = Π0 log(z) − 4 z k0 w k1
Γ(1 + k0 )8 Γ(k + 1)4 Γ(1 + k1 − k)4
k0 =1 k1 =0 k=0
h 
k1  (1) 
× k− ψ (1 + k1 − k + k0 ) − ψ (1) (1 + k + k0 )
2
+ ψ(1 + k1 − k + k0 ) (2(2k − k1 ) (ψ(1 + k1 − k + k0 ) − ψ(1 + k1 − k) − ψ(1 + k) − 2ψ(1 + k0 )) − 1)
+ (2(2k − k1 )(ψ(1 + k + k0 ) − ψ(1 + k)) + 1) (2ψ(1 + k0 ) − ψ(1 + k + k0 ))
i
− 2(−2k + k1 )ψ(1 + k1 − k) (2ψ(1 + k0 ) − ψ(1 + k + k0 )) (7.3)

then, the maps to flat coordinates are given by

(a)
a Π
2πit = 1 a = 0, 1 (7.4)
Π0

8 B-branes near a large volume point on the GN model

The hemisphere partition function for the GN model reads:
Z
ZD2 (B) = C(rΛ)ĉ/2 d3 τ (τ1 − τ2 ) sin(π(τ1 − τ2 ))Zmatter (τ )et(τ ) fB (τ ) (8.1)
γ

with

Zmatter (τ ) = Γ (τ0 + ε1 )8 Γ (−τ1 + ε2 )4 Γ (−τ2 + ε2 )4 Γ (−τ0 + τ1 + 1 − ε2 − ε2 )4

×Γ (−τ0 + τ2 + 1 − ε2 − ε2 )4 (8.2)

and

t(τ ) = t1 (τ1 + τ2 ) + t0 (τ0 ) tl = ζl − iθl (8.3)

In order to keep the R-charges inside the allowed range required for localization:

ε1,2 > 0 0 < ε1 + ε2 < 1 (8.4)

The divisors where the poles are located are given by:
[ [
D1 = {τ0 = −k − ε1 } D2 = {−τ0 + τ1 = −k − 1 + ε1 + ε2 }
k∈Z≥0 k∈Z≥0
[ [
D3 = {−τ0 + τ2 = −k − 1 + ε1 + ε2 } D4 = {−τ1 = −k − ε2 }
k∈Z≥0 k∈Z≥0
[
D5 = {−τ2 = −k − ε2 } (8.5)
k∈Z≥0

All poles are in the subspace (<(τ0 ), <(τ1 ), <(τ2 )) of the C3 spanned by (τ0 , τ1 , τ2 ). Poles are
located in the regions:

D1 → <(τ0 ) < 0 D2 → −<(τ0 ) + <(τ1 ) < 0

8
 D3 → −<(τ0 ) + <(τ2 ) < 0 D4 → −<(τ1 ) < 0
D5 → −<(τ2 ) < 0 (8.6)

This model has 3 large volume points. We will concentrate on the one located deep in the
phase:

Phase I : {ζ0 + 2ζ1 > 0} ∩ {ζ1 > 0} (8.7)

In this phase only φ and p get VEVs, therefore, the R-charges in the IR must be given by the
limit in the ε’s parameters that fix the R-charges of these fields to 0 i.e. ε1 → 0+ , ε2 → 1.
The convergence of the partition function, deep in a phase, is dominated by the exponential
(in general this can be not true, but when the gauge group is completely broken, or broken
to a discrete subgroup, at generic points inside the phase, this holds). Then, the range of τ
parameters where the exponential et(τ ) → 0 is

<(τ0 )  0 <(−2τ0 + τ1 + τ2 )  0 (8.8)

from an analysis of the Cartan model, we conclude that we must take Grothendieck residues
of the poles contained in the cone

C = {<(τ0 ) ≤ 0} ∩ {−<(τ0 ) + <(τ1 ) ≤ 0} ∩ {−<(τ0 ) + <(τ2 ) ≤ 0} (8.9)

then, we can take the cone in τ -space:

Π = iR3 + C (8.10)