Вы находитесь на странице: 1из 47

A UNIFIED APPROACH TO HOLOMORPHIC ANOMALY

EQUATIONS AND QUANTUM SPECTRAL CURVES


arXiv:1808.05343v1 [math-ph] 16 Aug 2018

ZHIYUAN WANG AND JIAN ZHOU

Abstract. We present a unified approach to holomorphic anomaly equations


and some well-known quantum spectral curves. We develop a formalism of
abstract quantum field theory based on the diagrammatics of the Deligne-
Mumford muduli spaces Mg,n and derive a quadratic recursion relation for
the abstract free energies in terms of the edge-cutting operators. This ab-
stract quantum field theory can be realized by various choices of a sequence
of holomorphic functions and suitable propagators, and the realized quantum
field theory can be represented by formal Gaussian integrals. Various applica-
tions are given.

Contents
1. Introduction 2
2. Diagrammatics of Moduli Space of Curves 4
2.1. Moduli space of stable curves and Feynman diagrams 4
2.2. Operators on vector space associated to the set of stable graphs 6
2.3. Abstract free energy 7
2.4. A recursion relation for the abstract free energy 8
3. A Generalization to Labelled Graphs 15
3.1. Diagrammatics for labelled graphs 15
3.2. A recursion relation for the abstract free energy 16
4. Realization of the Abstract Quantum Field Theory 17
4.1. Realization of the abstract QFT by Feynman rules 17
4.2. Representation by formal Gaussian integrals 18
4.3. Realization of the operators on stable graphs 20
4.4. Realization of the recursion relations 21
4.5. The general case 22
5. The holomorphic anomaly equation 23
6. Application to Topological 1D Gravity and Quantum Spectral Curves 25
6.1. Some results in topological 1D gravity 26
6.2. Topological 1D gravity as realization of the diagrammatics of stable
graphs 27
6.3. Example: The Airy function and quantum Airy curve 30
6.4. Example: The Kontsevich-Penner matrix model for N = 1 32
6.5. Example: Enumeration of stable graphs with genus zero vertices 34
6.6. Example: Enumeration of graphs 37
A. Some explicit expressions for[ Fg,n 43
References 46

Bibliography 46
1
2 ZHIYUAN WANG AND JIAN ZHOU

1. Introduction
There are various types of quadratic recursions which have played an important
role in Gromov-Witten theory. The Virasoro constraints are a sequence of recursion
relations [7, 10, 14] for the free energy defined on the big phase space. It involves
differential operators of infinitely many variables. There are also quadratic recursion
relations for the free energy restricted to the small phase space that involves only
finitely many variables. For example, the BCOV holomorphic anomaly equation
[3] and the Eynard-Orantin topological recursion relations [12]. The former was
developed by Bershadsky, Cecotti, Ooguri and Vafa to compute Gromov-Witten
invariants of the quintic Calabi-Yau threefold [2, 3]. The latter was first discovered
by Eynard and Orantin in the setting of matrix models and later formulated in
a form which is conjectured and proved to hold in general for n-point functions.
For an approach to the holomorphic anaomaly equation from the point of view of
Eynard-Orantin topological relation, see Eynard-Marinõ-Orantin [13]. In the case
of Witten-Kontsevich tau-function, it has been shown by the second author that
the Virasoro constraints are equivalent to the EO topological recursion relations
[28]. It is an interesting problem to establish a relationship between the Virasoro
constraints and the EO topological relations in general. The basic ingredient for EO
topological recursion is an algebraic curve with some extra data, called the spectral
curve. A suitable quantization of the spectral curve leads to a Schrödinger equation
satisfied by a partition function constructed from the EO topological recursion,
called the quantum spectral curve (cf. Gukov-Sulkowski [16]). It is a natural
problem to understand the relationship between the holomorphic anomaly equation
and the quantum spectral curves. In this work we will present a formalism that
gives a unified construction of the holomorphic anomaly equation and the quantum
spectral curves in some well-known examples.
To formulate the BCOV holomorphic anomaly equation, it is crucial to first ex-
tend the free energy Fg (t) to a non-holomorphic free energy Fbg (t, t̄) whose geometric
meaning is not mathematically clear at present. It is supposed to be modular-
invariant, and involve contributions from the boundary strata of the moduli spaces.
Originally the holomorphic anomaly is an equation for ∂t̄i Fbg which involves Fbh for
h < g of the following form:
 g−1
X 
1 (0)jk
∂¯t̄i Fbg = C̄ī Dj Dk Fbg−1 + Dj Fbr Dk Fbg−r .
2 r=1

It is shown in [3] that this system of equations can be recursively solved to get
(r<g)
(1) Fbg (t, t̄) = Γ(g) (∆ij , ∆i , ∆, Ci1 ,...,in ) + f (g) (t),
(r<g)
where Γ(g) is a polynomial equation in ∆ij , ∆i , ∆ (propagators) and Ci1 ,...,in (lower
genus vertices). Furthermore, the authors of that work present some explicit expres-
sions of Fbg for g = 2 and 3 and the corresponding Feynman graphs and Feynman
rules (cf. [3, (6.7), Figure 17] and [3, (6.8), Figure 18]). Inspired by Witten [23],
Aganagic, Bouchard and Klemm [1, (2.16)] obtain the following expression for Fbg :
 

−1 IJ
(2) b
Fg (t, t̄) = Fg (t) + Γg − (τ − τ̄ ) , ∂I1 · · · ∂In Fr<g (t) ,
HOLOMORPHIC ANOMALY EQUATIONS AND QUANTUM SPECTRAL CURVES 3

2
∂ F0
where τ = (τij ) = ∂t i ∂tj
). This was interpreted for matrix models using Eynard-
Orantin topological recursion by Eynard, Mariño and Orantin [13], and further-
more, they also reformulate the partition function as a formal Gaussian integral [13,
(4.27)] and present the Feynman graphs and Feynman rules for the terms that con-
tribute to Γg . These results have been generalized to other models by Grimm,
Klemm, Mariño and Weiss [15]. These authors reformulate the holomorphic anom-
aly equation as a quadratic recursion relation for the derivative of Fbg with respective
to the propagators ∆IJ (cf. [15, (7.50)]:
g−1
∂ Fbg 1 1X b
(3) IJ
= DI ∂J Fbg−1 + ∂I Fr ∂J Fbg−r .
∂∆ 2 2 r=1

Here the propagators ∆IJ has the form


1
(4) ∆IJ = − √ ((τ − τ̄ )−1 )IJ + E IJ ,
2 −1
where E IJ is a holomorphic function. They also derive the formal Gaussian integral
representation of the partition function and the Feynman expansions for Fbg .
To summarize, one can see the common features of the above works on holo-
morphic anomaly equations. One starts with a sequence of functions Fg (t) (g ≥ 0)
for finitely many variables t1 , . . . , tn , and seek for another sequence Fbg (t, t̄) such
that limt̄→∞ Fbg (t, t̄) = Fg (t), and such that the sequence Fbg (t, t̄) satisfies qua-
dratic recursion equations of the form (3). Furthermore, the partition function
b t̄) = exp P
Z(t, 2g−2 b
g≥0 λ Fg (t, t̄) can be represented as a formal Gaussian integral,
and there are Feynman rules that express Fbg as a polynomial in the propagators
and the derives of Fr for r ≤ g. As pointed out in [3], the holomorphic anomaly has
contributions from the degenerate Riemann surfaces, from lower boundary strata
of the Deligne-Mumford moduli spaces. It turns out that the Feynman graphs that
appear in the Feynman expansion of Fbg in the above works are just the stable
graphs that index the stratification of the Deligne-Mumford moduli spaces Mg .
In the present work, we will formulate a general construction of some kind of
abstract quantum field theory that have the above works on holomorphic anomaly
equations as special realizations. The key ingredient in our construction is the
diagrammatics for the stratification of the Deligne-Mumford moduli spaces Mg,n ,
using the language of the dual graphs. There are some natural operators acting
on these graphs, including an edge-cutting operator K. We define an abstract free
energy
X 1
Fbg = Γ,
c
| Aut(Γ)|
Γ∈Gg,0

and derive a quadratic relation


g−1
X
1
K Fbg = (D∂ Fbg−1 + ∂ Fbr ∂ Fbg−r )
2 r=1

for g ≥ 2 using the operators ∂ and D which correspond to adding external edges
on the graphs. For details, see Section 2. For generalizations that involves more
labelling on the graphs, see Section 3.
4 ZHIYUAN WANG AND JIAN ZHOU

Another ingredient in our formalism is what we call the realizations of the ab-
stract quantum field theory, we need a sequence of functions Fg (t), t = (t1 , . . . , tn ),
g ≥ 0, and a symmetric nondegenerate matrix κ that depends on t and t̄. They will
be used to give the Feynman rules that associate a weight ωΓ to each Γ. This kind
of realization of the above abstract quantum field theory can also be represented
as a formal Gaussian integral. We will refer to this as the formal Gaussian inte-
gral representation of the abstract quantum field theory. These will be presented in
Section 4. P 2g−2
Using this formalism, given a holomorphic function F (t) = λ Fg (t), we
will obtain a family of free energies Fbg (κ, t). Here Fbg (κ, t) is a polynomial in κ of
degree 3g − 3, whose coefficients are some differential polynomials in {Fk }0≤k≤g for
g ≥ 2. In particular, Fbg (0, t) = Fg (t) for g ≥ 2.
Let us now mention some applications of our formalism. First of all, for the choice
κ = (τ̄ − τ )−1 , we recover the holomorphic anomaly for matrix models [13] [11], or
more generally, when the propagator κIJ takes the form of ∆IJ in (4), we recover
the construction of holomorphic anomaly equation in Gromov-Witten theory as
discussed in [15]. For details, see Section 5. Secondly, the propagators κ can be also
chosen to be holomorphic. We will use this formalism for holomorphic propagators
to reinterpret some results of the second author [27] in the theory of topological
1D gravity. And as applications, we will reinterpret some well-known examples
of quantum spectral curves [16] in our formalism. We will deal with the case of
propagators of the form (4) in a subsequent work that generalizes the formalism of
this work [22].
The rest of this paper is arranged as follows. In Section 2 we formulate the
diagrammatics and derive a quadratic recursive relation for the free energies using
the edge-cutting operators. In Section 3 we generalize the diagrammatics to the
case of labelled graphs. We give a field theory realization of the diagrammatics in
Section 4, where the free energies can be obtained from formal Gaussian integrals
and Feynman rules. The holomorphic anomaly equation is derived in Section 5
in our formalism using non-holomorphic propagators. In Section 6 we apply our
formalism to topological 1D gravity and derive quadratic recursion relations for
this theory. As applications, we relate the recursion relations to quantum spectral
curves in some important examples. Other applications include recursion relations
in enumeration problems of stable graphs and general graphs.

2. Diagrammatics of Moduli Space of Curves


In this section we recall the diagrammatics related to the stratification of the
Deligne-Mumford moduli spaces Mg,n . We define the edge-cutting operators and
some other operators acting on the graphs, and use them to derive a recursive
relation of the abstract free energy.

2.1. Moduli space of stable curves and Feynman diagrams. We first recall
the stratification of the Deligne-Mumford moduli space Mg,n of stable curves [6,17].
The Deligne-Mumford moduli space Mg,n is a smooth orbifold of complex di-
mension 3g − 3 + n. There are two types of basic maps over these spaces, namely
the forgetful map
πn+1 : Mg,n+1 → Mg,n ,
HOLOMORPHIC ANOMALY EQUATIONS AND QUANTUM SPECTRAL CURVES 5

which simply forgets the (n + 1)-th marked point and contract the unstable com-
ponent; and the gluing maps
ξ1 : Mg1 ,n1 +1 × Mg2 ,n2 +1 → Mg1 +g2 ,n1 +n2 ,

ξ2 : Mg−1,n+2 → Mg,n ,
which glues two marked points together to produce a new nodal point. The union
of images of ξ1 and ξ2 is the boundary Mg,n − Mg,n ; moreover, it is of codimension
one in Mg,n . In this way, we get a stratification of the compactified moduli space
Mg,n by decomposing it to the union of some products of the moduli spaces Mh,m
of smooth stable curves.
An efficient way of describing this stratification is to use the language of graphs.
Recall that for a stable curve (C; x1 , · · · , xn ) of genus g with n marked points, its
dual graph is defined as follows. Let C̃ be its normalization, then we associate a
vertex to each of the connected components of C̃, and write down the genus of
this component at this vertex. For each nodal point of C, we draw an internal
edge connecting the corresponding vertices; and for each marked point, we draw
an external edge attaching to the corresponding vertex, and write the markings
i ∈ {1, 2, · · · , n} besides the external edges.
In what follows we will forget all the markings of the external edges, which
corresponds to the moduli space Mg,n /Sn parametrizing equivalent classes of stable
curves with n marked points while we do not distinguish these marked points. In
this way, we obtain a graph of genus g with n external edges, where the genus is
defined to be the sum of the genus of the underlying one-dimensional manifold and
all the numbers associated to the vertices, which is exactly the genus of the stable
curve C.
Example 2.1. All possible dual graphs of curves in M1,1 , and M2,0 are listed in
Figure 1 and Figure 2 respectively.

1 0

Figure 1. Dual graphs for M1,1

2 1 1 1 0 1 0 0 0 0 0

Figure 2. Dual graphs for M2,0

A graph Γ is called stable if all its vertices are stable, that is, the valence of each
vertex of genus 0 is at least three, and the valence of each vertex of genus 1 is at
least one. It is clear that the stability of curves is equivalent to the stability of its
dual graph. The stable graphs will be the Feynman graphs used in this paper.
Let Gg,n be the set of all stable graphs of genus g with n external edges (not
c
necessarily connected), and Gg,n be the subset of connected stable graphs.
Now let Γ be a stable graph of genus g with n external edges, and define MΓ to
be the subset of Mg,n /Sn consisting of all equivalent classes of curves whose dual
6 ZHIYUAN WANG AND JIAN ZHOU

graph is Γ, then
G
(5) Mg,n /Sn = MΓ
c
Γ∈Gg,n

gives the stratification of Mg,n /Sn .


