Patrick Lei

All notes

Higher genus GW theory

Simons Center, Jan 2025.

NOTES ON THE HIGHER GENUS GROMOV-WITTEN THEORY OF CALABI-YAU THREEFOLDS

[Uncaptioned image]

Lectures in January 2025 at the Simons Center for Geometry and Physics
Notes by Patrick Lei
  WHAT’S A PUBLISHER?

Preface

These notes, taken during the first two weeks of the program Recent developments in higher genus curve counting at the Simons Center for Geometry and Physics, explain how to prove important structural results in the higher-genus Gromov-Witten theory of compact Calabi-Yau threefolds.

One of the oldest problems in Gromov-Witten theory is to compute the Gromov-Witten invariants of compact Calabi-Yau threefolds. While there are a variety of physical methods which have allowed physicists to make spectacular predictions in a wide range of examples, mathematical progress has often been frustratingly slow due to a lack of satisfactory tools to attack the problem with. To illustrate this, we will briefly outline some developments in mathematics regarding Calabi-Yau threefolds with h2=1 which arise as complete intersections in weighted projective space.

For complete intersections in projective space, a genus-zero mirror theorem was proved by Givental in 1996 and Lian-Liu-Yau in 1997 [Giv96; LLY97]. For complete intersections in weighted projective space, a genus-zero mirror theorem was proved by Coates-Corti-Lee-Tseng in 2006 in the convex case and by Jun Wang in 2019 in the non-convex case [CCLT09; Wan20]. The main difficulty in the non-convex case is the failure of the quantum Lefschetz theorem [CG07], which is the main tool used in [Giv96; LLY97; CCLT09].

Meanwhile, a genus-one mirror theorem was proved for the quintic by Zinger in 2007 and for complete intersections in projective space by Popa in 2010 [Zin09; Pop13]. This used the theory of reduced invariants developed by Vakil-Zinger and Li-Zinger [VZ08; LZ09], which is a technique of performing birational modifications to the main component of the moduli space of stable maps to force the quantum Lefschetz theorem to hold so that computations can be performed using virtual localization [GP99]. Unfortunately, to this date no computations have been successfully performed using reduced invariants since.

Another approach is to consider the theory of GLSMs [Wit92], which were constructed mathematically by various authors [FJR18; FK21] during the 2010s. Unfortunately, the virtual cycle is supported on the moduli space of stable maps to the threefold, which does not carry a torus action. For the quintic, the ambient space of the relevant GLSM is the total space of 𝒪ℙ4⁢(−5). A standard way to gain a torus action is to compactify the moduli space at infinity and consider relative (or logarithmic) invariants. This leads to the theory of logarithmic GLSMs, which was introduced by Chen-Janda-Ruan in 2019 [CJR21]. Before this, Guo-Janda-Ruan used this theory [CJR23], including still conjectural foundational results [CJRa; CJRb], to prove a genus-two mirror theorem for the quintic in 2017 (up to the fact that the moduli space used in their 2017 paper is still not defined) and to prove the Yamaguchi-Yau finite generation conjecture, holomorphic anomaly equations, and orbifold regularity (as well as the conifold gap condition in low genus) for the quintic [GJR17; GJR18].

Because a GLSM is an enumerative theory of a critical locus in a GIT quotient, it depends on a stability parameter. We can vary the stability parameter and construct a master theory following the master space construction of Thaddeus [Tha96]. This leads to the theory of Mixed-Spin-P fields, which was first constructed by Chang-Li-Li-Liu in 2015 (for the quintic only) [CLLL19; CLLL22]. Unfortunately, the original theory is impossible to compute with when g≥2, so a parameter N was introduced by Chang-Guo-Li-Li in 2018 [CGLL21]. In the same year, Chang-Guo-Li proved the Yamaguchi-Yau finite generation conjecture, BCOV Feynman rule, and holomorphic anomaly equations for the quintic threefold [CGL21; CGL19]. These results were generalized to hypersurfaces in weighted projective space by the author in 2024 [Lei24b; Lei24a], where a genus-one mirror theorem was also proved.

The Castelnuovo bound was proved in 2022 by Liu-Ruan [LR22] for any Calabi-Yau threefold satisfying a conjectural Bogomolov-Gieseker-type inequality (including the quintic) due to Bayer-Macri-Toda [BMT14] and a weaker version of the result was proved by Zhiyu Liu in 2024 [Liu24] without assuming the conjectural inequality. These results were obtained using the GW/DT correspondence [MNOP06a; MNOP06b], which was proved for complete intersections in products of projective spaces by Pandharipande-Pixton in 2012 [PP17] and for all Calabi-Yau threefolds by Pardon in 2023 [Par24].

As we can see, two major breakthroughs were made around the year 2018, when mathematicians discovered ways which make proving results about Gromov-Witten invariants in arbitrary genus tractable.1 These notes explain how to use these new ideas to prove the most important conjectures which are provided to us by physicists, namely the Yamaguchi-Yau finite generation conjecture and the holomorphic anomaly equations. Chapter 1 covers the B-model topological string and the Givental formalism, Chapter 2 covers log GLSMs, and Chapter 3 covers Mixed-Spin-P fields. We will cover both the foundational theory and the calculations in both approaches – in particular, Section 2.1 and Section 3.1 discuss the geometry of the log GLSM and MSP moduli spaces while Section 2.2 and Section 3.2 discuss calculations.

Author’s note

In contrast to the genus zero situation, not much is known about higher-genus Gromov-Witten theory. This is in part because there are significantly fewer tools to study higher-genus invariants, and the ones which do exist seem to be considered extremely inaccessible. It is my sincere hope that these notes can help make the subject more accessible.

All errors are the sole responsibility of the author. Please email me if you find any mistakes or typos in these notes.

The lectures were divided as in Table 1. The notes from the lectures by Albrecht Klemm were taken partly during the lecture and partly by watching the videos on the SCGP website, the notes from my lectures are slightly expanded versions of my lecture notes, and the notes from all other lectures were taken during the lecture, with some material being provided by the speakers after their lectures. All references were added after the lectures.

Table 1: Schedule of lectures
Date Time Subject Speaker(s)
1/6/25 9:30 Log GLSM foundations I Qile Chen
1/6/25 11:30 Givental formalism I Patrick Lei
1/7/25 9:30 Log GLSM foundations II Qile Chen
1/7/25 11:30 Givental formalism II Patrick Lei
1/8/25 9:30 Log GLSM foundations III Qile Chen
1/9/25 9:30 Physics I Albrecht Klemm
1/9/25 11:30 Log GLSM virtual localization Felix Janda
1/10/25 9:30 Physics II Albrecht Klemm
1/13/25 9:30 MSP moduli space: quintic Wei-Ping Li
1/13/25 11:30 Genus two and LG/CY via log GLSM Shuai Guo, Felix Janda
1/14/25 9:30 Polynomiality via MSP Patrick Lei
1/14/25 11:30 GW theory of the quintic via log GLSM Shuai Guo, Felix Janda
1/15/25 9:30 Ω-stability I Yang Zhou
1/16/25 9:30 Ω-stability II Yang Zhou
1/17/25 9:30 BCOV Feynman rule via MSP Shuai Guo

Acknowledgements

I would like to thank Qile Chen, Felix Janda, Sheldon Katz, Melissa Liu, John Pardon, and Rachel Webb for organizing the program. I would also like to thank Konstantin Aleshkin, Shuai Guo, Melissa Liu, and Dimitri Zvonkine for helpful discussions regarding the material in my lectures. Further thanks goes to Qile Chen for generously providing the notes from his lectures, Felix Janda for pointing out some subtleties and correcting some errors related to the material in his lectures, Sheldon Katz for pointing out that the Kähler structure bundle is the dual of the Hodge bundle, Yang Zhou for making detailed suggestions, and the Simons Center for posting videos of all of the lectures.