2.2. Operators on vector space associated to the set of stable graphs.
Denote by V c the vector space over Q generated by all connected stable graphs, i.e.,
M
(6) Vc = QΓ;
c
Γ∈Gg,n
2g−2+n>0

similarly, denote by V the vector space over Q generated by all stable graphs, not
necessarily connected, i.e.,
M
(7) V= QΓ.
Γ∈Gg,n
2g−2+n>0

In this subsection we define some operators on V. These operators may be


understood as the inverse procedures of the gluing maps and forgetful maps.
Define an operator K acting on V:
X
(8) K : V → V, Γ ∈ Gg,n 7→ K(Γ) = Γ′ ,
Γ′

where the summation is over all graphs Γ obtained by cutting an internal edge of
Γ. We will refer to this operator as the edge cutting operator. It is easy to see that
the operator K does not affect the stability of the graph. It decreases the number
of internal edges by one, and increases the number of external edges by two. In
the picture of stable curves, this operator corresponds to breaking up a node and
regard it as two new marked points. So the operator K can be regarded as the
inverse operator of the gluing maps ξ1 and ξ2 .
Example 2.2.
K 2 =0,

K 0 0 =2 0 0 + 0 0 ,
K 0 0 =3 0 0 .
Let us define another operator ∂. This operator has two parts, one is to attach
an external edge to a vertex and sum over all vertices, and the other is to break up
an internal edge and insert a stable vertex of genus 0 with three edges.
Example 2.3.
∂ 1 = 1 ,

∂ 0 0 =2 0 0 +2 0 0 0 + 0 0 0 ,

0
∂ 0 1 = 0 1 + 0 1 +3
0 1.
HOLOMORPHIC ANOMALY EQUATIONS AND QUANTUM SPECTRAL CURVES 7

In the dual picture, the operator ∂ is just adding one marked point, or breaking
up a node of the curve and gluing a 3-pointed sphere.
Now we define a third operator γ, which acts on Γ by simply attach a stable
vertex of genus 0 with three edges to an external edge of Γ, and sum over all external
edges.
Example 2.4.
γ 1 = 1 0 ,

γ 0 =2 0 0 .

If the graph Γ has no external edge, then we may simply have γ(Γ) = 0. This
operator is to glue a 3-pointed sphere to the curve along marked points to get a
new node.
Clearly the operators ∂ and γ also preserves the stability, and they both increase
the number of external edges by one. Intuitively, ∂ + γ is the inverse of the forgetful
map. Write D = ∂ + γ. Then D preserves the subspaces V c .
2.3. Abstract free energy. Let us now define an abstract quantum field theory
based on the diagrammatics of stable graphs discussed above.
Definition 2.1. For g ≥ 2, we define the abstract free energy of genus g to be
X 1
(9) Fbg = Γ.
c
| Aut(Γ)|
Γ∈Gg,0

In general, for 2g − 2 + n > 0, define


X 1
(10) Fbg,n = Γ,
c
| Aut(Γ)|
Γ∈Gg,n

then the free energy Fbg is just Fbg,0 .


Example 2.5. Here we list some explicit expressions of Fbg,n for small (g, n).

Fb0,3 = 1
6 0
,

Fb0,4 = 1
24 0 + 18 0 0
,

Fb0,5 = 1
120 0 + 81 0 0 0 1
+ 12 0 0
,

Fb1,1 = 1 + 21 0
,

Fb1,2 = 1
2 1 + 14 0 + 12 1 0 + 41 0 0 + 14 0 0
,
8 ZHIYUAN WANG AND JIAN ZHOU

Fb1,3 = 1
6 1 1
+ 12 0 + 21 1 0 + 61 1 0 + 41 0 0

+ 41 0 0 1
+ 12 0 0 + 21 1 0 0 + 41 0 0 0

0
+ 41 0 0 0 + 61 0 0 ,

Fb2,0 = 2 + 12 1 + 21 1 1 + 18 0 + 12 1 0 + 81 0 0 1
+ 12 0 0
,

Fb2,1 = 2 + 12 1 + 1 1 + 81 0 + 12 1 0 + 21 1 0

+ 12 1 0 + 21 1 0 1 + 41 0 0 + 14 0 0

+ 61 0 0 + 21 0 0 1 + 21 0 0 1 + 41 0 0 0

0
+ 81 0 0 0 + 41 0 0.

More examples will be given in Appendix A.


2.4. A recursion relation for the abstract free energy.
Example 2.6. Using the expressions in the above example, one can directly check
the following identities:
9 b 2
K Fb0,4 = (F0,3 ) ,
2
K Fb0,5 = 12Fb0,3 Fb0,4 ,
K Fb1,1 = 3Fb0,3 ,
K Fb1,2 = 6Fb0,4 + 3Fb1,1 Fb0,3 ,
K Fb1,3 = 10Fb0,5 + 4Fb1,1 Fb0,4 + 6Fb1,2 Fb0,3 ,
1
K Fb2 = Fb1,2 + (Fb1,1 )2 ,
2
K Fb2,1 = 3Fb1,3 + 2Fb1,1 Fb1,2 .
In general, we have the following relation.
Theorem 2.1. For 2g − 2 + n > 0, we have
  X
n+2 b 1
(11) K Fbg,n = Fg−1,n+2 + (n1 Fbg1 ,n1 )(n2 Fbg2 ,n2 ),
2 2 g +g =g,n +n
1 2 1 2 =n+2,
n1 ≥1,n2 ≥1

where the sum on the right hand side is taken over all stable cases.
HOLOMORPHIC ANOMALY EQUATIONS AND QUANTUM SPECTRAL CURVES 9

c
Proof. First note that cutting off an internal edge of a stable graph in Gg,n gives
us two types of new stable graphs, one type is still connected, and must be of
genus g − 1 with n + 2 external edges; the other type of graphs have two connected
components, and have total genus g with n + 2 external edges. Thus we write

(12) K Fbg,n = Ig,n + Jg,n ,

where Ig,n and Jg,n are sums over connected graphs and disconnected graphs re-
spectively.
c
We already know that graphs appearing in Ig,n all belong to Gg−1,n+2 ; conversely,
c b
every graph in Gg−1,n+2 may be transformed to a graph appearing in Fg,n once we
glue two external edges together. This tells us that the graphs appearing in Ig,n are
the exactly the same graphs in Fbg−1,n+2 . Now we check the coefficients of every
c
graph in the two sums coincide. For a stable graph Γ ∈ Gg−1,n+2 , there are n+2
2
ways to choose two external edges to be glued together. Denote by Γ e one of the
equivalence classes of graphs obtained from Γ by gluing two external edges. Then
we need to show that
X   
′ 1 e n+2 1
(13) K ·Γ = · Γ,
e
| Aut(Γ)| 2 | Aut(Γ)|
e
Γ∈G Γ

where GΓ is the set of equivalence classes of stable graphs Γe such that one can
e ′ e e e is the number
obtain Γ by cutting an edge of Γ, and K (Γ) = N (Γ) · Γ, where N (Γ)
e
of ways to cut an edge of Γ to get Γ. It suffices to show that
X  
(14) e · | Aut(Γ)| = n + 2 .
N (Γ)
e
| Aut(Γ)| 2
e Γ
Γ∈G

Consider
 the set of the choices of picking a pair of external edges of Γ. This set has
n+2
2 elements. We partition this set into disjoint union of HΓe consisting of those
choices for which one can get Γ e by gluing the pair of external edges, where Γe runs
over the set GΓ of all possibilities. Therefore,
X  
n+2
(15) |HΓe | = .
2
e Γ
Γ∈G

Now we show that

(16) e · | Aut(Γ)| ,
|HΓe | = N (Γ)
e
| Aut(Γ)|
or equivalently,

|HΓe | e
N (Γ)
(17) = .
| Aut(Γ)| e
| Aut(Γ)|

Let Γ′′ be the graph obtained from Γ e by changing the internal edge that we
cut to get Γ into a dotted edge, and let Γ′ be the graph with two dotted external
edges obtained from Γ′′ by cutting the dotted edge. For examples, see Figure 3 and
Figure 4.
10 ZHIYUAN WANG AND JIAN ZHOU

1 1 1 1
v v v v
1 0 1 1 0 1 1 0 1 1 0 1
1 1 1 1

Γ Γ′ Γ′′ e
Γ

Figure 3. An example where the dotted edge is a loop.

v1 v2 v1 v2 v1 v2 v1 v2
0 0 0 0 0 0 0 0
1 1 1 1 1 1 1 1
Γ Γ ′
Γ ′′ e
Γ

Figure 4. An example where the dotted edge is not a loop.

When we talk about automorphisms of Γ′ and Γ′′ , the dotted edges can not be
mapped to solid edges. Now we claim:
|HΓe | 1 1 e
N (Γ)
(18) = = = .
| Aut(Γ)| ′
| Aut(Γ )| ′′
| Aut(Γ )| e
| Aut(Γ)|
The second equality is trivial. We only need to consider the first and third
equalities. Recall that |HΓe | is the number of ways to choose two external edges
of Γ to be glued together to get Γ,e thus |He | equals to the number of all possible
Γ
ways to change a pair of external edges of Γ to dotted edges to obtain Γ′ . Similarly,
e equals to the number of all possible ways to change an internal edge of Γ
N (Γ) e to
a dotted edge to obtain Γ′′ .
In the case where the dotted edge is a loop in Γ′′ , denote by v the vertex where
it is incident to. Then let Autv (Γ) be the group of automorphisms of Γ that fix the
vertex v (e.g., for Figure 3, Autv (Γ) ∼
= (S3 )3 × (S2 )2 ). Let Ov be the orbit of v in
Γ under Aut(Γ), then
| Aut(Γ)|
(19) = |Ov |.
| Autv (Γ)|
It is clear that any automorphism of Γ′ fixes the vertex v since there is no other
vertex with dotted half edges, and so Aut(Γ′ ) is a subgroup of Autv (Γ). To compute
the index of this subgroup, it suffices to consider the actions of these automorphisms
groups on the external half edges incident at v. Assume there are l external edges
attaching to v in Γ′ including two dotted edges, then we have
| Autv (Γ)| l! 1
(20) ′
= · ,
| Aut(Γ )| (l − 2)! 2
here the factor 21 comes from the symmetry between two dotted edges. Now (19)
and (20) give us
 
| Aut(Γ)| l
(21) = |Ov | · .
| Aut(Γ′ )| 2
HOLOMORPHIC ANOMALY EQUATIONS AND QUANTUM SPECTRAL CURVES 11

l

Noticing that |Ov | equals to the number of ways to choose the vertex v, and 2 is
the number of ways to choose two external edges on v, thus
 
l
|Ov | · = |HΓe |
2
and this proves the first equality in (18).
Using the same argument, we now consider Autv (Γ), e i.e., the group of automor-
phisms of Γ e ∼
e that fix v (e.g., for Figure 3 Autv (Γ) ev be the
= (S3 )2 × (S2 )3 ). Let O
e e
orbit of v in Γ under Aut(Γ), then
e
| Aut(Γ)|
(22) ev |.
= |O
e
| Autv (Γ)|
Assume there are k loops attaching to v in Γ′′ including a dotted one, then
e
| Autv (Γ)| k! · 2k 1
(23) ′′
= k−1
· = k.
| Aut(Γ )| (k − 1)! · 2 2
Thus (22) and (23) give us
e
| Aut(Γ)| ev |,
= k · |O
| Aut(Γ′′ )|
e Therefore the third equality in (18) also
and the right hand side equals to N (Γ).
holds.
The case where the dotted edge in Γ′′ is not a loop can be proved similarly. Let
v = (v1 , v2 ) be the ordered pair of end points of the dotted edge. As above, we
consider Autv (Γ) (here the order of (v1 , v2 ) should be preserved, e.g., for Figure
4, Autv (Γ) ∼ = S2 ), and the orbit Ov of the ordered pair v in V (Γ) × V (Γ) under
Aut(Γ). We still have the equality:
| Aut(Γ)|
(24) = |Ov |.
| Autv (Γ)|
Assume there are li external edges attached to vi in Γ′ , then
| Autv (Γ)| l1 ! l2 !
(25) ′
= · · δ = l1 l2 · δ,
| Aut(Γ )| (l1 − 1)! (l2 − 1)!
1
here δ = 2 if (v2 , v1 ) ∈ Ov , and δ = 1 if (v2 , v1 ) 6∈ Ov . The equations (24) and (25)
give us
| Aut(Γ)|
(26) = l1 l2 |Ov | · δ,
| Aut(Γ′ )|
which equals |HΓe |.
e to be the group of automorphisms of Γ
Similarly define Autv (Γ) e that fix v (e.g.,
e ∼ e e e under Aut(Γ),
for Figure 4, Aut(Γ) = S3 ). Let Ov be the orbit of v in V (Γ) × V (Γ) e
then
e
| Aut(Γ)|
(27) ev |.
= |O
e
| Autv (Γ)|
12 ZHIYUAN WANG AND JIAN ZHOU

Assume there are k internal edges connecting v1 and v2 in Γ′′ including a dotted
one, then
e
| Autv (Γ)| k!
(28) = · δ = k · δ.
| Aut(Γ′′ )| (k − 1)!
Then the equations (27) and (28) imply
e
| Aut(Γ)| ev | · δ,
= k|O
| Aut(Γ′′ )|
e Therefore the first and third equalities in (18) also hold in
which equals to N (Γ).
this case.
In summary, we’ve proved the identity
 
n+2 b
(29) Ig,n = Fg−1,n+2 .
2
The relation
1 X
(30) Jg,n = (n1 Fbg1 ,n1 )(n2 Fbg2 ,n2 )
2g
1 +g2 =g,n1 +n2 =n+2,
n1 ≥1,n2 ≥1

can be established in a similar fashion. For two graphs Γ1 ∈ Ggc1 ,n1 , Γ2 ∈ Ggc2 ,n2 ,
denote by Γe one of the equivalence classes of graphs obtained from Γ1 and Γ2 by
gluing two external edges, one from each graph. Then we need to show that
X 
1 e = n1 n2
(31) K ′′ ·Γ Γ1 · Γ2 ,
e
| Aut(Γ)| | Aut(Γ1 )| | Aut(Γ2 )|
e
Γ∈G ′′
Γ

e such that one can


where GΓ′′ is the set of equivalence classes of stable graphs Γ
obtain Γ1 and Γ2 by cutting an edge of Γ,e and
e = N ′′ (Γ)
K ′′ (Γ) e · Γ1 Γ2 ,

e is the number of ways to cut an edge of Γ


where N ′′ (Γ) e to get Γ1 and Γ2 . It suffices
to show that
X 1 n1 n2
(32) e ·
N ′′ (Γ) = · .
e
| Aut(Γ)| | Aut(Γ 1 )| | Aut(Γ 2 )|
e
Γ∈G
′′
Γ

Consider the set of the choices of picking one external edge from each of Γ1 and
Γ2 . This set has n1 n2 elements. We partition this set into disjoint union of HΓe′′
consisting of those choices for which one can get Γ e by gluing the pair of external
edges, where Γe runs over the set G ′′ of all possibilities. Therefore,
Γ
X
(33) |HΓe′′ | = n1 n2 .
e
Γ∈G ′′
Γ

Now we show that


e · | Aut(Γ1 )| · | Aut(Γ2 )|
(34) |HΓe′′ | = N ′′ (Γ) ,
| Aut(Γ)|e
HOLOMORPHIC ANOMALY EQUATIONS AND QUANTUM SPECTRAL CURVES 13

or equivalently,
|HΓe′′ | e
N ′′ (Γ)
(35) = .
| Aut(Γ1 )| · | Aut(Γ2 )| e
| Aut(Γ)|
e by changing the internal edge that we cut
Let Γ′′ be the graph obtained from Γ
to get Γ into a dotted edge, and Γ be the disconnected graph obtained from Γ′′ by

cutting the dotted edge. Then the equality


|HΓe′′ | 1 1 e
N ′′ (Γ)
(36) = = =
| Aut(Γ1 )| · | Aut(Γ2 )| ′
| Aut(Γ )| ′′
| Aut(Γ )| e
| Aut(Γ)|
can be proved using exactly the same method as the proof of (18). Suppose that
the two external edges to be glued together are incident at v1 in Γ1 and v2 in Γ2
respectively. Denote by v = (v1 , v2 ) the ordered pair of these two vertices. Then the
e and the orbits Ovi , O
subgroups Autvi (Γi ), Autv (Γ) ev are similarly defined. Then
the following equalities hold:
| Aut(Γi )| | Autv1 (Γ1 )| · | Autv1 (Γ1 )|
= |Ovi |, = δ ′′ ,
| Autvi (Γi )| | Aut(Γ′ )|
e
| Aut(Γ)| e
ev |, | Autv (Γ)| = δ ′′ ,
= |O
e
| Autv (Γ)| | Aut(Γ′′ )|

where δ ′′ = 12 if Γ1 = Γ2 and v1 ∈ Ov2 under this identification, and δ ′′ = 1


otherwise. Furthermore, we have
|Ov1 | · |Ov2 | · δ ′′ = |HΓe′′ |, ev | · δ ′′ = N ′′ (Γ).
|O e

These proves 36, therefore the relation 30 holds.


The equalities (12), (29), (30) completes the proof of the theorem. 

Furthermore, the right hand side of (11) can be rewritten in terms of operators
D. To explain this we need the following observation.

Lemma 2.1. For 2g − 2 + n > 0, we have DFbg,n = (n + 1)Fbg,n+1 .


Proof. The proof is similar to the proof of Theorem 2.1. First the graphs appear-
ing in the two sums are the same, since we’ve already taken all possible unstable
contractions into consideration in the definition of the operator D. Thus we only
need to check the coefficients.
c c
Let Γ ∈ Gg,n+1 be a stable graph, and GΓ be the subset of Gg,n consisting of
graphs obtained from Γ by removing an external edge and stabilizing. We only
need to show
X 1 1
D′ ( e = (n + 1) ·
Γ) Γ,
e
| Aut(Γ)| | Aut(Γ)|
e Γ
Γ∈G

e · Γ with N (Γ)
where D′ (Γ) = N (Γ) e the number of ways to obtain Γ from Γ̃ via one
of the operations in D = ∂ + γ. Then it suffices to show
X
(37) e · | Aut(Γ)| = n + 1.
N (Γ)
e
| Aut(Γ)|
e Γ
Γ∈G
14 ZHIYUAN WANG AND JIAN ZHOU

Consider the set of the choices of picking an external edge of Γ, and partition it
e after removing
to subsets HΓe consisting of those choices for which one can obtain Γ
this edge and stabilizing. Clearly
X  
n+1
|HΓe | = = n + 1.
1
e
Γ∈G Γ

Now we show that


e · | Aut(Γ)|
N (Γ) = |HΓe |,
e
| Aut(Γ)|
or equivalently,
|HΓe | e
N (Γ)
= .
| Aut(Γ)| e
| Aut(Γ)|
Denote by Γ′ the graph obtained from Γ by changing the external edge that we
remove to a dotted edge. We will show that
|HΓe | 1 e
N (Γ)
(38) = = .
| Aut(Γ)| ′
| Aut(Γ )| e
| Aut(Γ)|
Let v be the vertex of Γ and Γ′ that the dotted edge attaches to. The first equality
follows from
| Aut(Γ)| | Autv (Γ)| k!
= |Ov |, = ;
| Autv (Γ)| | Aut(Γ′ )| (k − 1)!
 
k
|HΓe | = · |Ov |,
1
where Autv (Γ) is the set of automorphisms of Γ that preserve v, Ov is the orbit of
v in Γ under Aut(Γ), and k is the number of external edges attaching to v in Γ.
e that is, ṽ = v if the graph is still
Let ṽ be the corresponding vertex of v in Γ,
stable after removing the dotted edge, and ṽ the vertex that v contracts to if the
graph is unstable. The second equality in (38) follows from
e
| Aut(Γ)| eṽ |,
= |O e = | Aut(Γ′ )|;
| Autṽ (Γ)|
e
| Autṽ (Γ)|

e = |O
N (Γ) eṽ |,
e is the set of automorphisms of Γ
where Autṽ (Γ) e that preserve ṽ, and O
eṽ is the orbit
e under Aut(Γ).
of ṽ in Γ e
Therefore the equality (38) holds, which proves the lemma. 

Even though so far we have not defined Fb1 , Fb0,2 , and Fb0,1 , it will be convenient
for later use to define DFb1 , DFb0,2 , DDFb0,1 by formally applying Lemma 2.1 as
follows:
DFb1 := Fb1,1 ,
(39) DFb0,2 := 3Fb0,3 ,
DDFb0,1 := 6Fb0,3 .
HOLOMORPHIC ANOMALY EQUATIONS AND QUANTUM SPECTRAL CURVES 15

Now using Lemma 2.1 and conventions in (39), our Theorem 2.1 can be rewritten
in the following way.

Theorem 2.2. For 2g − 2 + n > 0, we have


1 X
(40) K Fbg,n = (DDFbg−1,n + DFbg1 ,n1 DFbg2 ,n2 ).
2 g +g =g,
1 2
n1 +n2 =n

In particular, by taking n = 0 we get a recursion relation for the abstract free


energy:
g−1
X
b 1 b
(41) K Fg = (D∂ Fg−1 + ∂ Fbr ∂ Fbg−r ).
2 r=1

3. A Generalization to Labelled Graphs


In this section, we generalize the construction of Section 2 to the case where all
the half-edges of graphs may have labels.

3.1. Diagrammatics for labelled graphs. Fix a positive integer N , for a stable
graph Γ ∈ Gg,n we label some indices on the edges as follows. For an external edge,
we label an index in {1, 2, · · · , N }; and for an internal edge, we label an index in
{1, 2, · · · , N } at each of its end points. In other words, we label each half edge
incident at a vertex.
As generalizations of the edge-cutting operator K defined in Section 2.2, we
define the edge-cutting operators Kij as the operator that cuts off an internal
edges with two labels i and j, then sum over all such internal edges.

Example 3.1. Take N = 2, then we have the following examples:

1 1 2 1 1 1 2 1
K11 0 0 =2 0 0
1 1 1 1 ,
1 1 2 1 1 1 2 1
K12 0 0 = 0 0
1 1 1 1 ,
1 1 2 1
K22 0 0 =0
1 1 .

Similarly, we also have operators ∂i and γi generalizing the operators ∂ and γin
Section 2.2 respectively. The operator ∂i has two parts, one is to add an external
edge labelled by i, and the other is to break up an internal edge and attach a 3-
pointed Riemann sphere such that the new external edge has label i. The operator
γi is to attach a 3-pointed Riemann sphere to an external edge e, and move the
label of e to be the label of one of the new external edge, while the other new
external edge has label i. The two labels of the new internal edge can be chosen
arbitrarily.
16 ZHIYUAN WANG AND JIAN ZHOU

Example 3.2. For N = 2, here are some examples of the operator ∂i .


2 2
∂1 1 = 1 1,
2 1 2
∂1 0 = 0 1 + 1 01 20 1 + 1 02 20 1 + 2 1 01 20 1
1 1 2 2 2 2 2
2 2 .
Example 3.3. For N = 2, here are some examples of the operator γi .

γ1 2 1 1 2 + 1 2 2 + 2 1 2 + 2 2 2
1 = 1 0 1 0 1 0 1 0
1 1 1 1,
2 1 2 11 2 12 2 21 2 22
γ1 0 = 0 0 1 + 0 0 1 + 0 0 1 + 0 0 1
1 1 1 1
2 2 2 2 2 .
Now let Gg,n (N ) be the set of all labelled stable graphs of genus g with n external
c
edges and N possible indices, and Gg,n (N ) be the set of connected ones. In the
labelled case we also define the abstract free energy in the same way.
Definition 3.1. For g ≥ 2, the abstract free energy Fbg is defined by:
X 1
(42) Fbg = Γ.
c
| Aut(Γ)|
Γ∈Gg,0 (N )

For example, take N = 2, then we have


1 1 1 1 2 2 1 2
Fb2 = 2 + 2 1 1 + 2 1 1 + 1 1

1
1 1
2 1
+ 2 1 + 2 1 + 1 +···
1 2 2
The complete expressions for Fb2 can be obtained by adding all possible labels on
the graphs of Fb2 for the N = 1 case and modifying the number of automorphisms.
3.2. A recursion relation for the abstract free energy. In the labelled case,
we also have a recursion relation generalizing (41). Here we define Di = ∂i + γi .
Theorem 3.1. For g ≥ 2, we have
g−1
X
(43) Kij Fbg = Di ∂j Fbg−1 + ∂i Fbr ∂j Fbg−r
r=1
for i 6= j, and
g−1
X
1
(44) Kii Fbg = (Di ∂i Fbg−1 + ∂i Fbr ∂i Fbg−r ).
2 r=1

The proof is an easy modification of the proof of Theorem 2.2 and is omitted.
For the case of g = 2 we use the following convention
P j P k j P k j
1
∂j Fb1 =: j( 1 + k 2 0 + k6=l 0 )
(45) k l
HOLOMORPHIC ANOMALY EQUATIONS AND QUANTUM SPECTRAL CURVES 17

4. Realization of the Abstract Quantum Field Theory


In this section we formulate a Feynman rule for the diagrammatics discussed in
the previous sections, in such way we are able to define a new free energy Fb from a
given holomorphic free energy F and a choice of propagator κ. We also represent
this new field theory as formal Gaussian integrals. This construction realizes the
abstract free energy and its quadratic recursion relations discussed in preceding
sections.
4.1. Realization of the abstract QFT by Feynman rules. In this subsection
we will present a construction to realize the abstract quantum field theory defined
in Section 2.3 and Section 3.1 by Feynman rules that assign the contribution of a
vertex and the contribution of edge:
Y Y
(46) Γ 7→ ωΓ = ωv · ωe .
v∈V (Γ) e∈E(Γ)

For ωv , we need to fix a sequence of holomorphic functions Fg in t1 , . . . , tN as input;


for ωe , we need fix a nondegenerate symmetric matrix κ = (κij )1≤i,j≤N of functions
in t1 , . . . , tN as propagators.
For a given positive integer N , let t1 ,t2 ,· · · ,tN be the coordinates on an N -
dimensional vector space, and κ = (κij ) be a nondegenerate symmetric matrix of
size N × N . Moreover, we fix a sequence of holomorphic function Fg (t1 , · · · , tN )
for g ≥ 0. Let

X
(47) F (t; λ) = λ2g−2 Fg (t).
g=0
c
Now for a labelled stable graph Γ ∈ Gg,0 (N ), we associate a polynomial ωΓ as
ext
follows. Let V (Γ), E (Γ), and E(Γ) be the sets of vertices, external half edges,
and internal edges of Γ respectively. For each e ∈ E ext (Γ), write l(e) as the label
on the external e; and for each e ∈ E(Γ), write l1 (e) and l2 (e) as the two labels
on e (the order is not important). For each v ∈ V (Γ), write g(v) as the genus
associated
P to v, and vali (v) as the number of half-edges labelled by i incident at v,
then i vali (v) is the valence of v. Then ωΓ is defined by
Y (val (v),··· ,valN (v))
Y
(48) ωΓ = Fg(v)1 (t) κl1 (e)l2 (e) ,
v∈V (Γ) e∈E(Γ)
(val (v),··· ,valN (v))
where Fg(v)1 (t) := ( ∂t∂1 )val1 (v) · · · ( ∂t∂N )valN (v) Fg(v) (t). For N = 1, this
expression simplifies to
Y (val(v))
(49) ωΓ = ( Fg(v) (t))κ|E(Γ)| .
v∈V (Γ)

Definition 4.1. For g ≥ 2 we define the free energy to be


X 1
(50) Fbg = ωΓ .
c
| Aut(Γ)|
Γ∈Gg,0 (N )

In particular, the degree of κ in the expression of Fbg is 3g − 3.