1 Mathematical and physical preliminaries

We begin these notes with discussion of the topological B-model. Using mirror symmetry, physicists are able to transport questions in Gromov-Witten theory (the A-model) to the B-model, which lacks a satisfactory mathematical definition. The benefit of doing this is that it is easier to calculate in the B-model, where genus-zero invariants are simply period integrals. Using other B-model techniques,1 physicists are able to make far-reaching predictions about the Gromov-Witten invariants of Calabi-Yau threefolds. Note that there are proposed mathematical definitions of the B-model – the analytic approach of Costello-Li [CL12b; Li11] and the categorical enumerative invariants of Caldararu-Tu [CT24] – but they are extremely difficult to compute with.

We then proceed to a discussion of the mathematical prerequisites for the calculations which appear later in these notes. The most important tool for performing calculations in Gromov-Witten theory is the formalism introduced by Givental in a highly influential series of papers [Giv01b; Giv01a; Giv04] in the early 2000s. This formalism allows us to package large formulae, for example coming from virtual localization, in a compact way, which allows us to reason about higher-genus invariants in a relatively streamlined way. For example, polynomiality of GW generating series will be a direct corollary of polynomiality of the entries of an R-matrix which plays a central role in Givental’s formalism.

1.1 Introduction to the topological B-model

1.1.1 Mirror symmetry and the role of Calabi-Yau threefolds

Let X⁡(Ω,ω) be a Calabi-Yau n-fold, where here Ω is a holomorphic n-form and ω is a Kähler form. Recall that this is equivalent to having S⁢U⁢(n) holonomy or to KX=0. By a result of Yau [Yau78], there exists a Kähler-Einstein metric g in the class of ω with vanishing Ricci curvature. For our purposes, we will mostly consider the case when n=3.

We will really consider families of Calabi-Yau varieties

where ℳ⁡(z¯,t¯) is parametrized by a combination of complex structure moduli z¯ and Kähler structure moduli t¯ℝ complexefied by a Neveu-Schwarz harmonic 2-form B to t¯. We will denote the fiber by

On the moduli of complex structures ℳ⁡(z¯), the tangent space H(0,1)⁢(X,TX) has a basis given by harmonic forms

for k=1,…,hn−1,1⁢(X). Because we are in the Calabi-Yau setting, contracting with Ω gives us a basis

of Hn−1,1⁢(X). By a theorem of Tian-Todorov [Tia87; Tod89], the moduli space is unobstructed. The B-model is built from mathematical structures on ℳ⁡(z¯).

There is a moduli space ℳg2 parameterized by δ⁢g subject to the condition

To first order, we have

The indices with pure Hodge type correspond to harmonic forms in H(0,1)⁢(X,T⁢X) with components

yielding the Kuranishi family over ℳ⁡(z¯). The mixed indices correspond to real harmonic (1,1)-forms, and expanding the Kähler form linearly we obtain

in terms of real Kähler parameters

which are volumes of holomorphic curves.

The Kähler moduli space ℳ⁡(t¯ℝ) is the real Kähler cone subject to positivity conditions from integration over k-dimensional holomorphic submanifolds, namely

The bosonic part of the string action contains the harmonic antisymmetric Neveu-Schwarz background field bi⁢ȷ¯

where α′ is the string coupling constant. Its critical values measure complexified volumes of holomorphic curves by

Conjecture 1.1.1.

For non-rigid Calabi-Yau threefolds X with h2,1>0, there exists a mirror Calabi-Yau X^ with h1,1⁢(X^)=h2,1⁢(x) and h2,1⁢(X^)=h1,1⁢(X) such that the moduli spaces satisfy

and all relevant physical and mathematical structures can be identified using locally invertible mirror maps

Mirror symmetry identifies Type IIA compactifications on X with Type IIB compactifications on X^, and vice versa. Additional Ramond-Ramond background fields and axio-dilaton fields with modulus j¯ and a extend the moduli spaces as

and

where 𝒬 denotes a quaternionic extension of ℳ. The RR (k+1)-form fields are sourced from Dk branes. k is even for Type IIA and odd for Type IIB. The D2⁢m correspond to coherent sheaves and D2⁢m+1 correspond to special Lagrangian branes. Here, 𝒬 is the 4d N=2 hyper-multiplet moduli space, and ℳ is the 4d N=2 vector multiplet moduli space.

One kind of mathematical structure is Hodge numbers. In the traditional picture of the Hodge diamond, Poincaré duality corresponds to a reflection along the horizontal axis, Dolbeaut symmetry is a reflection along the vertical axis, and mirror symmetry is a reflection along a line with slope 1.

Superstring theory is defined by maps

weighted by an action S which is a supersymmetric extension of the area of 𝒞β. It is easy to quantize the Neveu-Schwarz-Ramond action, and the Green-Schwarz action incorporates the RR fields. The first quantized theory is defined by a variatonal integral with partition function

Superstring theory is Weyl invariant in ten dimensions, or in other words

so we obtain a discrete sum of finite-dimensional integrals. This implies that the compact part X of the spacetime M must be a complex threefold. If X is Calabi-Yau, we obtain an extended (2,2) world-sheet SCFT, which has four nilpotent operators

The A-twist corresponds to taking

and the B-twist corresponds to

The topological A-model yields a cohomological topological theory depending only on the Kähler structure, while the topological B-model is a homological topological theory depending only on the complex structure. Mirror symmetry then exchanges the A-model and B-model.

1.1.2 The topological A-model and B-model

In the A-model, terms depending on the complex structure are QA-exact, so the variatinoal integral simplifies to

Here, we have

and these holomorphic maps are stationary points of the action. Moreover, taking the logarithm, we obtain

where rgβ are the GW invariants. Rewriting these in terms of GV invariants, we obtain

In the B-model, the terms depending on the Kähler structure are QB-exact and the variational integral localizes to constant maps albeit with a nontrivial measure depending on the complex structure. Mirror symmetry is supposed to be an exact duality, so we should have

Period integrals

define a nondegenerate pairing between middle homology and cohomology by Stokes’ theorem. This pairing is antisymmetric if n is odd and symmetric if n is even. For example, if X is a K3 surface, then the lattice H2⁢(X,ℤ) is

If n is odd, we can fix an integral symplectic basis Γ¯={Aℓ,Bℓ}, which is defined only up to the action of Sp⁡(bn⁢(X),ℤ).

Example 1.1.2.

If we consider an elliptic curve p3=w⁢y2−x⁢(x−w)⁢(x−w⁢z)=0⊂ℙ2, then we obtain

and

Then the integrals

are elliptic integrals. The periods are annihilated by the Picard-Fuchs operator, and by definition thus satisfy the equation

The main constraints which govern the periods of a Calabi-Yau n-fold are the Riemann bilinear relations

This defines the exponential of the Kähler potential K⁡(z) for the Weil-Petersson metric

on ℳ⁡[X]⁢(z¯). There are also holomorphic bilinear relations

which follow from Griffiths transversality [Gri68a; Gri68b]. Here, the integrand in the left hand side are arbitrary combinations of derivatives of Ω with respect to the zi. We will see later that these give rise to propagators, the holomorphic anomaly equations, and other structures. The Cℓn⁢(z) are rational functions labelled by ℓn up to permutations.They are also determined by differential ideals ℒ⁢Π→ also determine the Cℓn⁢(z) up to normalization.

Remark 1.1.3.

In terms of the periods Π→, if we write them in a basis compatible with the intersection form Σ, the quantities in the relations may be written as

1.1.3 The quintic

Consider the mirror quintic W, which is given by the equation

The period vector Π⁡(z) (with an appropriate choice of integral cycles) fulfills the Picard-Fuchs equation

where θ=z⁢ddz.

We want to find a basis where the monodromies around the singular points are integral symplectic matrices, so we look at the Riemann symbol and see that it is given by

Here, z=0 is a point of maximal unipotent monodromy, z=∞ is the orbifold (or Landau/Ginzburg, or Gepner) point, and z=5−5 is the conifold point. At a point of maximal unipotent monodromy, we can expand the mirror map and go to the large volume limit point in the A-model.

Using special geometry, Bryant and Griffiths [BG83] showed that the periods may actually be expressed using a prepotential ℱ as

These correspond to triple logarithmic, double logarithmic, analytic, and logarithmic solutions, respectively. Using this, we can make the identification

where the mirror map is given by

This was generalized to multi-parameter models by Hosono et. al [HKTY95], who related the classical terms to CTC Wall data κ=D3 and σ=κmod22 in the formula

Later, we will use this to find the integral symplectic basis without calculating any monodromy.

We now turn our attention to calculating the numbers ngβ. We already saw the genus-zero invariants, but we are also interested in the higher-genus ones. Some invariants in low genus and degree are given in Table 1.1.

Table 1.1: Low genus GV invariants of the quintic.
g d=1 d=2 d=3 d=4 d=5
0 2875 609250 317206375 242467530000 22930588887625
1 0 0 609250 3721431625 12129909700200
2 0 0 0 534750 75478987900
3 0 0 0 8625 −15663750
4 0 0 0 0 49250
5 0 0 0 0 1100
6 0 0 0 0 10

Predictions were made by Candelas et al. [COGP92] for g=0, Bershadsky et al. [BCOV94] for g≤3, and Huang et al. [HKQ09] for g≤53. The numbers n11, n12, and n13 were calculated mathematically by Schubert, Katz [Kat86], and Ellingsrud-Strømme [ES96], respectively. Kontsevich [Kon95] gave a mathematical proof of some numbers in g=0 before the genus-zero invariants were completely determined by Givental and Lian-Liu-Yau [Giv96; LLY97]. The formula for the genus-one invariants was proved by Zinger [Zin09].

1.1.4 Fourteen one-parameter hypergeometric families

There are fourteen hypergeometric Picard-Fuchs operators for one-parameter families of Calabi-Yau threefolds X^ given by

where the values of μ and ak are given in Table 1.2.

Table 1.2: Data of one-parameter hypergeometric families
N ak μ−1 Mirror X κ c2⋅D χ⁡(X)
8 12,12,12,12 28 X2,2,2,2⁢(18) 16 64 −128
9 14,13,23,34 26⁢33 X4,3⁢(15⁢21) 6 48 −156
16 14,12,12,34 210 X4,2⁢(16) 8 56 −176
25 15,25,35,45 55 X5⁢(14) 5 50 −200
27 13,13,23,23 36 X3,3⁢(16) 9 54 −144
32 14,14,34,34 212 X4,4⁢(14⁢22) 4 40 −144
36 13,12,12,23 24⁢33 X3,2,2⁢(17) 12 60 −144
72 16,12,12,56 28⁢33 X6,2⁢(15⁢31) 4 54 −256
108 16,13,23,56 24⁢36 X6⁢(14⁢21) 3 42 −204
128 18,38,58,78 216 X8⁢(14⁢41) 2 44 −296
144 16,14,34,56 210⁢33 X6,4⁢(13⁢22⁢31) 2 32 −158
200 110,310,710,910 28⁢55 X10⁢(13⁢21⁢51) 1 34 −288
216 16,16,56,56 28⁢36 X6,6⁢(12⁢22⁢32) 1 22 −120
864 112,512,712,1112 212⁢36 X12,2⁢(14⁢41⁢61) 1 46 −484

Here, the notation for the mirror X means the complete intersection in the weighted projective space with weights given in parentheses with degrees given by the subscripts. The local exponents are all 0 at the MUM point, 0,1,1,2 at the conifold point μ, and are given by a1,a2,a3,a4 at the orbifold point ∞. The monodromy around a singular point ∗ is captured by the minimal exponent3 1≤k<∞ such that

for some 0≤p≤dimℂX. When k>1 and p=0, then we have an orbifold point. When p>0 we have an infinite shift monodromy. If p=1 and the local exponents take the form a,b,b,c, then we have a single vanishing period dual to a logarithmic degenerating period and hence a conifold point, and if the local exponents are take the form a,a,b,b, then we have two vanishing periods and two logarithmic degenerating periods, and this case is called a K-point. The case p=2 cannot occur by Schmid’s SL⁡(2,ℂ) orbit theorem, and when p=3, we have a point of maximal unipotent monodromy with local exponents a,a,a,a.

At the point z=0, there is a Frobenius basis of solutions

where we normalize the power series to have f0⁢(0)=1 and fk>0⁢(0)=0. Therefore, we can find an integral basis

In the integral basis, all monodromies are integral symplectic matrices. At the point z=0, we have

and at the conifold point, we have

At z=∞, we have M∞=(M0⁢Mμ)−1.

The work of Alexandrov-Feyzbakhsh-Klemm-Pioline-Schimannek [Ale+24] solves the topological string on these models up to g𝗆𝗈𝖽 given in Table 1.3 using D4 brane and wall-crossing.4 Note that g𝖺𝗏𝖺𝗂𝗅 is the maximum genus for which data appears at http://www.th.physik.uni-bonn.de/groups/klemm/data.php and g𝗂𝗇𝗍𝖾𝗀 is the maximum genus solved by the work of Huang-Klemm-Quackenbush [HKQ09].

Table 1.3: Current state of the art in physics literature as of January 2023.
X χD n1p n1c Type g𝗂𝗇𝗍𝖾𝗀 g𝗆𝗈𝖽 g𝖺𝗏𝖺𝗂𝗅
X5⁢(15) 5 7 0 O 53 69 64
X6⁢(14⁢21) 4 4 0 O 48 66 48
X8⁢(14⁢41) 4 4 0 O 60 84 64
X10⁢(13⁢21⁢51) 2 7 0 O 50 70 68
X4,3⁢(15⁢21) 5 9 0 O 20 24 24
X6,4⁢(13⁢22⁢31) 3 3 0 O 14 17 17
X3,3⁢(16) 6 14 1 K 29 33 33
X4,4⁢(14⁢21) 4 6 1 K 26 34 34
X6,6⁢(12⁢22⁢32) 2 1 0 K 18 21 21
X6,2⁢(15⁢31) 5 7 0 C 63 84 49
X4,2⁢(16) 6 15 1 C 50 64 40
X3,2,2⁢(17) 7 21 1 C 14 ? 14
X2,2,2,2⁢(18) 8 33 3 M 17 ? 32
Remark 1.1.4.

Even inserting at least one D4 brane requires going beyond the topological string, which only gives the usual DT or PT invariants (one D6 brane, no D4 branes, and arbitrary D2 and D0 brane charges). Hopefully, we will be able to explore the entire space of Bridgeland stability conditions in the future.