Example 4.1. For N = 1, we have
1 1 1 (4) 1 5
(51) Fb2 = F2 + κ[ F1′′ + (F1′ )2 ] + κ2 ( F0 + F1′ F0′′′ ) + κ3 (F0′′′ )2 .
2 2 8 2 24
18 ZHIYUAN WANG AND JIAN ZHOU

And for N = 2, we have


1 (2,0) 1 (1,0) 2 1 (0,2) 1 (0,1) 2
Fb2 = F2 + ( F1 + (F1 ) )κ11 + ( F1 + (F1 ) )κ22
2 2 2 2
1 (4,0) 1 (1,0) (3,0) 2 1 (0,4) 1 (0,1) (0,3) 2
+ ( F0 + F1 F0 )κ11 + ( F0 + F1 F0 )κ22
8 2 8 2
(1,1) (1,0) (0,1) 1 (2,2) (1,0) (1,2) (0,1) (2,1)
+ (F1 + F1 F1 )κ12 + ( F0 + F1 F0 + F1 F0 )κ212
2
1 (2,2) 1 (1,0) (1,2) 1 (0,1) (2,1)
+ ( F0 + F1 F0 + F1 F0 )κ11 κ22
4 2 2
1 (3,1) 3 (1,0) (2,1) 1 (0,1) (3,0)
+ ( F0 + F1 F0 + F1 F0 )κ11 κ12
2 2 2
1 (1,3) 3 (0,1) (1,2) 1 (1,0) (0,3)
+ ( F0 + F1 F0 + F1 F0 )κ22 κ12
(52) 2 2 2
5 (3,0) 2 3 5 (0,3) 2 3 1 (3,0) (0,3) 3 (2,1) (1,2) 3
+ (F0 ) κ11 + + (F0 ) κ22 + ( F0 F0 + F0 F0 )κ12
24 24 6 2
3 (2,1) 2 1 (3,0) (1,2) 2
+ ( (F0 ) + F0 F0 )κ11 κ22
8 4
3 (1,2) 2 1 (0,3) (2,1) 2
+ ( (F0 ) + F0 F0 )κ22 κ11
8 4
3 (2,1) 2 (3,0) (1,2) 3 (1,2) (0,3) (2,1)
+ ( (F0 ) + F0 F0 )κ11 κ212 + ( (F0 )2 + F0 F0 )κ22 κ212
2 2
5 (3,0) (2,1) 5 (0,3) (1,2)
+ F0 F0 κ211 κ12 + F0 F0 κ222 κ12
4 4
1 (3,0) (0,3) 9 (2,1) (1,2)
+ ( F0 F0 + F0 F0 )κ11 κ22 κ12 .
4 4
4.2. Representation by formal Gaussian integrals. In fact, the free energy
Fbg defined in last subsection has a representation using formal Gaussian integrals.
Consider the following partition function
Z  
−2 ∂F0 (t) 1 ∂ 2 F0 (t)
Z(t, κ) = exp F (η) − λ (η − t) + (η − t)T (η − t)
∂t 2 ∂t2
(53) 
1
+ (η − t)T κ−1 (η − t) dη,
2

where the integral is over η ∈ RN . Now do the Taylor expansion of F (η), we get
Z  X YN
λ2g−2 (l(1) ,··· ,l(N ) ) (i)
Z(t, κ) =Z(t) exp QN (i) Fg (t) (ηi − ti )l
i=1 l !
P
2g−2+ l(i) >0 j=1
−2

λ
− (η − t)T κ−1 (η − t) dη
2
(54) P k
X X 1 λ2 k gk −2k Y (l
(1) (N )
,··· ,lj )
=Z(t) Q Q (i)
Fgj j ×
k!
k≥0 2gj −2+P l(i) >0 1≤j≤k i lj ! j=1
i j

Z Y
N Pk (i)  λ−2 
(ηi − ti ) j=1 lj · exp − (η − t)T κ−1 (η − t) dη,
i=1
2
HOLOMORPHIC ANOMALY EQUATIONS AND QUANTUM SPECTRAL CURVES 19

where Z(t) = exp(F (t)) is the holomorphic partition function. Then evaluate the
Gaussian integral on the right-hand side of the last equality, we have
Z Y N Pk (i) λ−2
(ηi − ti ) j=1 lj · exp[− (η − t)T κ−1 (η − t)]dη
i=1
2
(55) l
(2π)n 1 1 X Y 2
=( ) · l
2 (λ κaσ(2j−1) aσ(2j) ),
det(λ−2 κ−1 ) 2 · l! j=1 σ∈S2l
P (i) P (i)
for i,j lj = 2l even, and equals zero if is odd, where the sequence
i,j lj
P (i)
(a1 , · · · , a2l ) consists of indices in {1, · · · , N } with the number of i equals j lj .
Let the normalized partition function be
(56) e κ) = Z(t, κ)/(2πλ2 ) n2 ,
Z(t,
then this relates to our free energy (50) in the following way.
Theorem 4.1. We have
X∞
(57) e κ) = exp(
Z(t, λ2g−2 Fbg (t)),
g=0

where Fb0 and Fb1 are defined by


1
(58) Fb0 = F0 , Fb1 = F1 + log(det(κ)).
2
p
e κ)/(Z(t) det(κ)) is
Proof. From (54) and (55), the coefficient of λn in Z(t,
(1) (N )
k (l ,··· ,lj ) l
X 1 1 Y Fgj j X Y
· l · κaσ(2j−1) aσ(2j) .
P k! 2 · l! j=1 l(1) ! · · · l(N ) !
2 gj −2k+2l=n j j σ∈S2l j=1

This corresponds to the summation of ωΓ over all possible ways to glue some sta-
(1) (N )
(l ,··· ,l )
ble vertices of types {Fgj j j
}1≤j≤l together to obtain a stable graph Γ (not
1
necessarily connected), with coefficients | Aut(Γ)| . Then since
l = |E(Γ)|, k = |V (Γ)|,
we have
k
X
g(Γ) = 1 + l − k + gj .
j=1

Thus
X λ2g−2
e κ) =
Z(t, ωΓ
| Aut(Γ)|
Γ∈Gg,0

and
X X ∞
e = λ2g−2
log(Z) ωΓ = λ2g−2 Fbg .
c
| Aut(Γ)| g=0
Γ∈Gg,0


20 ZHIYUAN WANG AND JIAN ZHOU

Example 4.2. For the case of N = 1, the partition function is

X∞
Z̃(t, κ) = exp( λ2g−2 Fbg (t))
g=0

(59) X X X 1 λ2(g1 +···+gk )−2k+2l


=Z(t) ×
k! l1 ! · · · lk !
k≥0 2gi −2+li >0 l

(2l − 1)!! · Fg(l11 ) · · · Fg(lkk ) · κl · δl1 +···+lk −2l,0 .

As mentioned in the Introduction, the formal Gaussian integral (53) appeared in


earlier work on holomorphic anomaly equations [11, 13, 15], where the propagator
κ was chosen to be the Zamolodchikov metric which is non-holomorphic. Later
in this paper we will see that the propagator κ admits different choices, which
give us different theories in mathematical physics. In particular, by taking suitable
holomorphic propagators we will see that partition functions for some well-known
quantum spectral curves can be constructed in our formalism.

4.3. Realization of the operators on stable graphs. Now we discuss the re-
alization of the operators K, ∂ and γ defined in Section 2.2.
The operator K is realized by the differential operator ∂κ . In fact, by (49),
 Y 
(val(v))
∂κ ωΓ = ∂κ ( Fg(v) (t))κ|E(Γ)|
v∈V (Γ)
Y (val(v))
= |E(Γ)| · ( Fg(v) (t))κ|E(Γ)|−1
v∈V (Γ)
X
= ωΓ′ = ωKΓ ,
Γ′

where the summation on the right-hand side of the second equality is over all graphs
Γ′ obtained from Γ by cutting an edge.
The operator γ is realized by the operator of multiplication by |E ext (Γ)| · κF0′′′
on ωΓ . In fact, the effect of γ is to change each external edge incident at v of a
stable graph Γ to an internal edge joining the vertex v to an additional trivalent
vertex of genus 0. Let Γ′ be one of stable graphs obtained from Γ in this way, then
by the Feynman rule,

(60) ωΓ′ = κF0′′′ · ωΓ .

Therefore,
X
(61) ωγΓ = ωΓ′ = |E ext (Γ)| · κF0′′′ · ωΓ .
Γ′

The operator ∂ is realized by the differential operator ∂t . In fact, ωΓ is a product


of contributions from the vertices and the contributions from the internal edges.
Hence ∂t ωΓ is a summation of the contributions of the actions on each vertices and
HOLOMORPHIC ANOMALY EQUATIONS AND QUANTUM SPECTRAL CURVES 21

the contributions on each internal edges. In fact by (49),


 Y 
(val(v))
∂t ωΓ = ∂t ( Fg(v) (t))κ|E(Γ)|
v∈V (Γ)
X (val(v)+1)
Y (val(v ′ ))
= Fg(v) (t) · Fg(v′ ) (t)) · κ|E(Γ)|
v∈V (Γ) v ′ ∈V (Γ)−{v}
Y (val(v))
+ Fg(v) (t)) · |E(Γ)|κ|E(Γ)|−1 · ∂t κ.
v∈V (Γ)

On the other hand,


X X
(62) ∂Γ = Γv + Γe ,
v∈V (Γ) e∈E(Γ)

where Γv is obtained from Γ by attaching an additional half-edge at a vertex v ∈


V (Γ), and Γe is obtained from Γ by changing an internal edge e ∈ E(Γ) to two
internal edges incident at an addition trivalent vertex w of genus zero. Therefore,
by the Feynman rule we get:
X X
ω∂Γ = ωΓv + ωΓe
v∈V (Γ) e∈E(Γ)
X (val(v)+1)
Y (val(v ′ ))
= Fg(v) (t) · Fg(v′ ) (t)) · κ|E(Γ)|
v∈V (Γ) v ′ ∈V (Γ)−{v}
X Y (val(v))
+ Fg(v) (t)) · κ|E(Γ)|+1 · F0′′′ (t).
e∈E(Γ) v∈V (Γ)

Therefore, in order to have ω∂Γ = ∂t ωΓ , we require the propagator κ to be a function


of t which satisfies
∂κ
(63) = κ2 · F0′′′ (t),
∂t
here κ needs not to be holomorphic. A general solution to this equation is
1
κ(t) = ,
C − F0′′ (t)
where C may be either a constant or an anti-holomorphic function of t.

4.4. Realization of the recursion relations. By applying Feynman rules to


the quadratic recursion relations in Theorem 2.2 and using the realizations of the
operators on graphs discussion in the preceding subsection, we obtain the recursion
relations for Fbg immediately.
Theorem 4.2. When the propagator κ satisfies the condition (63), the free energy
Fbg for g ≥ 2 satisfies the equations
g−1
X
1
(64) ∂κ Fbg = (Dt ∂t Fbg−1 + ∂t Fbr ∂t Fbg−r ),
2 r=1

where Dt = ∂t + κF0′′′ is the covariant derivative.


22 ZHIYUAN WANG AND JIAN ZHOU

Example 4.3. For the expression (51) of Fb2 , we have


1 ′′ 1 ′ 2 1 (4) 5
∂κ Fb2 = F + (F1 ) + κ( F0 + F1′ F0′′′ ) + κ2 (F0′′′ )2 ,
2 1 2 4 8
1 1 (5) 1 (4) 1
∂t Fb2 = F2′ + κ[ F1′′′ + F1′ F1′′ ] + κ2 ( F0 + F1′ F0 + (F1′ )2 F0′′′ + F1′′ F0′′′ )
2 8 2 2
2 (4) 5
+ κ3 [F1′ (F0′′′ )2 + F0′′′ F0 ] + κ4 (F0′′′ )3 .
3 8
Here in the second equality we have used (63).
Example 4.4. Direct computation gives
1 (4)
Dt ∂t Fb1 = F1′′ + κ( F0 + F1′ F0′′′ ) + κ2 (F0′′′ )2 ,
2
one may check that
1
∂κ Fb2 = (Dt ∂t Fb1 + (∂t Fb1 )2 ).
2
4.5. The general case. Now we return to the general case for an arbitrary dimen-
sion N . Similar to N = 1 case, the operators Kij and ∂i defined in Section 3 are
realized by ∂κij and ∂ti respectively. To realize the operator ∂i as the differential
operator ∂ti , we need to impose some conditions on the propagators κij analogous
to (63):
∂κjk X ∂ 3 F0
(65) = κjl κkm ,
∂ti ∂ti ∂tl ∂tm
l,m

or in matrix form:
∂κ ∂H(F0 )
(66) =κ κ
∂tk ∂tk
2
∂ F0
where H(F0 ) = ( ∂ti ∂tj
) is the Hessian of F0 . A general solution is given by

κ(t) = (C − H(F0 ))−1 ,


where C is a symmetric anti-holomorphic or constant matrix such that (C − H(F0 ))
is invertible.
The realization of the operator Di = ∂i + γi is more subtle. For our purpose,
it suffices to define the operator Dti ∂tj as a realization of the operator Di ∂j as
follows.
c
Definition 4.2. For Γ ∈ Gg,0 , we define
Dti ∂tj ωΓ =ωDi ∂j Γ
N
X
(67) ∂2 ∂ ∂ 3 F0
= ωΓ + ωΓ · κlm (t) ,
∂ti ∂tj ∂tl ∂tm ∂ti ∂tj
l,m=1

where
Y (val (v),··· ,valN (v))
Y
ωΓ = Fg(v)1 (t) · κl1 (e)l2 (e) (t, t̄).
v∈V (Γ) e∈E2 (Γ)

In particular, we see that Dti ∂tj ωΓ = Dtj ∂ti ωΓ .


Then by applying the Feynman rules, Theorem 3.1 gives us
HOLOMORPHIC ANOMALY EQUATIONS AND QUANTUM SPECTRAL CURVES 23

Theorem 4.3. When the condition (66) is satisfied, for g ≥ 2, we have the equa-
tions
g−1
X
(68) ∂κij Fbg = Dti ∂tj Fbg−1 + ∂ti Fbr ∂tj Fbg−r
r=1

for i 6= j, and
g−1
X
1
(69) ∂κii Fbg = (Dti ∂ti Fbg−1 + ∂ti Fbr ∂ti Fbg−r )
2 r=1

for the free energy Fbg defined by (50).