1.1.5 More on periods

Consider the Picard-Fuchs operator

of a one-parameter Calabi-Yau n-fold with middle Hodge structure of type 1,1,…,1. Defining the adjoint operator

then Griffiths transversality implies that the operator is essentially self-adjoint, or in other words

Here, the Yukawa coupling Cz¯ satisfies the differential equation

For the quintic, we in fact obtain

Equivalently, self-adjointness implies

When n=3, this yields

In the multi-parameter case, the periods Π→ span the kernel of the Picard-Fuchs differential ideal

generated by the differential operators

This system is complete if all three-point functions

can be integrated from the Griffiths transversality relations.

Example 1.1.5.

For the example X18⁢(13⁢61⁢91) with middle Hodge structure 1,2,2,1, the Picard-Fuchs equation is generated by

and the Ci⁢j⁢k are given by

where the components of the discriminant are given by

and δ=1−432⁢z1.

If we shift the holomorphic form by

this induces a Kähler gauge transformation

Also note that the Yukawa couplings Ci⁢j⁢k⁢(z¯) are sections of

where ℒ is the dual of the Hodge bundle.

By the local Torelli theorem [Gri68a; Gri68b], the XI components of the period vector

are homogeneous coordinates on ℳ⁡(z¯). Griffiths transversality implies that

is a homogeneous prepotential such that

and the Yukawa couplings are given by

Dividing by X0, we will get inhomogeneous coordinates ti=XiX0, and we obtain the inhomogeneous prepotential

The Kähler potential and Yukawa coupling become

In addition, we have

In inhomogeneous coordinates, we can obtain the period vector from ℱ⁡(t¯) by

The Picard-Fuchs equation is equivalent to the Gauss-Manin connection

where i=1,…,r and Ai⁢(z¯)∈ℚ⁡[z¯]. If we set

this becomes

1.1.6 Special geometry

On the moduli space of complex structures ℳ𝖼𝗌≔ℳ⁡(z¯), we have metric connections from the Weil-Petersson metric and the Kähler line bundle connection. On sections

the covariant derivatives are given by

Using the covariant derivatives, we see that χi∈Hn−1,1⁢(X) and χ¯ı¯∈H1,n−1⁢(X).

Proceeding more systematically, repeated applications of Di yield

Using the relation

we further deduce

It follows that

where Cȷ¯k⁢ℓ=e2⁢K⁢Cȷ¯⁢k¯⁢ℓ¯⁢Gk⁢k¯⁢Gℓ⁢ℓ¯. This equation is known as the special geometry equation and is the integrability condition for the existence of the prepotential ℱ.

1.1.7 Genus one predictions

In a (2,2) theory, the topological torus partition function is defined by

as an integral of the fermion number projected partition function over the fundamental region of the torus. Using its relationship to the family index of Ray-Singer analytic torsion [FLY08], it satisfies the holomorphic anomaly equation

Using Gi⁢ȷ¯=∂ȷ¯∂iK=∂ȷ¯Ki and the special geometry equation, we can integrate the holomorphic anomaly equation to obtain

1.1.8 Propagators

The propagators are an-holomorphic sections Si⁢j, Si, and S of ℒ−2⊗Sym2⁡(T⁢ℳ𝖼𝗌1,0), ℒ−2⊗T⁢ℳ𝖼𝗌1,0, and ℒ−2, respectively, defined by

We can integrate the first equation using the special geometry equation and the observation that the only contributions to Cȷ¯k⁢m⁢Ci⁢ℓ⁢m are derivatives in ∂¯ȷ¯. Therefore, if there exists an index i such that [C(i)]m is invertible, we obtain

where q(i)⁢ℓm is the holomorphic propagator ambiguity.

Following Alim-Länge [AL07], it is convenient to shift the remaining propagators as

Applying the special geometry equation, we obtain

Here, all of the ambiguities are in fact rational functions in z¯ with rational coefficients. This allows us to obtain explicit formulae

Taking the holomorphic limit of the propagators, we obtain

where we take the holomorphic limits e−K→e−𝒦=X∗0, Ki→𝒦i=−∂ilog⁡(X∗0), and Γj⁢ki→Υj⁢ki=∂zi∂t∗a⁢∂2t∗a∂zj⁢⁡∂zk.

Returning to the genus one situation, the holomorphic anomaly equation becomes

where we define Ci=12⁢Sj⁢k⁢Ci⁢j⁢k+fi(1). This can be integrated to obtain Equation 1.1. The holomorphic limit is given by

where the holomorphic ambiguity is given by

1.1.9 Higher-genus predictions

In higher genus, the holomorphic anomaly equation becomes

This can be rewritten as the system of equations

assuming that Si⁢j, Si, S, and K are algebraically independent. Using the shifted propagators, we obtain ∂Fg∂Kj=0 and

Using the method of direct integration, due to Yamaguchi-Yau, Grimm-Klemm-Mariño-Weiss, and Alim-Länge [YY04; GKMW07; AL07], we obtain

Here, we set

Therefore, we can solve for a polynomial Fg⁢(Si⁢j,S~i,S~,z), which is a weighted polynomial of degree 3⁢g−3 with the weights 1,2,3,0. For example, in the one-parameter case, we obtain

1.1.10 Boundary conditions

The most difficult part of computing Fg is the degree-zero part fg⁢(z¯), which is known as the holomorphic or modular ambiguity. In the hypergeometric cases, we have

where the number c∞ depends on the type of singularity at z=∞. At z=0, the boundary conditions come from the asymptotics

The second equality uses the computation of degree-zero GW invariants by Faber-Pandharipande.

In the one-parameter case, the Castelnuovo bound implies that the GV invariants ngd are nonzero only when the genus is strictly less than gmax⁢(d)≤⌊d22⁢κ+d2⌋+1 in general and gmax⁢(d)≤⌊2⁢d23⁢κ+d3⌋+1 when 0<d<κ. This was first observed by Huang-Klemm-Quackenbush [HKQ09] and proved in 2023 by Feyzbakhsh (in the paper [Ale+24] of Alexandrov et. al). Because curves of degree β of genus gmax⁢(β) are smooth, the associated invariants may be obtained by the formula

where ℳβ denotes the deformation space of the curve. For the one-parameter hypergeometric cases, we obtain

The most important boundary condition is the gap condition at the conifold point z=μ, where a 3-cycle with the topology of a lens space vanishes. Here, this says that we have

for all g>1, where t𝖼𝗈𝗇 is the local coordinate at the conifold point given by the ratio of a vanishing and non-vanishing period.

1.2 An axiomatic approach to enumerative geometry

1.2.1 Moduli of curves and CohFTs

We will denote by ℳ¯g,n the moduli space of stable curves of genus g with n marked points. This is nonempty if and only if 2⁢g−2+n>0, in which case it is a Deligne-Mumford stack of dimension 3⁢g−3+n. There is a combinatorial structure to the collection of ℳ¯g,n, which is given by a collection of morphisms.

  • •

    There is the gluing morphism

    which takes two curves and glues them along the last marked point to form a node;

  • •

    There is the self-gluing morphism

    which glues the last two marked points.

There is of course another interesting map, which is the forgetful map

given by deleting the last marked point and then stabilizing. We are now able to define our enumerative theories of interest, which include Gromov-Witten theory.

Definition 1.2.1.

Given a vector space V with a nondegenerate symmetric bilinear form η and a unit element 𝟏∈V, a Cohomological Field Theory (CohFT) [KM94] Ω on V is a collection Ωg,n of Sn-equivariant linear maps

which satisfy the basic identity

and the following combinatorial identities (we will let eμ be a basis of V such that e1=𝟏 and eμ be the dual basis):

If in addition the identity

is satisfied, then we say the CohFT has a flat unit (or satisfies the string equation).

All of this can be generalized to super vector spaces, but for simplicity we will not deal with this case.