5. The holomorphic anomaly equation


In the rest of this paper we will present some examples of our formalism developed
above. In this section we will focus on the case of holomorphic anomaly equations.
As we have mentioned in the Introduction, the holomorphic anomaly equation
was introduced by Bershadsky et al in physics literatures [2] [3] in order to compute
the Gromov-Witten invariants of the quintic Calabi-Yau threefold. They developed
a method to solve the non-holomorphic free energies Fg (t, t̄) recursively using the
holomorphic anomaly equations
g−1
X
1
(70) ∂¯ī Fg = C̄īj̄ k̄ e2K Gj j̄ Gkk̄ (Dj Dk Fg−1 + Dj Fr Dk Fg−r ), g ≥ 2,
2 r=1

and by formally ‘taking a limit’


Fg (t) = lim Fg (t, t̄)
t̄→∞

they got the generating series of Gromov-Witten invariants F (t) of the threefold
on the small phase space. The non-holomorphic free energies Fg (t, t̄) are supposed
to have modularity while the holomorphic free energies Fg (t) does not, and this
has been exploited to use the theory of modular forms. Since the ring of modular
forms are often polynomial algebras with finitely many generators, this suggests
some polynomiality properties of the non-holomorphic free energies [25]. This has
led to many recent progress on the computation of Gromov-Witten invariants of
quintic Calabi-Yau 3-folds.
Unfortunately the mathematical or geometric definition of the nonholomorphic
free energy is lacking in the mathematical literature. According to [3], the nonholo-
morphic free energy should be closely related to some geometric structures on the
relevant moduli spaces. In the genus zero case, it is known that the free energy is
related to the special Kähler geometry of the moduli space [4]. In the genus one
case, the free energy is related to the tt∗ -geometry of the moduli space [5] [8] and
the theory of analytic torsions [2]. In the higher genus case, the nonholomorphic
free energy is supposed to be a suitable non-holomorphic section of a holomorphic
line bundle on the moduli space [3]. Such an approach depends heavily on the
special properties of Calabi-Yau 3-folds.
It is interesting to further clarify the geometric meaning of non-holomorphic free
energy, and furthermore, to find analogue of holomorphic anomaly equation for
Gromov-Witten theory of general symplectic manifolds or other related theories.
The formalism we develop in this paper shed some lights on this problem. In
24 ZHIYUAN WANG AND JIAN ZHOU

fact, our formalism is inspired by the direct integration approach to holomorphic


anomaly equations, and it does not need to start with a Calabi-Yau 3-fold.
Let us briefly recall the direct integration approach. According to [3], the non-
holomorphic free energy should count the contributions from degenerate Riemann
surfaces, and so it should involve the boundary strata of the Deligne-Mumford
moduli spaces of algebraic curves. Witten [23] interpreted the holomorphic anomaly
equation from the point of view of geometric quantization of symplectic vector
spaces and background independence. Inspired by this work, Aganagic, Bouchard
and Klemm [1] obtained via change of polarizations the following general form for
Fbg :
 
(71) b IJ
Fg (t, t̄) = Fg (t) + Γg ∆ , ∂I1 · · · ∂In Fr<g (t) ,

in some particular cases, ∆IJ can be taken to be −(τ − τ̄ )IJ , where τ = (τij ) =
∂ 2 F0
∂ti ∂tj ). These particular cases were interpreted for matrix models using Eynard-
Orantin topological recursion by Eynard, Mariño and Orantin [13]. Furthermore,
they also represented the partition function as a formal Gaussian integral [13, (4.27)]
and presented the Feynman graphs and Feynman rules for the terms that contribute
to Γg . These results have been generalized to other models by Grimm, Klemm,
Mariño and Weiss [15]. These authors reformulated the holomorphic anomaly equa-
tion as a quadratic recursion relation for the derivative of Fbg with respective to the
propagators ∆IJ (cf. [15, (7.50)]:
X g−1
∂ Fbg 1 bg−1 + 1
(72) = D I ∂J F ∂I Fbr ∂J Fbg−r .
∂∆IJ 2 2 r=1

From the consideration of modularity of Fbg , the propagators ∆IJ has the form
1
(73) ∆IJ = √ ((τ − τ̄ )−1 )IJ + E IJ ,
−1
where E IJ is a holomorphic function. They also derived the formal Gaussian integral
representation of the partition function and the Feynman expansions for Fbg .
It is clear that our formalism is inspired by [1,13,15]. The results in these works
when the propagators are
(74) κ = −(τ − τ̄ )−1
are special cases of our formalism. Indeed, the condition (66) clearly holds for
such propagators, and so our formalism can be applied to the free energy Fg of
Gromov-Witten theory of a compact Calabi-Yau 3-fold to get a sequence Fbg that
satisfies the holomorphic anomaly equations. It is remarkable that the application
of our formalism does not need the Calabi-Yau condition, and so it applies to the
Gromov-Witten theory of any compact symplectic manifold, any Gromov-Witten
type theory as the FJRW theory.
t30
Example 5.1. For FJRW theory of type A1 , F0 (t0 ) = 3! , and so we have
∂ 2 F0
(75) τ= = t0 ,
∂t20
HOLOMORPHIC ANOMALY EQUATIONS AND QUANTUM SPECTRAL CURVES 25

and so
1
(76) κ= .
t̄0 − t0
The free energy functions Fg (t0 ) are generating functions of intersection numbers
on Mg,n , restricted to the small phase space ti = 0 for i > 0. Therefore, for g ≥ 1,
(77) Fg (t0 ) = 0.
Then our construction yields a sequence Fbg (t, t̄):
Example 5.2. For the Gromov-Witten theory of P1 , the free energy function re-
stricted to the small phase space is given by
t20 t1
F0 (t0 , t1 ) = + qet1 ,
2
t1
F1 (t0 , t1 ) = − ,
24
Fg (t0 , t1 ) = 0, g ≥ 2,
where q is the degree tracking parameter. The period matrix τ can be directly com-
puted as follows:
 
t1 t0
(78) τ=
t0 qet1
and so the propagator is
 −1
t1 − t̄1 t0 − t̄0
(79) κ=−
t0 − t̄0 qet1 − qet̄1
It is an interesting problem to extend our construction to the big phase space that
includes all the gravitational descendants of primary observables. It is well-known
that Gromov-Witten theory and its various generalizations in the mathematical
literature describes what physicists refer to as the topological string theory. The
mathematical meaning or definition of the partition function of what physicists refer
to as the string theory needs to be clarified at present. For example, for Gromov-
Witten theory, the free energy Fg depends on infinitely many coupling constants
{ti,n }1≤i≤N,n≥0 , we expect the string theory behind it to have free energy Fbg that
also depends on {t̄i,n }1≤i≤N,n≥0 . We hope our construction shed some lights on
this problem.
When the propagators take the form
(80) κ = −(τ − τ̄ )−1 + E,
where E is a symmetric matrix whose entries are holomorphic functions, the condi-
tion (66) do not hold in general. This means that we need to modify the diagram-
matics we discuss in this paper. We will do this in a subsequent work [22].

6. Application to Topological 1D Gravity and Quantum Spectral


Curves
In this section we present the second class of applications of our formalism.
We will apply it to topological 1D gravity [27]. The propagator in this case is
holomorphic. By making particular choices of coupling constants, this gives us
examples concerning the quantum spectral curves [16].
26 ZHIYUAN WANG AND JIAN ZHOU

6.1. Some results in topological 1D gravity. In [27], the second author defined
the partition function of the topological 1D gravity as a formal Gaussian integral:
Z
1 1 
(81) Z 1D = 1 dx exp 2 S(x) ,
(2πλ2 ) 2 λ

where the action S(x) is


1 X xn
(82) S(x) = − x2 + tn−1 .
2 n!
n≥1

The following explicit formula is given in [27, Prop. 4.5]:

X X Pk k
Y
(2n − 1)!! 2n−2 mj m j
(83) Z= Qk λ j=1 · tj−1 .
n≥0
Pk
mj j=2n j=1 (j!)mj m j! j=1
j=1

Using Wick’s theorem, it is also straightforward to write down the Feynman graphs
and the Feynman rules
X 1 Y
(84) Z= λval(v)−2 tval(v)−1 ,
| Aut(Γ)|
Γ∈G v∈V (Γ)

where the sum is taken over the set G all possible graphs, with the following Feyn-
man rules:

(85) w(v) = λval(v)−2 tval(v)−1 ,


(86) w(e) = 1.

One of the results obtained in [27] is that after a natural change of variables, the
free energy functions Fg have nice simple expressions. This coordinate change can
be obtained by the idea of renormalization. Alternatively, one can formally apply
the stationary phase method and find the critical point of S(x) and consider the
Taylor expansion of S(x) at its critical point. This gives us the following change of
coordinates:

X X
1 tp1 tp
I0 = ··· k ,
k p1 ! pk !
k=1 p1 +···+pk =k−1
(87)
X In
Ik = tn+k 0 , k ≥ 1,
n!
n≥0

and
X∞
(−1)n I0n
(88) tk = In+k ,
n=0
n!

where I0 is the critical point of S(x), and In −δn,1 (n ≥ 1) are the Taylor coefficients
at x = I0 . Using the new coordinate {Ik }k≥0 , the action (82) can be rewritten as

X X∞
(−1)k (In−1 − δn,2 )
(89) S(x) = (Ik + δk,1 )I0k+1 + (x − I0 )n .
(k + 1)! n=2
n!
k=0
HOLOMORPHIC ANOMALY EQUATIONS AND QUANTUM SPECTRAL CURVES 27

Then the partition function (81) can be rewritten in terms of another formal Gauss-
ian integral:
 X ∞ 
1D 1 (−1)k k+1 1 1
Z = exp 2 (Ik + δk,1 )I0 + log
λ (k + 1)! 2 1 − I1
k=0
(90) Z  
1 1 1 2 X In−1 xn
· 1 dx exp (− x + n · ) .
(πλ2 ) 2 λ2 2 (1 − I1 ) 2 n!
n≥3

This is interpreted in [27] as giving a mean field theory of the P topological 1D


gravity. Using this new equivalent theory, the free energy F 1D = g≥0 λ2g−2 Fg1D =
log(Z 1D ) of this theory can be written down as follows:

X (−1)k
F01D = (Ik + δk,1 )I0k+1 ,
(k + 1)!
k=0
1 1
(91) F11D = log ,
2 1 − I1
X 2g−1
Y  lj
l 1 Ij
Fg1D = hτ2l2 · · · τ2g−1
2g−1
ig , g ≥ 2.
P2g−1 j−1 j=2
lj ! (1 − I1 )(j+1)/2
j=2 2 lj =g−1

By rewriting the theory in the new coordinates {Ik }, the free energy F 1D has a
different Feynman diagram expansion [27, (130)]:
X 1 Y Y 1
(92) Fg1D = Ival(v)−1 · ,
c,0
| Aut(Γ)| 1 − I1
Γ∈Gg,0 v∈V (Γ) e∈E(Γ)

c,0 c
for g ≥ 2, where Gg,0 is the subset of Gg,0 consists of stable graphs whose vertices
are all of genus 0.

6.2. Topological 1D gravity as realization of the diagrammatics of stable


graphs. Now we see that the Feynman rules (92) is indeed a special case of the
realization of diagrammatics of stable graphs described in Section 4. I.e., Fg1D = Fbg
(g ≥ 2) for suitable κ and a suitable sequence Fg (X) for g ≥ 0. By comparing
(92) with the Feynman rules (49), it is clear that for topological 1D gravity, the
propagator is taken to be
1
(93) κ= .
1 − I1
Since in (92), the summation is taken over stable graphs with only genus zero
vertices, so we have Fg (X) = 0 for g ≥ 1, and
(val(v))
(94) F0 (X) = Ival(v)−1 ,
where val(v) ≥ 3. In other words, we need a function F0 (X) such that
dn
(95) F0 (X) = In−1
dX n
for n ≥ 3. Since In−1 involves an infinite set of indeterminates t0 , t1 , . . . , it is not
clear how to choose one of them or a combination of them to be X. We will take
28 ZHIYUAN WANG AND JIAN ZHOU

a different approach as follows. Regard {Ik }k≥0 as some parameters, and X as a


new independent variable. Now we define a function F0 (X) by
X∞ X∞
(−1)k (In−1 − δn,2 )
(96) F0 (X) = (Ik + δk,1 )I0k+1 + (X − I0 )n .
(k + 1)! n=2
n!
k=0

Write X̃ = X − I0 , then

X X∞
(−1)k (In−1 − δn,2 ) n
(97) F0 (I0 + X̃) = (Ik + δk,1 )I0k+1 + X̃ .
(k + 1)! n=2
n!
k=0
This is equivalent to say

X (−1)k
F0 (I0 ) = (Ik + δk,1 )I0k+1 ,
(k + 1)!
k=0
dF0 (X)
(98) = 0,
dX X=I0

dn F0 (X)
= In−1 − δn,2 , n ≥ 2.
dX n X=I0

Now using the formalism in Section 4, we can construct a sequence Fbg (X) from
F (X) = λ−2 · F0 (X), and the propagator
1
(99) κ = − ′′
F0 (X)
which clearly satisfies the condition (63). Hence (93), (94)and (95) hold when
X = I0 . For g ≥ 2 we have
X  Y 
1 (val(v))
(100) Fbg (X) = F0 (X) · κ|E(Γ)| ,
c,0
| Aut(Γ)|
Γ∈Gg,0 v∈V (Γ)

Then evaluate this expression at X = I0 using (98) and (99), we get


(101)
X  Y 
1 (val(v)) 1
Fbg (I0 ) = F0 (X) · (− )|E(Γ)|
c,0
| Aut(Γ)| X=I0 F0′′ (X)|X=I0
Γ∈Gg,0 v∈V (Γ)

X  Y   |E(Γ)|
1 1
= Ival(v)−1 · ,
c,0
| Aut(Γ)| 1 − I1
Γ∈Gg,0 v∈V (Γ)

this is exactly the Feynman rule (92), i.e., we have proved the following
Theorem 6.1. The free energy of the topological 1D gravity can be realized by the
abstract free energy:
(102) Fg1D (I0 , I1 , I2 , · · · ) = Fbg (X)|X=I0
for g ≥ 2. Here F = λ−2 F0 and κ are given by (96) and (99) respectively.
Example 6.1. F21D and F31D are explicitly given by:

1
F21D = 12 + 18 + 18
,
HOLOMORPHIC ANOMALY EQUATIONS AND QUANTUM SPECTRAL CURVES 29

1 1 1 1
F31D = 48 + 48 + 16 + 16

1
+ 12 + 81 1
+ 16

1
+ 12 + 81 + 81 1
+ 48

1 1 1
+ 24 + 16 + 16 + 81
,
or,
5 2 3 1
F21D = I κ + I3 κ2 ,
(103) 24 2 8
1 1 7 25 5
F31D = I5 κ + ( I32 + I2 I4 )κ4 + I22 I3 κ5 + I24 κ6 ,
3
48 12 48 48 16
1
where κ = 1−I1 .

As a corollary to Theorem 4.2 and Theorem 6.1, we get a quadratic recursion


relation for F 1D as follows. First we have
g−1
X
1
(104) ∂κ Fbg = ((∂X + κF0′′′ (X))∂X Fbg−1 + ∂X Fbr ∂X Fbg−r ),
2 r=1

where Fb1 = F11D = 1


2 log κ, next we restrict to X = I0 to get:
g−1
X
1
(105) ∂κ Fg1D = 1D
((dX + κI2 )dX Fg−1 + dX Fr1D dX Fg−r
1D
),
2 r=1

where dX is a differential operator such that


(106) dX Ik = Ik+1 , k ≥ 1, dX κ = κ2 · I2 .

This also provides a way to compute Fg1D recursively without listing all possible
Feynman graphs.

Example 6.2. For g = 4, we have


∂F41D
2 = (dX + κI2 )∂I0 F31D + 2dX F11D · dX F31D + (dX F21D )2
∂κ
1 5 25 21 113 2 33
= I7 κ3 + ( I2 I6 + I3 I5 + I42 )κ4 + ( I I5 + I33 + 7I2 I3 I4 )κ5
48 16 48 64 48 2 24
3115 2 2 1127 3 985 4 1105 6 8
+( I I + I I4 )κ6 + I I3 κ7 + I κ ,
144 2 3 96 2 24 2 64 2
30 ZHIYUAN WANG AND JIAN ZHOU

Then F41D can be obtained by integrate with respext with κ:


1 1 5 21 2 5
F41D = I7 κ4 + ( I2 I6 + I3 I5 + I )κ
384 32 96 640 4
113 2 11 7 445 2 2 161 3
(107) +( I I5 + I33 + I2 I3 I4 )κ6 + ( I I + I I4 )κ7
576 2 96 12 288 2 3 192 2
985 4 1105 6 9
+ I I3 κ8 + I κ .
384 2 1152 2
6.3. Example: The Airy function and quantum Airy curve. Our first ex-
ample is the Airy function
Z
 t3 
(108) Z A (t) = C dη exp − λ−2 (−ηt + ) ,
3
3
where C is a suitable normalization constant. Let f (η) = −ηt + t3 , then the critical
1
points of f (η) is η = ±t 2 . Denote by c one of the critical points and η̃ = η − c,
then
1 2 1
(109) f (η) = −(η̃ + c)c2 + (η̃ + c)3 = − c3 + cη̃ 2 + η̃ 3 .
3 3 3
Therefore in this case
I0 = c, I1 = 1 − 2c, I2 = −2;
(110)
Ik = 0, k ≥ 3.

And the propagator is


1 1
(111) κ= = .
1 − I1 2c
Using the coordinate change (88), this particular point (110) in the space of
coupling constants is

t0 = c 2 , t1 = 1, t2 = −2,
(112)
tk = 0, k ≥ 3.

In what follows we will use ∂c to replace the operator dT since we have ∂I
∂c = Ik+1
k

∂κ 2
for k ≥ 1 and ∂c = κ I2 . Now using our formalism we can compute the free energies
FgA associated with Z A . The Feynman graphs of this integral only involves cubic
graphs without external edges, whose vertices are all of genus zero.

Example 6.3. The free energies F0A and F1A are given by
2 1 1
(113) F0A = − c3 , F1A = log .
3 2 2c
And F2A and F3A are given by

1
F2A = 12 + 18
,
HOLOMORPHIC ANOMALY EQUATIONS AND QUANTUM SPECTRAL CURVES 31

1 1 1
F3A = 8 + 16 + 16

1 1
+ 24 + 48

,
or,
5 5
(114) F2A = , F3A = .
48c3 64c6
Using (111), for g ≥ 1 we have
∂FgA
∂κ FgA = (−2c2 ) · ,
∂c
thus Theorem 4.2 give us the following recursion relation.
Theorem 6.2. For g ≥ 2 we have
g−1
X ∂F A ∂Fg−r
∂FgA 1  ∂ A
1 ∂Fg−1 r
A 
(115) =− 2 ( − )+ .
∂c 4c ∂c c ∂c r=1
∂c ∂c
1
Note that in this case FgA must be a monomial in κ = 2c of degree 3g − 3, thus
A
Fg can be solved recursively using this relation. For example,
1105 565
(116) F4A = , F5A = , ···
9216c9 2048c12
The recursion relation (115) is equivalent to the equation
∂ 2 
(117) (~ ) − t Z A (t) = 0
∂t
if we set ~ = λ2 . This is the Schrödinger equation of the quantum Airy curve.
Now let us recall some results about the quantum Airy curve [16,26] and compare
with the above construction. First let u(z) = 21 z 2 , v(z) = z be a parametrization
of the Airy curve
1 2
A(u, v) = v − u = 0,
2
then using the Eynard-Orantin topological recursion [12] one may get a family of
differentials
Wg,n (p1 , · · · , pn ) = Wg,n (z1 , · · · , zn )dz1 · · · dzn .
These invariants associated to the Airy curve encode the information of the Witten-
Kontsevich tau-function [18, 24, 28]. Following [16], define

X
(118) ZAiry = exp( ~n−1 Sn ),
n=0
32 ZHIYUAN WANG AND JIAN ZHOU

where
Z z
S0 (z) = v(z)du(z),
1 du
(119) S1 (z) = − log( ),
2 dz
X Z z Z z
(−1)k
Sn (z) = ··· Wg,k (z1′ , · · · , zk′ )dz1′ · · · dzk′ , n ≥ 2.
k!
2g−1+k=n

Then Sk can be used to construct the quantization of the Airy curve. The following
result was proposed by Gukov-Sulkowski in [16], and proved in [26].
Theorem 6.3. ( [26]) The function ZAiry satisfies
(120) ÂZAiry = 0,
where
1 2 1
(121) v̂ − û = (~∂u )2 − u·
 =
2 2
is the quantization of the Airy curve.
One computes
1 3 1
S0 =
z , S1 = − log(z),
(122) 3 2
5 5 1105
S2 = , S3 = , S4 = , ···
24z 3 16z 6 1152z 9
We make a comparison between this and (114), (116), then we might simply expect
FgA = Sg for g ≥ 2 once we set z 3 = 2c3 . In fact, this can be proved by the following
recursion relation.
Lemma 6.1. ( [26]) For n > 2, we have
1 2 1 X
(123) ∂u Sn−1 + ∂u S0 · ∂u Sn + ∂u S1 · ∂u Sn−1 + ∂u Si · ∂u Sj = 0.
2 2 i+j=n
i,j≥2

This lemma can be obtained from Theorem 6.3 by expanding (120) directly as
a series in ~ and comparing the coefficients. Using z 3 = 2c3 and u = 12 z 2 , this
recursion is equivalent to (115).
In conclusion, for the particular choices (110) of coupling constants, the pro-
cedure to produce FgA is equivalent to the solution of quantization of Airy curve
obtained by the Eynard-Orantin topological recursion.

6.4. Example: The Kontsevich-Penner matrix model for N = 1. Our next


example is the Kontsevich-Penner matrix model for N = 1 [18,21], which is related
to the quantization of the Catalan curve
1
x=z+ ,
z
see [16, 19]. The Schrödinger equation in this case is
∂2 ∂
(124) [~2 + ~t + (1 − ~)]Z(t) = 0,
∂t2 ∂t
HOLOMORPHIC ANOMALY EQUATIONS AND QUANTUM SPECTRAL CURVES 33

where the partition function is the formal integral


Z
 1 
(125) Z (t) = dη exp − λ−2 ( η 2 + ηt + logη) ,
C
2

up to a normalization constant, and we set ~ = λ2 .


Now let us understand this integral using the model of topological 1D gravity. √
2
Write f (η) = 21 η 2 + ηt + log(η), we expand f (η) at its critical points c = −t± 2 t −4 .
Let η̃ = η − c, then

f (η) =f (η̃ + c)
(126) c2 1 η̃ 3 η̃ 4 η̃ 5
=(− − 1 + log(c)) + (1 − c−2 )η̃ 2 + 3 − 4 + 5 − · · ·
2 2 3c 4c 5c
By taking
1
(127) κ=
1 − c−2
and
(−1)k+1 · k!
(128) Ik = , k ≥ 1,
ck+1
we see Ik′ = Ik+1 and κ′ = κ2 · I2 where the prime means taking derivative with
respect to c. Thus we get a special case of the theory in Section 6.2, thus we obtain
a sequence of free energies FgC . Using the formula for FgC for g = 2, 3, 4 given by
(103) and (107), we get:

c2 1 1
F0C = − − 1 + log(c), F1C = log ,
2 2 1 − c−2
3 5
F2C = 2 2
+ ,
4(c − 1) 6(c − 1)3
2

5 10 25 5
F3C = + 2 + + 2 ,
2(c2 − 1)3 (c − 1)4 2(c2 − 1)5 (c − 1)6
105 507 6391 767
F4C = 2 4
+ 2 5
+ 2 6
+
8(c − 1) 5(c − 1) 24(c − 1) 2(c − 1)7
2

985 1105
+ + .
4(c2 − 1)8 18(c2 − 1)9

The quadratic recursion relation in Theorem 4.2 reads


g−1
X ∂
1 ∂ ∂ C ∂ C
(129) ∂κ FgC = [( + κ · I2 ) Fg−1 + FrC · Fg−r ], g ≥ 2.
2 ∂c ∂c r=1
∂c ∂c


To write this relation completely in terms of ∂c , we need the following lemma.

Lemma 6.2. For g ≥ 2, we have


∂ C 2κ
(130) F = (κ2 · I2 − )∂κ FgC .
∂c g c
34 ZHIYUAN WANG AND JIAN ZHOU

Proof. Using the homogeneity we could see that every monomial in Fg1D is of the
form ag,i1 ···ik Ii1 · · · Iik · κl with i1 + · · · ik + k = 2l. Thus
Pk l l−1

∂c (Ii1 · · · Iik κ )
l
j=1 Ii1 · · · Iij +1 · · · Iik κ + Ii1 · · · Iik lκ · κ2 I2
=
∂κ (Ii1 · · · Iik κl ) Ii1 · · · Iik lκl−1
k
1 X Iij +1
= · κ + κ2 · I2
l j=1 Ij
k
1 X (−1)(ij + 1)
= · κ + κ2 · I2
l j=1 c
2
= − · κ + κ2 · I2 ,
c
which proves the lemma. 

Now using this lemma and


2κ 2c3
κ2 · I2 − =− 2 ,
c (c − 1)2
we rewrite the recursion relation as follows.
Theorem 6.4. For the Kontsevich-Penner model, for g ≥ 2, we have
g−1
X ∂
∂ C c3  ∂ 2 ∂ C ∂ C 
(131) Fg = − 2 − Fg−1 + FrC · Fg−r .
∂c (c − 1)2 ∂c c(c2 − 1) ∂c r=1
∂c ∂c

Or in terms of t = −(c + 1c ),
g−1
X ∂
1 ∂ ∂2 C ∂ C
(132) (c − ) · FgC = 2 Fg−1 + FrC · Fg−r .
c ∂t ∂t r=1
∂t ∂t

The second equation in the above theorem is exactly the same as the recursive
formula for the quantization of the Catalan curve [19, (4.13)]. Similar to the case
of quantum Airy curve, the free energies FgC here can also be constructed from the
Catalan curve using the Eynard-Orantin topological recursion.

6.5. Example: Enumeration of stable graphs with genus zero vertices. In


this subsection we apply topological 1D gravity and our formalism to study the
problem of enumeration of stable graphs with only genus zero vertices. In this
model, all vertices of the Feynman graphs have valence ≥ 3. The partition function
is given by
Z X xn
st κ1/2 −2 x2
(133) Z (T, κ) = dx exp[λ (T − )].
(2πλ2 )1/2 n≥3
n! 2κ

Write the free energy F st = log Z st in the following expansion:


X 2g−2
X
(134) F st (T, κ) = agd T d κg−1+d .
g≥2 d=1
HOLOMORPHIC ANOMALY EQUATIONS AND QUANTUM SPECTRAL CURVES 35

By (83)
X X X 
st λ2l−2k
(135) F (T, κ) = log · (2l − 1)!! · T k · κl · δl1 +···+lk ,2l .
k! · l1 ! · · · lk !
k≥0 li ≥3 l

The coefficients agd counts the number of connected stable graphs of genus g with d
vertices of marked genus 0, weighted by the inverse of the orders of their automor-
phism groups. Such graphs has g − 1 + d edges. Using this interpretation one gets
the following explicit formulas for agd , d = 1, 2:
1 1
ag1 = = g ,
(2g)!! 2 g!
g+1
(136) 1X X 1
ag2 = .
2 k! · (2l)!! · (2g + 2 − 2k − 2l)!!
k=1 k+2l≥3
2g+2−k−2l≥3

By an elementary calculation, one can simplify the second equality to get


22g − 2g−1 − (g + 1)2
(137) ag2 = .
2g · (g + 1)!
Since ag2g−2 counts trivalent graphs of genus g, it equals to bg−1 in [9, (65), (66)],
and so
X g X∞ 
g−1 (6m)! x m
(138) a2g−2 x = log .
m=0
(3m)!(2m)! 288
g≥2

It is interesting to compute agd for 3 ≤ d < 2g − 2. Formula (135) and the direct
counting of graphs are not very effective for this purpose. We will apply the qua-
dratic recursion relations developed in Section 6.2 to derive some recursion relations
for agd .
In (105) and (106), we take Ik = T for all k ≥ 1, and then dX changes to an
operator D such that
DT = T, Dκ = κ2 T
then we get from (105) the following recursion relation:
g−1
X
1 
(139) ∂κ Fgst st
= (D + κT )DFg−1 + DFrst · DFg−r
st
,
2 r=1

where we have to add the following convention:


1
(140) F1st = log κ.
2
From this we recursively:
T κ2 5T 2 κ3
F2st = + ,
8 24
T κ3 11T 2 κ4 25T 3κ5 5T 4 κ6
F3st = + + + ,
48 48 48 16
T κ4 223T 2κ5 515T 3κ6 1373T 4κ7 985T 5κ8 1105T 6κ9
F4st = + + + + + .
384 1920 576 576 384 1152
36 ZHIYUAN WANG AND JIAN ZHOU

We have
X X
DFgst = agd dT d κg−1+d + agd (g − 1 + d)T d+1 κg+d ,
d d
X X
D 2
Fgst = agd d2 T d κg−1+d + agd (g − 1 + d)dT d+1 κg+d
d d
X X
+ agd (g − 1 + d)(d + 1)T d+1 g+d
κ + agd (g − 1 + d)(g + d)T d+2 κg+d+1 .
d d