Example 1.2.2.

The most important example of a CohFT is the Gromov-Witten theory of a smooth projective variety X. Here, recall that the source of a stable map is a prestable curve, so there is a stabilization morphism

which forgets the map and stabilizes the curve. Then, working over the Novikov ring, we will set V=H∗⁢(X), η to be the Poincaré pairing, and 𝟏 to be the fundamental class of X. The linear maps Ωg,nX are given by the formulae

where evi:ℳ¯g,n⁢(X,β)→X is the i-th evaluation map.

Example 1.2.3.

Let π:𝒞g,n→ℳ¯g,n be the universal curve. Then the Hodge bundle is the vector bundle

We then consider the vector space V=ℂ with the usual pairing and define the CohFT Ω𝔼 by the formula

The gluing axioms are satisfied by the equation q∗⁢𝔼g+h=𝔼g⊕𝔼h and the exact sequence

Here, because genus zero components do not contribute to global sections of ω and 𝔼 does not depend on the marked points, this CohFT has a flat unit.

Remark 1.2.4.

We will see later that this is related to the GW CohFT of a point by the quantum Riemann-Roch theorem [CG07].

Given a CohFT Ω, we may produce invariants by pairing the classes Ωg,n⁢(𝐯) with various cohomology classes on ℳ¯g,n. The most important ones to consider are the classes

where si:ℳ¯g,n→𝒞g,n is the i-th marked point.

Remark 1.2.5.

In Gromov-Witten theory, these are called ancestors. There are also descendants, which are given by the same formula except using the moduli space 𝔐g,n of prestable curves instead of ℳ¯g,n, where we factor the morphism st defined above as

The descendant classes are denoted by ψi.

Example 1.2.6.

We will calculate an invariant which will appear in the GW theory of Calabi-Yau threefolds. For any CohFT Ω, consider the invariant

Using the second gluing equation, the degree zero part of Ω1,1⁢(𝟏) is equal to

which yields the (graded) dimension χ⁡(V) of V. Therefore, we obtain

Example 1.2.7.

Consider the GW CohFT of a point. This is given by V=ℂ with the usual pairing and the formula

Then we define the invariants

Theorem 1.2.8 ([Kon92]).

The function

is annihilated by the operators

Remark 1.2.9.

This result (Virasoro constraints for a point) is equivalent to 𝒟 being a tau-function for the KdV hierarchy and has been generalized by various authors.

1.2.2 The genus-zero picture

A CohFT in genus zero defines a Frobenius manifold [Dub96]. In particular, there is a product structure defined by the formula

and the quantum connection, which is defined by the formula

The structure of a Frobenius manifold comes from a function ℱ0, which is known as the genus-zero descendant potential and satisfies a set of PDEs, which are the string equation, dilaton equation, and an infinite set of topological recursion relations, where we write 𝐯=∑μ,ntnμ⁢eμ⁢zn∈V⁡⟦z⟧:

To state the following result, we consider the infinite-dimensional vector space V⁢⦇z−1⦈ (or a completion of this) with the symplectic form

We also consider the new variable5 𝐪=𝐯−z, so ℱ0 is now a function near 𝐪=−z.

Theorem 1.2.10 ([Giv04]).

ℱ satisfies the above PDEs if and only if the graph ℒ of dℱ is a Lagrangian cone with vertex at 𝐪=0 such that its tangent spaces L are tangent to ℒ exactly along z⁢L⊂L.

We may recover the Lagrangian cone ℒ by the following procedure. We find a fundamental solution

to the quantum connection, which satisfies the equation

Then, setting J≔z⁢S∗⁢(z)⁢𝟏, we can then recover the Lagrangian cone by the following procedure:

  • •

    The derivatives ∂∂tμ⁢J form a basis of L∩V⁢⟦z−1⟧;

  • •

    This implies that z⁢∂2∂tμ⁢⁡∂tν∈L∩V⁡⟦z−1⟧. Writing these in terms of the first derivatives and using the fact that J solves the quantum connection, we recover the Frobenius structure and hence the Lagrangian cone.

In Gromov-Witten theory, there is an explicit formula for the fundamental solution in terms of descendant invariants. It is given by

Example 1.2.11.

Let X be the quintic threefold. We will use the genus-zero mirror theorem of Givental to compute the quantum product on H∗⁢(X). Let

be the (very small) I-function of X. Setting Q⁡(q)≔q⁢eI1⁢(q)I0⁢(q), the mirror theorem states that

Because the mirror map q↦Q⁡(q) corresponds to setting τ=I1⁢(q)I0⁢(q)⁢H by the divisor equation, we can use the mirror theorem to compute ⋆τ. Because of our nonstandard choice of the I-function, the quantum connection becomes the ODE

where D≔q⁢ddq (here, the coordinate is log⁡q) and I11≔1+D⁡(I1⁢(q)I0⁢(q)). An explicit computation using the mirror theorem and the results of Zagier-Zinger [ZZ08] yields

Because Sτ begins with the identity, this computes H⋆τ.

1.2.3 R-matrix action

In the early 2000s, Givental [Giv01a; Giv04] discovered a remarkable property of axiomatic enumerative theories, namely that one can transform CohFTs by the action of matrices R⁡(z)=R0+R1⁢z+R2⁢z2+⋯∈Hom⁡(V,V′)⁢⟦z⟧ which satisfy the property

Traditionally, the literature [PPZ15] requires that V=V′ and R0=Id, but the MSP group [CGL19] has removed this restriction and also allows dimV<dimV′.

In order to define the action of R on a CohFT Ω defined on V, we need to review some of the combinatorial structure of ℳ¯g,n. For a curve C∈ℳ¯g,n⁢(ℂ), we may consider the dual graph of C, which has vertices, edges, and legs, which are defined as follows:

  • •

    The vertices correspond to irreducible components of C and are labelled by a non-negative integer, which is the genus;

  • •

    The edges correspond to nodes of C (in particular, we allow loops);

  • •

    The legs correspond to marked points.

Any graph which appears as the dual graph of a stable curve is called a stable graph.

230αβ312ψψψ2ψ
Figure 1.1: Example of a stable graph in ℳ¯7,3 and associated tautological class. This stable graph describes the image of a map ℳ¯2,4×ℳ¯3,2×ℳ¯0,5→ℳ¯7,3.

The first type of action on CohFTs is a translation action. Let T∈V⁢⟦z⟧. For a CohFT Ω, we define

whenever the infinite sum makes sense.

Example 1.2.12.

If ΩX is the GW CohFT of a smooth projective variety X and τ∈H2⁢(X) is a divisor class, then the infinite sum makes sense and we can define the shifted GW CohFT ΩX,τ of X. For example, if X is the quintic threefold and τ=I1I0⁢H, then shifting to τ is the same as the mirror map q↦Q⁡(q)=q⁢eI1I0.

The second type of action is the action of a matrix R as in the beginning of this section. Let Gg,n define the set of all stable graphs of genus g with n legs. Then we define

where ContΓ is defined by the following construction:

  • •

    At the i-th leg, we place the map

  • •

    At every edge, we place the bivector

  • •

    At every vertex, we place the linear map

  • •

    Finally, we consider the pushforward in cohomology along the gluing morphism

Definition 1.2.13.

Let R be as above. Then we define the translation

Whenever it makes sense, we define

Theorem 1.2.14 ([CGL19]).

Suppose we work with coefficients in ℂ⁢⟦q⟧. Then if

R.Ω is a well-defined CohFT. Moreover, if dimV=dimV′, 𝟏′ is a unit for R.Ω.