Then (139) gives us

X
2 agd (g − 1 + d)T d κg−2+d
d
g−1
X
st
= (D + κT )DFg−1 + DFrst DFg−r
st

r=1
X X
= ag−1
d d2 T d κg−2+d + ag−1
d (g − 2 + d)dT d+1 κg−1+d
d d
X
+ ag−1
d (g − 2 + d)(d + 1)T d+1 κg−1+d
d
X
+ ag−1
d (g − 2 + d)(g − 1 + d)T d+2 κg+d
d
X X 
+2 ag−1
d dT d+1 κg−1+d + ag−1
d (g − 2 + d)T d+2 g+d
κ
d d
XX
g−2 X 
+ ard dT d κr−1+d + ard (r − 1 + d)T d+1 r+d
κ ·
r=2 d d
X X 
ag−r
d dT d κg−r−1+d + ag−r
d (g − r − 1 + d)T d+1 κg−r+d
d d

for g ≥ 3. Therefore the following quadratic recursion relations hold for agd :

X X g−1
1
agd = ag−1
d d2 + ad−2 (g − 4 + d)(g − 1 + d)
2(g − 1 + d)
d d
X g−1  
+ ad−1 (g − 3 + d)(2d − 1) + 2(d − 1)
d
g−2
X X X
+ ard1 ag−r
d2 d1 d2 + ard1 ag−r
d2 (r − 1 + d1 )d2
r=2 d1 +d2 =d d1 +d2 =d−1
X
+ ard1 ag−r
d2 d1 (g − r − 1 + d2 )
d1 +d2 =d−1
X 
+ ard1 ag−r
d2 (r − 1 + d1 )(g − r − 1 + d2 ) .
d1 +d2 =d−2
HOLOMORPHIC ANOMALY EQUATIONS AND QUANTUM SPECTRAL CURVES 37

6.6. Example: Enumeration of graphs. In this subsection we study another


example: the model describing enumeration of graphs, not necessarily stable, in-
troduced and studied in [9]. The partition function for this case is
Z
1  η2 
(141) Z(t) = √ dη exp λ−2 (t · eη − ) .
2πλ2 2
η2
The critical point η = T of the function f (η) = t · eη − 2 is given by
T
(142) T =t·e ,
whose solution may be explicitly given by the Lambert series [9]:

X dd−1
(143) T = td .
d!
d=1

Write x = η − T , then the expansion f (η) at T is



T2 teT − 1 2 X teT n
f (η) = (teT − )+ x + x
2 2 n=3
n!

T2 T −1 2 X T n
= (T − )+ x + x .
2 2 n=3
n!
Therefore in this case, the propagator is
1
(144) ,
κ=
1−T
and the point on the space of coupling constants is specified by
(145) I0 = T ; Ik = T, k ≥ 1,
or in the coordinates {tk },
(146) tk = t, k ≥ 0.
By (92), the free energy is
X 1 T |V (Γ)|
(147) Fggr = · ,
c,0
| Aut(Γ)| (1 − T )|E(Γ)|
Γ∈Gg,0

For example,
T2 1 1
F0gr = T− , F1gr = log ,
2 2 1−T
T 5T 2
F2gr = + ,
8(1 − T )2 24(1 − T )3
T 11T 2 25T 3 5T 4
F3gr = + + + ,
48(1 − T )3 48(1 − T )4 48(1 − T )5 16(1 − T )6
T 223T 2 515T 3 1373T 4
F4gr = 4
+ 5
+ 6
+
384(1 − T ) 1920(1 − T ) 576(1 − T ) 576(1 − T )7
985T 5 1105T 6
+ 8
+ .
384(1 − T ) 1152(1 − T )9
These match with Gk in [9, (63)].
38 ZHIYUAN WANG AND JIAN ZHOU

By comparing (141) with (133), it is clear that for g ≥ 2,


1
(148) Fggr = Fgst (T, ).
1−T
Therefore, for g ≥ 2 the free energy Fggr can be written in the form
2g−2
X agd · T d
Fggr = .
(1 − T )d+g−1
d=1

Moreover, note that for g ≥ 2, Fggr is of the form


fg (T )
(149) Fggr = ,
(1 − T )3g−3
where fg (T ) is a polynomial of degree ≤ 2g −2 without constant term. For example,
T
F2gr = (3 + 2T ),
24(1 − T )3
T
F3gr = (1 + 8T + 6T 2 ),
48(1 − T )6
T
F4gr = (15 + 594T + 2624T 2 + 2144T 3 + 164T 4 − 16T 5),
5760(1 − T )9
T
F5gr = (3 + 465T + 6730T 2 + 21940T 3 + 18940T 4
11520(1 − T )12
+ 3012T 5 − 240T 6),
T
F6gr = (63 + 35568T + 1349298T 2 + 11582816T 3 + 31178616T 4
2903040(1 − T )15
+27897072T 5 + 6526912T 6 − 266448T 7 − 36576T 8 + 2304T 9).
Let
2g−2
X
fg (T ) = bgn · T n ,
n=1

then the two type of coefficients {agk }


and {bgk } are related by
k
X  
2g − 2 − l g
(150) bgk = (−1)k+l al .
k−l
l=1

In particular,
1
bg1 =ag1 = ,
(2g)!!
(151) bg2 =ag2 − (2g − 3)ag1
22g − 2g−1 − (g + 1)2 2g − 3
= − .
2g · (g + 1)! (2g)!!
P3g−3
As pointed out in [9, (64), (65)], Fggr is of the form Fggr = n=g−1 λg,n κn , and
Bg
λg,g−1 = , λg,3g−3 = bg−1 ,
g(g − 1)
HOLOMORPHIC ANOMALY EQUATIONS AND QUANTUM SPECTRAL CURVES 39

where Bg are the Bernoulli numbers, and bg are given by


X∞ X∞ 
r (6m)! x m
br x = log .
n=1 m=0
(3m)!(2m)! 288

The numbers {bg } are exactly the coefficients appearing in the expansion of the
Airy function (122). Here are some examples of Fggr written as polynomials in κ:
5 3 7 1
F2gr = κ − κ2 + κ,
24 24 12
5 6 35 5 13 4 1 3
F3gr = κ − κ + κ − κ ,
16 48 24 8
1105 9 1225 8 2273 7 313 6 227 5 17 4 1 3
F4gr = κ − κ + κ − κ + κ − κ − κ ,
1152 384 576 144 480 1440 360
565 12 14665 11 76367 10 11191 9 2557 8 7993 7
F5gr = κ − κ + κ − κ + κ − κ
128 768 2304 384 192 2880
37 6 1
+ κ + κ5 ,
320 48
82825 15 441245 14 493235 13 16116187 12 2827135 11
F6gr = κ − κ + κ − κ + κ
3072 3072 1536 41472 10368
1884983 10 567289 9 7489 8 10249 7 47 6 1 5
− κ + κ − κ − κ + κ + κ ,
17280 25920 6480 60480 10080 1260
19675 18 7969325 17 65405005 16 453853985 15
F7gr = κ − κ + κ − κ
96 6144 18432 82944
215237149 14 64035527 13 23126555 12 11204309 11
+ κ − κ + κ − κ
41472 20736 20736 51840
1352989 10 6481 9 1927 8 1
+ κ + κ − κ − κ7 .
103680 4032 20160 72
1
Plug T = 1 − κ into (135), we get the generating series for {λg,k }:

X 3g−3
X
λ2g−2 λg,d κd
g=2 d=g−1
(152) X X X 
λ2l−2k k l−k
= log · (2l − 1)!! · (κ − 1) κ · δl1 +···+lk −2l,0 .
k! · l1 ! · · · lk !
k≥0 li ≥3 l

The relations among {agk }, {bgk } and {λg,k } are given by (150) and
X l g
2g−2
k
λg,3g−3−k = (−1) b ,
k l
l=k
X 3g − 3 − j 
3g−3
g n
bn = (−1) λg,j ,
j=g−1
n
k
X  
k+l 2g − 2 − l
λg,3g−3−k = (−1) a2g−2−l .
k−l
l=0

Now we apply the quadratic recursion relations for topological 1D developed


in Section 6.2 to this problem. For this purpose, we need to find an operator that
∂ ∂
satisfies the conditions in (106). We can take dX to be D = ∂(log T ) = T ∂T , because
40 ZHIYUAN WANG AND JIAN ZHOU

we have
1
Ik = T, κ= ,
1−T
the conditions in (106) are satisfied by D. So we get the following equation from
(105):
g−1
X
1 gr gr
(153) ∂κ F gr = ((D + κT )DFg−1 + DFrgr DFg−r ),
2 r=1

Lemma 6.3. For g ≥ 2, we have


(154) (κ + κ2 T )∂κ Fggr − DFggr = (g − 1)Fggr .

Proof. Every monomial in Fggr for g ≥ 2 is of the form ag,k T k κk+g−1 . Then the
lemma follows from
κ∂κ (T k κk+g−1 ) = (k + g − 1)T k κk+g−1
and
D(T k κk+g−1 ) = kT k κk+g−1 + (k + g − 1)T k κk+g−2 · κ2 T.

1
Since κ + κ2 T = (1−T )2 , (153) and Lemma 6.3 give us:

Theorem 6.5. For g ≥ 2, we have


g−1
X
1 T gr gr
(155) (D + g − 1)Fggr = [(D + )DFg−1 + (DFrgr )(DFg−r )].
2(1 − T )2 1−T r=1

This theorem provides a way to compute Fggr recursively. In fact, the left-hand
side of (155) can be rewritten as
(D + g − 1)Fg1D =T Fg′ (T ) + (g − 1)Fg (T )
(156) d
=T 2−g (T g−1 · Fg (T )),
dT
thus Fg1D is determined by the formula
(157)
Z T  g−1
X 
1D 1−g T g−2 T 1D 1D 1D
Fg (T ) = T · 2
(D + )DFg−1 + (DFr )(DFg−r ) dT.
0 2(1 − T ) 1−T r=1

Moreover, for g ≥ 2, when Fggr (T ) is written as a polynomial in the propagator


κ = (1 − T )−1 using T = 1 − κ1 , the recursion relation (155) can be written as
follows.
Theorem 6.6. For g ≥ 2, we have
 gr gr
 2 d  gr κ2 d2 Fg−1 dFg−1
(κ − κ) + g − 1 Fg = (κ2 − κ)2 + (3κ − 2)(κ 2
− κ)
dκ 2 dκ2 dκ
(158) g−1 gr 
X dF gr dFg−r
r
+ (κ2 − κ)2 · .
r=1
dκ dκ
HOLOMORPHIC ANOMALY EQUATIONS AND QUANTUM SPECTRAL CURVES 41

In particular, for g ≥ 3,
 gr gr
 2 d  κ2 d2 Fg−1 dFg−1
(κ − κ) + g − 1 Fggr = (κ2 − κ)2 + (4κ − 3)(κ 2
− κ)
dκ 2 dκ2 dκ
(159) g−2 gr 
X dF gr dFg−r
r
+ (κ2 − κ)2 · .
r=2
dκ dκ

The equation (159) can be rewritten as


 
d g−1 g
F gr
(1 − κ)
dκ κg−1
 
1 1 g−2 gr 2
dFggr
=− 2 −1 (g − 1)Fg + (κ − κ)
κ κ dκ
(160) g−2
 2 gr gr
1 (1 − κ) 2 2
d F g−1 2
dFg−1
=− (κ − κ) + (4κ − 3)(κ − κ)
2 κg−2 dκ2 dκ
g−2
X dF gr dFg−r gr 
r
+ (κ2 − κ)2 · .
r=2
dκ dκ
Now if we want to integrate this equality with respect to κ, we need to determine
Fggr
the constant of integration. Note that (1 − κ)g−1 κg−1 is a polynomial in κ whose
Bg
constant term is λg,g−1 = g(g−1) , therefore
 Z κ  gr
gr κg−1 Bg 1 (1 − κ)g−2 2 2
d2 Fg−1
Fg = + − (κ − κ) +
(1 − κ)g−1 g(g − 1) 0 2 κg−2 dκ2
(161) gr g−2 gr  
dFg−1 X dFrgr dFg−r
2 2 2
(4κ − 3)(κ − κ) + (κ − κ) · dκ .
dκ r=2
dκ dκ
Now let us derive some recursion relations for the coefficients {λg,k } using The-
orem 6.6. Expand the two sides of (159), we get
X 
l(κ − 1) + (g − 1) λg,l κl
l

κ2 X X
= (κ − 1)2 l(l − 1)λg−1,l κl + (4κ − 3)(κ − 1)lλg−1,l κl
2
l l
g−2
XX 
2 l1 +l2
+ (κ − 1) l1 l2 λr,l1 λg−r,l2 κ ,
r=2 l1 ,l2

which gives us
(l − 1)λg,l−1 + (g − l − 1)λg,l
1 1 1
= (l − 4)(l − 1)λg−1,l−4 − (l − 3)(2l − 1)λg−1,l−3 + (l − 2)lλg−1,l−2
2 2 2
g−2
X  X X
(162) 1
+ l1 l2 λr,l1 λg−r,l2 − 2 l1 l2 λr,l1 λg−r,l2
2 r=2
l1 +l2 =l−4 l1 +l2 =l−3
X 
+ l1 l2 λr,l1 λg−r,l2 .
l1 +l2 =l−2
42 ZHIYUAN WANG AND JIAN ZHOU

Note that λg,n = 0 unless g − 1 ≤ n ≤ 3g − 3, thus the above relation indeed


determines {λg,n } uniquely. In fact, first we set l = 3g − 2, then (162) gives
1 (3g − 5)(6g − 5)
λg,3g−3 = (3g − 6)λg−1,3g−6 − λg−1,3g−5
2 2(3g − 3)
(3g − 4)(3g − 2) 1 X X
g−2
+ λg−1,3g−4 + l1 l2 λr,l1 λg−r,l2
2(3g − 3) 2(3g − 3) r=2
l1 +l2 =3g−6
X X 
− 2 l1 l2 λr,l1 λg−r,l2 + l1 l2 λr,l1 λg−r,l2 .
l1 +l2 =3g−5 l1 +l2 =3g−4

Then set l = 3g − 3, (162) gives


(3g − 4)λg,3g−4 + (2 − 2g)λg,3g−3
1 1
= (3g − 7)(3g − 4)λg−1,3g−7 − (3g − 6)(6g − 7)λg−1,3g−6
2 2
g−2
1 1X X
+ (3g − 5)lλg−1,3g−5 + l1 l2 λr,l1 λg−r,l2
2 2 r=2
l1 +l2 =3g−7
X X 
− 2 l1 l2 λr,l1 λg−r,l2 + l1 l2 λr,l1 λg−r,l2 ,
l1 +l2 =3g−6 l1 +l2 =3g−5

from which we can solve λg,3g−4 using λg,3g−3 and lower genus λr,n . Similarly λg,n
can be solved recursively from n = 3g − 3 to n = g − 1 using (162).