We would like to remark a bit more about the translation action when R0≠Id. For simplicity, we will assume that V=V′ and R0⁢𝟏=c⋅𝟏 for some constant c. Then we set T~R≔z⁡(𝟏−c⁢R⁢(z)−1⁢𝟏) and use the dilaton equation to compute

whenever this makes sense.

1.2.4 Reconstruction theorem

Recall that every CohFT defines a Frobenius algebra. We will call a CohFT semisimple if the corresponding Frobenius algebra is semisimple.

Theorem 1.2.15 ([Tel12]).

Let Ω be a semisimple CohFT with flat unit and ω be its topological part. If Ω is semisimple, there exists a unique

such that

Example 1.2.16.

Recall the Hodge bundle CohFT from before. Recall that it is given by the formula

Taking the degree zero part, we see that ω𝔼 is the GW CohFT of a point. Using Mumford’s computation

where κm=p∗⁢ψ¯n+1m+1, B2⁢k are the Bernoulli numbers, and ι is the inclusion of the boundary up to a 2:1 étale cover, and the formula

for any vector bundle E (or by just using the quantum Riemann-Roch theorem6 directly), we obtain the R-matrix

As a sanity check, we will compute Ω1,1𝔼 using the R-matrix. First, we consider the stable graphs in Figure 1.2.

10
Figure 1.2: Stable graphs for g=1,n=1

The first graph Γ1 gives us the contribution

Using the formula B2=16 and the fact that dimℳ¯1,1=1, this becomes

The second graph does not receive tail contributions because dimℳ¯0,3=0. The constant term of the edge contribution is 112, so considering the automorphism and pushing forward to ℳ¯1,1 gives us

where Δ denotes the boundary divisor (with the correct stack structure). Because ψ1=κ1 in this case, we obtain

1.2.5 Operator formalism and geometric quantization

We will return to the symplectic formalism of Givental. This is more convenient for certain computations, but it is in fact equal to what we have before, at least when we want to calculate generating functions. Recall that we had a Frobenius manifold structure on V and we considered the vector space 𝒱≔V⁡⦇z−1⦈ with symplectic form

If we consider the polarization given by 𝒱+=V⁡[z] and 𝒱−=z−1⁢V⁢⟦z−1⟧, then letting 𝐪 be as before (with the dilaton shift), let 𝐩 be coordinates on 𝒱− such that (𝐪,𝐩) form a system of Darboux coordinates for 𝒱.

Definition 1.2.17.

We will call any formal function of the form

an asymptotic element of the Fock space.

Then given such an asymptotic element of the Fock space, we will quantize (infinitesimal) symplectic transformations on 𝒱 by the following formulae:

In particular, this will allow us to understand expressions like R^⁢𝒟. However, we need to be careful because our formulae will involve both the fundamental solution Sτ and the R-matrix, and Sτ is a power series in z−1.7

Theorem 1.2.18 ([Giv01a]).

An operator of the form S⁡(z−1)=Id+S1⁢z−1+⋯ acts on (asymptotic) elements of the Fock space by the formula

where W=∑η⁡(Wm⁢n⁢qm,qn) is defined by the formula

Operators of the form R⁡(z)=Id+R1⁢z+⋯ act by the formula

where V=∑η⁡(pm,Vm⁢n⁢pn) is defined by

For a semisimple CohFT, there is a system of canonical coordinates uα such that the 1-form du is a homomorphism of algebras Tv⁢V→ℂ. Then near a semisimple point, there is a asymptotic solution to the quantum connection of the form

where Ψ switches from flat coordinates to canonical coordinates and U=diag⁡(uα) is the matrix of canonical coordinates. Finally, define

A corollary of Teleman’s reconstruction theorem is the formula

for the Gromov-Witten theory of any smooth projective variety with semisimple quantum cohomology.

Example 1.2.19.

As a final example, we will apply Teleman’s theorem to compute F1 of any semisimple CohFT. This result was first proved by Givental [Giv98] for the equivariant Gromov-Witten theory of GKM orbifolds and extended to the orbifold case by Zong [Zon16]. The argument we give is due to Guo-Ross [GR19]. Let eμ denote an idempotent basis for the quantum product. We will compute

Using the reconstruction theorem, we obtain

Using the identities

where Δμ≔η⁢(eμ,eμ)−1, we obtain the result

which can be placed in the suggestive form

We will now prove the identities in Equation 1.2. For the first identity, we will compute Ω0,4 using the reconstruction theorem. There are two stable graphs, given in Figure 1.3.

000
Figure 1.3: Stable graphs for g=0,n=4

The first graph gives the contribution

The second graph gives the contribution

This has multiplicity 3, so taking the sum and integrating over ℳ¯0,4≅ℙ1, we obtain

where we have used the fact that

This proves the first identity.

To prove the second equality in Equation 1.2, we note that

because eμ is idempotent. Differentiating with respect to uβ, we obtain

Using again the fact that eμ is idempotent, we see that

This implies that

which proves the identity.

2 Higher-genus computations via log GLSM

In algebraic geometry, one profitable way to deal with non-compactness issues is to forcibly compactify our spaces and then understand the contributions coming from the divisor at infinity. For example, this technique was used by Deligne to construct mixed Hodge structures [Del71]. In enumerative geometry, one profitable application of this technique is the capped topological vertex [OP10], which calculates relative invariants of a (partial compactification) of ℂ3 relative to the divisor at infinity and is related to the usual topological vertex by a capping operator. While Jun Li’s technique of expanded degenerations [Li01] allows us to calculate enumerative invariants relative to a smooth divisor, generalizing this to the case of a simple normal crossings divisor requires logarithmic geometry (at least if we want to do it using algebraic geometry). Now, Chen-Janda-Ruan [CJR21] have developed the theory of log GLSMs, which define stable map-type invariants for any GLSM-type theory defined over an arbitrary projective Deligne-Mumford stack (and resolves all compactness issues therein). This theory has been used by Guo-Janda-Ruan [GJR18] to prove finite generation, holomorphic anomaly equations, orbifold regularity, and the LG/CY correspondence for the quintic threefold and some other Calabi-Yau threefold complete intersections in projective space.

2.1 Geometry of log GLSM moduli spaces

Our motivation is the quantum Lefschetz theorem. Here, we let 𝒳 be a smooth projective variety or Deligne-Mumford stack over ℂ and E be a vector bundle over 𝒳. We are interested in computing the Gromov-Witten invariants of a smooth complete intersection 𝒵⊂𝒳 defined by a regular section of E. Because the ambient space is generally easier to work with, it is desirable to find a way to compute the GW invariants of 𝒵 in terms of the data of (𝒳,E).

Question 2.1.1.

Is there a way to compute the Gromov-Witten invariants of 𝒵 using the ambient data of (𝒳,E), possibly involving some correction terms?

While quantum Lefschetz holds in genus zero in the convex case, and the approach of desingularization of the moduli of stable maps to force quantum Lefschetz to hold works well in genus one, this is intractable in higher-genus. Instead, our approach will be to use GLSMs, which were introduced by Witten [Wit92] in the physics literature, and by Fan-Jarvis-Ruan and Kiem-Li, Chang-Li, Chang-Li-Li, and other authors in mathematics [FJR18; KL13; CL12a; CLL15]. We will combine this with the theory of punctured log maps due to Abramovich-Chen-Gross-Siebert [ACGS24] to obtain the theory of log GLSM.

2.1.1 R-maps and log targets

Definition 2.1.2.

An R-map is a commutative diagram

where the morphism 𝔓→B⁢ℂω× is a proper, log-smooth DM-type morphism. If the source is a twisted curve, this is the underlying R-map. If the source is a log curve, this is a log R-map, and if the source is a punctured curve, this is a punctured R-map.

Example 2.1.3.

Consider the diagram

Taking the base change of the underlying R-map to C, consider the diagram

Therefore, the data of an R-map to this target is equivalent to the data of a morphism C→ℙn and a section ρ∈H0⁢(f∗⁢𝒪⁢(−d)⊗ωlog). In particular, if n=0, we are left with the data of a differential ρ∈H0⁢(ωlog).

Example 2.1.4.

We will now give some examples of some log targets. An easy way to compactify a vector bundle is to turn it into a weighted projective space bundle, so for example we may consider

We can give this the log structure of functions which vanish only on the boundary and call it 𝔓𝖦𝖶.

Another example is to consider the example of 5-spin curves. Here, 𝒳=B⁢𝔾m with the degree 5 map to B⁢ℂω×. Then we set

Again, we give this the divisorial log structure and call it 𝔓𝖫𝖦.

A feature of this is that there is a geometric LG/CY correspondence, namely that ∞𝖫𝖦≃∞𝖦𝖶 as log stacks.

Example 2.1.5.

For the (3,3) complete intersection in ℙ5, we consider

with the divisorial log structure.

Example 2.1.6.

We will now consider a general hybrid target. The input is a smooth projective DM stack X, a vector bundle E=⨁Ei over X, a line bundle L over X, and r∈ℤ>0. The spin is the diagram

The target is

with the divisorial log structure at infinity.

We will now turn our discussion to the discrete data. This consists of a genus, curve class, and twisted sector for each marking. We will now construct a dual graph, which has vertices and half-edges. There is an involution

and a vertex map vG:H→V. Legs are the half-edges which are fixed by the involution, and we will label the legs by the marking map

The decorations are the curve classes, sectors, and degrees.

Consider a log map

where Ph is a special point. Because ωlog|Ph is the trivial bundle, we obtain a commutative diagram

which exhibits Ph as a gerbe over 𝔓ℂ. Assuming that there are no orbifold points, theere is a sector map

We require that Ph→𝔓ℂ factors through r¯h, and say that h is compact type if r¯h=0𝔓.

Example 2.1.7.

Consider again X=ℙ0. Then 𝔓=ℙ⁡(ℒω⊕𝒪). If g=1, then a log map is simply η∈H0⁢(ωC), which cannot be stable. Adding a marked point, we obtain a section of ωClog, which we will force to vanish at the marked point.

The stability of underlying R-maps has two conditions:

  1. 1.

    First, that the morphism is representable;

  2. 2.

    Second, that

    where H is a polarization on the target. Here, k≫1≫δ>0.

Example 2.1.8.

Again considering X=ℙ0, we will let C be of genus 2. the limit limλ→∞λ⋅η is the union of a genus-zero curve with a differential which vanishes at one point with multiplicity 2 connected to a genus 2 curve which sits entirely at infinity. Even though the rational component is not a priori stable, it is stabilized by its intersection with the zero section.

Remark 2.1.9.

In the ℙ0 case, stability is equivalent to the condition that

whenever k≫1.

2.1.2 Log geometry and tropicalization

Definition 2.1.10.

Let Y¯ be a scheme or a stack. A log structure on Y¯ is a sheaf of monoids ℳY on Y¯, together with a morphism

inducing an isomorphism α−1⁢𝒪Y¯×≃𝒪Y¯×. The pair Y=(Y¯,ℳY) is called a log scheme/stack and the sheaf

is called the characteristic sheaf or ghost sheaf.

Example 2.1.11.

Any log GLSM target is a log stack.

Example 2.1.12.

Another standard target is a toric variety with ℳY being all functions which vanish only on the toric boundary. This is usually called the divisorial log structure.

Example 2.1.13.

For a toric monoid P, we can consider

The tropical data we will consider is the category 𝖢𝗈𝗇𝖾𝗌 of rational polyhedral cones (σ,N). This has a distinguished object (ℝ≥0,ℤ). Locally, we will define

with the natural lattice structure P∨≔Hom⁡(P,ℕ). In general, the tropicalization Σ⁡(X) of a log scheme X is the generalized cone complex given by gluing local pictures along face maps.

Example 2.1.14.

Consider 𝔸1 with the log structure given by 0. Then

Here, there is only a log structure at 0, which is a copy of ℕ measuring the vanishing order at the origin. The map

sends 1 to 0, so induces 0↪ℕ∨.

Example 2.1.15.

Consider 𝔸2 with the toric log structure. Then Σ⁡(𝔸2) is simply ℝ≥02, which is the same as the support of the toric fan. At the monoid level, there is an ℕ2 at the origin, a copy of ℕ on each axis, and 0 at a general point.

2.1.3 Log curves

Definition 2.1.16.

An n-pointed log curve over a log scheme S consists of

such that

  1. 1.

    The underlying morphism of π is an n-pointed twisted curve;

  2. 2.

    π is log smooth and integral;

  3. 3.

    On the smooth locus of C, the log structure is given by M¯S⊕⨁i=1nℕ⋅pi

Intuitively, the log structure is given by the log structure M¯S at smooth unmarked points, M¯S⊕ℕ at marked points, and M¯S⊕ℕℕ2 at the nodes, which means the pushout diagram

Here, ℓ is the edge length parameter, and (1,0) and (0,1) are the two components.

Because tropicalization is functorial, the morphism C→S induces a morphism

For example, if C has two components, then the node will correspond to the largest cone, each component corresponds to a copy of M¯S, and each marking will give a leg of infinite length. Above a point in Σ⁡(S), the distance between the two components is the edge length parameter evaluated at the point x. To see this more clearly, see Figure 2.1.

C⟶S⟹Σ⁡(C)⟶Σ⁡(S)ℓ⁡(x)xM¯S∨(M¯C,z1)∨=M¯S∨(M¯C,q)∨=(M¯S⊕ℕℕ2)∨(M¯C,z2)∨=M¯S∨M¯C,pi∨=(M¯S⊕ℕ⁢pi)∨legs of infinite lengthz1z2qpi
Figure 2.1: Tropicalization of a log curve.

We will now consider punctured curves. We will consider a diagram

Here, P is a morphism of log schemes which is an isomorphism away from Pi and P¯ is an isomorphism. In addition, we will have an inclusion

This corresponds to allowing poles at pi, and in the tropicalization makes edges finite-length. A picture is given in Figure 2.2.

ℓ⁡(x)(M¯C∘,pi∘)∨xyM¯S∨(M¯C,z1)∨=M¯S∨(M¯C,q)∨=(M¯S⊕ℕℕ2)∨(M¯C,z2)∨=M¯S∨M¯C,pi∨=(M¯S⊕ℕ⁢pi)∨edge offinite lengthz1z2qpipi∘ : puncturingpi
Figure 2.2: Tropicalization of a punctured curve.
Definition 2.1.17.

A punctured curve is a log curve with punctures

A tropical punctured curve is a tropical curve with the additional data of lengths of punctured legs.

2.1.4 Superpotentials

Definition 2.1.18.

A superpotential is a commutative diagram

where 𝔓∘=𝔓∖∞. We say that W has proper critical locus if

is proper.

Equivalently, a superpotential is a ℂω×-equivariant function

In this formulation, W has proper critical locus if and only if Crit⁡Wℂ is proper (as a DM stack). This implies that the critical locus is contained in 0𝔓.

Example 2.1.19.