Acknowledgements. The second author is partly supported by NSFC grant


11661131005.
HOLOMORPHIC ANOMALY EQUATIONS AND QUANTUM SPECTRAL CURVES 43

A. Some explicit expressions for Fbg,n


In this appendix, we give more examples of the abstract free energies.
For the case N = 1, here are the explicit expressions for Fb2,2 and Fb3 in terms of
Feynman graphs.

Fb2,2 = 1
2 2 + 14 1 + 12 1 1 + 12 2 0 + 12 1 1 1
+ 16 1

+ 41 1 0 + 21 0 1 + 21 0 1 + 41 0 1 + 41 1 0

+ 14 0 1 + 12 1 1 0 + 1 0 1 + 14 1 0 1
1 + 12 0 0

+ 18 0 0 + 18 0 0 + 18 0 0 + 18 0 0 + 18 0 0

0
+ 14 0 0 + 14 0 1 0 + 12 1 + 12 1 0 0
0

+ 12 0 0 1 + 14 1 0 0 + 12 0 0 1 + 14 1 0 0

1
+ 14 0 0 1 + 14 1 0 0 + 12 1 0 0 + 14 0 0
1

+ 12 1 0 0 1 + 18 0 0 0 + 18 0 0 0 + 14 0 0 0

+ 18 0 0 0 1
+ 16 0 0 0 + 14 0 0 0 1
+ 12 0 0 0

0 0
+ 14 0 + 14 0 0 0 + 18 0 + 18 0 0 0
0 0

0 0
+ 12 0 + 12 0 0 0 1 + 14 1 0 0 0 + 12 1 0
0 0

1 0 0
+ 21 1 0 0 0 + 41 0 0 1
+ 16 0 0 + 41 0 0
0 0 0

0 0 0 0
+ 18 0 0 0 0 + 14 + 18 0 0 + 14 0 0
0 0 0 0

+ 14 0 0 0 0 + 81 0 0 0 0 + 18 0 0 0 0
.
44 ZHIYUAN WANG AND JIAN ZHOU

1
Fb3 = 3 + 2 2 + 2 1 + 81 1 + 41 1 1 + 21 2 0 + 21 1 1

+ 12 1 1 1
1 + 48 0 + 14 1 0 + 16 1 0 + 14 1 0

1
+ 18 1 0 + 12 1 1 0 + 12 1 0 1 + 14 1 0 1 + 16 1 0
1

1 1 1 1
+ 48 0 0 + 16 0 0 + 16 0 0 + 12 0 0 + 81 0 1 0

0
+ 14 1 + 41 1 0 0 + 14 1 0 0 + 14 1 0 0 + 16 1 0 0
0

1
+ 14 0 0 + 14 1 0 0 1 + 18 0 0 0 1
+ 16 0 0 0
1

1
0 0
+ 12 0 0 0 + 18 0 + 18 0 + 14 1 0 0 0
0 0

0 0 0 0
+ 14 1 0 + 18 1 0 1
+ 48 0 0 1
+ 24 0 0
0 0 0 0

1
0 0 1
0
+ 16 + 16 0 0 0 0 + 18 0 0
0 0 0.

And in terms of the derivatives of the holomorphic free energy F (t) together
with the propagator κ, the explicit expressions for Fb3 and Fb4 are as follows.

1
Fb3 = F3 + ( F2′′ + F1′ F2′ )κ
2
1 (4) 1 1 1 1
+[ F1 + (F1′′ )2 + F0′′′ F2′ + F1′ F1′′′ + (F1′ )2 F1′′ ]κ2
8 4 2 2 2
1 (6) 1 (4) 5 1 (5)
+[ F0 + F0 F1′′ + F0′′′ F1′′′ + F0 F1′ + F0′′′ F1′ F1′′
48 4 12 8
1 (4) ′ 2 1 ′′′ ′ 3 3
+ F0 (F1 ) + F0 (F1 ) ]κ
4 6
1 (4) 2 7 ′′′ (5) 5 ′′′ 2 ′′ 2 ′′′ (4) ′ 1 ′′′ 2 ′ 2 4
+[ (F0 ) + F0 F0 + (F0 ) F1 + F0 F0 F1 + (F0 ) (F1 ) ]κ
12 48 8 3 2
25 ′′′ 2 (4) 5 ′′′ 3 ′ 5 5 ′′′ 4 6
+[ (F0 ) F0 + (F0 ) F1 ]κ + (F0 ) κ .
48 8 16
HOLOMORPHIC ANOMALY EQUATIONS AND QUANTUM SPECTRAL CURVES 45

1 1
Fb4 = F4 + [ (F2′ )2 + F1′ F3′ + F3′′ ]κ
2 2
1 1 1 1 1
+[F1′ F1′′ F2′ + (F1′ )2 F2′′ + F1′′ F2′′ + F0′′′ F3′ + F1′′′ F2′ + F1′ F2′′′
2 2 2 2 2
1 (4)
+ F2 ]κ2
8
1 1 1
+[ (F1′ )2 (F1′′ )2 + (F1′′ )3 + F0′′′ (F1′ )2 F2′ + F0′′′ F1′′ F2′ + F0′′′ F1′ F2′′
2 6 2
1 ′ 3 ′′′ 5 5 1 (4)
+ (F1 ) F1 + F1 F1 F1 + (F1′′′ )2 + F0′′′ F2′′′ + F0 F1′ F2′
′ ′′ ′′′
6 24 12 2
1 (4) ′′ 1 ′ 2 (4) 1 ′′ (4) 1 (5) ′ 1 (5) ′ 1 (6)
+ F0 F2 + (F1 ) F1 + F1 F1 + F0 F2 + F1 F1 + F1 ]κ3
4 4 4 8 8 48
1 3 5
+[ F0′′′ (F1′ )3 F1′′ + F0′′′ F1′ (F1′′ )2 + (F0′′′ )2 F1′ F2′ + (F0′′′ )2 F2′′ + F0′′′ (F1′ )2 F1′′′
2 2 8
5 1 (4) 3 (4) 3 (4)
+ F0′′′ F1′′ F1′′′ + F0 (F1′ )4 + F0 (F1′ )2 F1′′ + F0 (F1′′ )2
4 24 4 8
2 ′′′ (4) ′ 2 (4) ′ ′′′ 2 ′′′ ′ (4) 1 (4) (4) 1 (5)
+ F0 F0 F2 + F0 F1 F1 + F0 F1 F1 + F0 F1 + F0 (F1′ )3
3 3 3 6 12
3 (5) 7 (5) 7 (5) 1 (6) 1 (6)
+ F0 F1′ F1′′ + F0 F1′′′ + F0′′′ F1 + F0 (F1′ )2 + F0 F1′′
8 48 48 16 16
1 (7) 1 (8) 4
+ F0 F1′ + F ]κ
48 384 0
1 5 5
+[ (F0′′′ )2 (F1 )4 + 2(F0′′′ )2 (F1′ )2 F1′′ + (F0′′′ )2 (F1′′ )2 + (F0′′′ )3 F2′
8 4 8
15 ′′′ 2 ′ ′′′ 7 ′′′ (4) ′ 3 8 ′′′ (4) ′ ′′ 25 ′′′ (4) ′′′
+ (F0 ) F1 F1 + F0 F0 (F1 ) + F0 F0 F1 F1 + F0 F0 F1
8 12 3 24
1 (4) 2 ′ 2 1 (4) 2 ′′ 25 ′′′ 2 (4) 25 ′′′ (5) ′ 2
+ (F0 ) (F1 ) + (F0 ) F1 + (F0 ) F1 + F0 F0 (F1 )
3 3 48 48
7 ′′′ (5) ′′ 5 (4) (5) ′ 21 (5) 2 5 ′′′ (6) ′
+ F0 F0 F1 + F0 F0 F1 + (F ) + F0 F0 F1
12 16 640 0 24
5 (4) (6) 1 ′′′ (7) 5
+ F0 F0 + F0 F0 ]κ
96 32
2 ′′′ 3 ′ 3 25 ′′′ 3 ′ ′′ 5 ′′′ 3 ′′′ 109 ′′′ 2 (4) ′ 2
+[ (F0 ) (F1 ) + (F0 ) F1 F1 + (F0 ) F1 + (F ) F0 (F1 )
3 8 4 48 0
125 ′′′ 2 (4) ′′ 11 ′′′ (4) 2 ′ 11 (4) 3 53 ′′′ 2 (5) ′
+ (F ) F0 F1 + F0 (F0 ) F1 + (F0 ) + (F0 ) F0 F1
48 0 8 96 48
7 ′′′ (4) (5) 113 ′′′ 2 (6) 6
+ F0 F0 F0 + (F ) F0 ]κ
12 576 0
25 15 185 ′′′ 3 (4) ′ 445 ′′′ 2 (4) 2
+[ (F0′′′ )4 (F1′ )2 + (F0′′′ )4 F1′′ + (F ) F0 F1 + (F ) (F0 )
16 8 48 0 288 0
161 ′′′ 3 (5) 7
+ (F ) F0 ]κ
192 0
15 985 ′′′ 4 (4) 8 1105 ′′′ 6 9
+[ (F0′′′ )5 F1′ + (F ) F0 ]κ + (F ) κ .
8 384 0 1152 0
(k+1)
In particular, by setting Fg = 0 for all g ≥ 1 and F0 = Ik , we recover the
expressions (103) and (107) for F31D and F41D .
46 ZHIYUAN WANG AND JIAN ZHOU

References
[1] Aganagic M, Bouchard V, Klemm A. Topological Strings and (Almost) Modular Forms. Com-
munications in Mathematical Physics, 2008, 277(3):771-819.
[2] Bershadsky M, Cecotti S, Ooguri H, Vafa C. Holomorphic anomalies in topological field theo-
ries. arXiv preprint hep-th/9302103, 1993.
[3] Bershadsky M, Cecotti S, Ooguri H, Vafa C. Kodaira-Spencer theory of gravity and exact
results for quantum string amplitudes. Communications in Mathematical Physics, 1994, 165(2):
311-427.
[4] Candelas P, de la Ossa X C, Green P S, Parkes L. A pair of Calabi-Yau manifolds as an exactly
soluble superconformal theory. Nuclear Physics B, 1991, 359(1): 21-74.
[5] Cecotti S, Vafa C. Topologicalanti-topological fusion. Nuclear Physics B, 1991, 367(2):359-461.
[6] Deligne P, Mumford D. The irreducibility of the space of curves of given genus. Publications
Mathmatiques de l’Institut des Hautes tudes Scientifiques, 1969, 36(1): 75-109.
[7] Dijkgraaf R, Verlinde H, Verlinde E. Topological strings in d¡ 1. Nuclear Physics B, 1991,
352(1): 59-86.
[8] Dubrovin B. Geometry and integrability of topological-antitopological fusion. Communications
in Mathematical Physics, 1993, 152(3):539-564.
[9] Dubrovin B, Yang D, Zagier D. Classical Hurwitz numbers and related combinatorics. Moscow
Mathematical Journal, 2017, 17(4): 601-633.
[10] Eguchi T, Hori K, Xiong C S. Quantum cohomology and Virasoro algebra. arXiv preprint
hep-th/9703086, 1997.
[11] Eynard B. Large N expansion of convergent matrix integrals, holomorphic anomalies, and
background independence. Journal of High Energy Physics, 2009, 2009(03): 003.
[12] Eynard B, Orantin N. Invariants of algebraic curves and topological expansion. Communica-
tions in Number Theory & Physics, 2007, 1(2):347-452.
[13] Eynard B, Mariño M, Orantin N. Holomorphic anomaly and matrix models. Journal of High
Energy Physics, 2007, 2007(06): 058.
[14] Getzler E. The Virasoro conjecture for Gromov-Witten invariants. Contemporary Mathemat-
ics, 1999, 241: 147-176.
[15] Grimm T W, Klemm A, Mariño M, Weiss M. Direct integration of the topological string.
Journal of High Energy Physics, 2007, 2007(08): 058.
[16] Gukov S, Sulkowski P. A-polynomial, B-model, and quantization. Journal of High Energy
Physics, 2012, 2012(2):70.
[17] Knudsen F F. The projectivity of the moduli space of stable curves, II: The stacks Mg,n .
Mathematica Scandinavica, 1983, 52(2): 161-199.
[18] Kontsevich M. Intersection theory on the moduli space of curves and the matrix Airy function.
Communications in Mathematical Physics, 1992, 147(1): 1-23.
[19] Mulase M, Sulkowski P. Spectral curves and the Schroedinger equations for the Eynard-
Orantin recursion. arXiv preprint arXiv:1210.3006, 2012.
[20] Nishigaki S, Yoneya T. A nonperturbative theory of randomly branching chains. Nuclear
Physics B, 1991, 348(3): 787-807.
[21] Penner R C. Perturbative series and the moduli space of Riemann surfaces. Journal of Dif-
ferential Geometry, 1988, 27(1): 35-53.
[22] Wang Z., Zhou J., in preparation.
[23] Witten E. Quantum Background Independence In String Theory. arXiv preprint
arXiv:hep-th/9306122.
[24] Witten E. Two-dimensional gravity and intersection theory on moduli space. Surveys in
differential geometry, 1990, 1(1): 243-310.
[25] Yamaguchi S, Yau S T. Topological string partition functions as polynomials. Journal of High
Energy Physics, 2004, 2004(07): 047.
[26] Zhou J. Intersection numbers on Deligne-Mumford moduli spaces and quantum Airy curve.
arXiv preprint arXiv:1206.5896, 2012.
[27] Zhou J. On topological 1D gravity. I. arXiv preprint arXiv:1412.1604, 2014.
[28] Zhou J. Topological Recursions of Eynard-Orantin Type for Intersection Numbers on Moduli
Spaces of Curves. Letters in Mathematical Physics, 2013, 103(11):1191-1206.
HOLOMORPHIC ANOMALY EQUATIONS AND QUANTUM SPECTRAL CURVES 47

Department of Mathematical Sciences, Tsinghua University, Beijing, 100084, China


E-mail address: zhiyuan-14@mails.tsinghua.edu.cn

Department of Mathematical Sciences, Tsinghua University, Beijing, 100084, China


E-mail address: jzhou@math.tsinghua.edu.cn