Let X be a smooth projective Deligne-Mumford stack and E be a vector bundle on X with section s. We will write

Then Crit⁡Wℂ is proper if and only if Z=(s=0) is smooth of codimension equal to the rank of E.

Example 2.1.20.

Let X=ℙ0 and E=ℂ. Let s∈H0⁢(E) be nonzero. Now Wℂ is simply multiplication by s, so the critical locus is empty.

2.1.5 Punctured R-maps

Definition 2.1.21.

A punctured R-map is an R-map with domain a punctured R-map.

Definition 2.1.22.

Let h be a half-edge. Then the contact order at h is defined by

uh^uhuh′Trop⁡fℝ≥00
Figure 2.3: Contact orders. Note the last downward edge has finite length.
Example 2.1.23.

Consider the genus 2 picture from before (target being ℙ0) and consider the same stable limit as an R-map. If we calculate with z=s−1, then

so the pole order at the node must be 3. Tropically, if h^ is the half edge of the node attached to the g=2 vertex, we have c⁡(h^)=−3, so restricting to the g=2 component, we obtain a punctured R-map.

The discrete data of punctured R-maps is

Theorem 2.1.24.

The moduli stack ℛ⁡(𝔓,𝛕) is a proper log DM stack admitting a canonical perfect obstruction theory.

Definition 2.1.25.

We will consider 𝝉 where V⁡(G)={⋆} and L⁡(G)=H⁡(G). These are called vertex-type moduli. If σ=0, then we will denote

for the stack of stable log R-maps, and if σ=ℝ≥0, the we will denote

for the stack of punctured R-maps.

2.1.6 Obstruction theories

Definition 2.1.26.

Recall that we have a diagram

The canonical perfect obstruction theory is defined by

Because it is very complicated, we will not discuss 𝔐⁡(𝝉).

Unfortunately, this is not the obstruction theory that we really want. We will consider the superpotential and then use the cosection localization technique of Kiem-Li [KL13]. Here, we have the diagram

whenever 𝝉 is of compact type. The compact-type condition implies that

factors through the sheaf ω of holomorphic differentials. Because R1⁢π∗⁢ω≅𝒪, we obtain a cosection

By the results of Kiem-Li, this gives us a virtual cycle

supported on R-maps to Crit⁡W, which coincides with the canonical virtual cycle after pushing forward to the stack of all R-maps.

Remark 2.1.27.

In Gromov-Witten theory, the cosection localized virtual cycle satisfies the relation

by work of Chang-Li, Chang-Li, Kim-Oh, Picciotto, and Chen-Janda-Webb [CL12a; CL20b; KO22; Pic21; CJW21].

Of course, we have the problem that ℛg,𝒞⁢(𝔓∘,β) is not proper, so we need to find a way to extend the cosection along the boundary

We need to understand how to differentiate

which will require compact-type legs and a principalization of the boundary.

Definition 2.1.28.

We will say discrete data is compact type if for all h∈L⁡(G), either

  • •

    c⁡(h)=0 and r¯=0𝔓;

  • •

    c⁡(h)≤−1. In this case r¯=∞𝔓.

Example 2.1.29.

Consider the ℙ0 example in genus 1. If we consider η∈H0⁢(ωC), we can view it as η∈H0⁢(ωClog). Then, at the marking, we see that c⁡(h)=0 and r¯h=0𝔓, so the leg is of compact type. In the tropical picture, the entire infinite leg gets contracted to the origin.

Example 2.1.30.

Now consider the same example but with g=2. We will impose that there are no markings by (η=0)=2⁢p. The stable limit as we scale η to infinity had a genus 0 component with a zero of order 2 and a genus 2 component mapping entirely to ∞. If we consider the half-edge emerging from the genus 0 vertex, it touches ∞ with contact order 3, so it is not compact type. On the other hand, the half-edge coming from the genus 2 vertex touches ∞ with contact order −3, so it is compact type (and in the tropical picture the leg has finite length).

Remark 2.1.31.

The data of η0|C2 is equivalent to a fixed isomorphism 𝒪C2⁢(2⁢p)≃ωC2, where C2 is the genus 2 vertex.

Remark 2.1.32.

Tropical curves with only compact-type legs have compact image in Σ⁡(𝔓)=ℝ≥0. In this picture, they are either pointing downward or contracted to the origin.

Having defined compact type insertions, we will now define a modular principalization of the boundary. There is an edge length

and a degeneracy

which intuitively records where v is sent in ℝ≥0.

Definition 2.1.33.

For a toric monoid P, define a partial order a≤b if there exists c∈P such that a+c=b. If we consider a punctured R-map, consider the collection of degeneracies

It has uniform maximal degeneracy if there is a unique maximum

Example 2.1.34.

Let M¯S=ℕ2. Suppose that there are v1,v2 and a vertex v0 which is sent to 0∈ℝ≥0. Then we compute

Similarly, ev2=ℓ2, where ℓ1 and ℓ2 are the generators of ℕ2. These cannot be compared, so we do not have uniform maximal degeneracy.

To uniformize this, we consider a subdivision into three subcones. The first is when ℓ2>ℓ1, the second is when ℓ1=ℓ2, and the third is when ℓ1>ℓ2. After this subdivision, each subcone has uniform maximal degeneracy.

Example 2.1.35.

A family of tropical curves over M¯S∨ with compact-type legs and uniform maximal degeneracy emax∈M¯S has images in ℝ≥0 uniformly bounded from above by emax.

We now define a new stack

to be the stack of punctured R-maps with discrete data 𝝉 and having uniform maximal degeneracy. By removing the condition of uniform maximal degeneracy, we obtain a morphism

which satisfies the following:

  • •

    F is log étale, proper, and surjective;

  • •

    There is as canonical perfect obstruction theory;

  • •

    We have

    for vertex-type moduli.

The boundary Δg,c→⁢(𝔓,β)⋏ defined by the Cartesian diagram

is a log Cartier divisor.

Theorem 2.1.36.

Assume all legs are of compact type and W:𝔓∘→ℒω has proper critical locus. Then

  1. 1.

    𝒰⁡(𝔓,𝝉) has a reduced perfect obstruction theory;

  2. 2.

    In the log R-map case, we have

  3. 3.

    The boundary Δg,c→⋏⁢(𝔓,β) also has a reduced perfect obstruction theory;

  4. 4.

    There is the relation

    where r~ is the pole order of W at infinity.

The reduced perfect obstruction theory for log R-maps is given by the triangle

and for emax>0 is given by

The virtual components of Δg,𝒞⋏ are given by the formula

due to Abramovich-Chen-Gross-Siebert [ACGS24], where

is the tropical type of rigid tropical curves with uniform maximal degeneracy. Here, this implies that 𝝉⋏ is bipartite and rigidity means there is no deformation fixing 𝝉 and Vmax⁢(G) besides scaling emax

Decomposing this further, we have

This is not true if we replace everything with the canonical obstruction theory, so this is quite interesting. Putting these two formulae together, we obtain the tropical decomposition formula

Example 2.1.37.

Consider the ℙ0 example again. When g=2, we obtain six bipartite graphs, which are given in Figure 2.4.

g=2g=1g=0g=0g=0g=0g=0g=1g=0g=0g=1g=2g=2ABCDEF11111223
Figure 2.4: Graphs of tropical types for X=ℙ0 when g=2.

Choosing any nonzero superpotential, we obtain the tropical decomposition formula

Note here that the critical locus is empty, so the virtual cycle must be zero.

In GW theory, the tropical decomposition formula becomes

2.1.7 Effective invariants

We will consider the reduced virtual cycle [𝒰g,𝒞⁢(∞,β)]red. For each h∈L, the evaluation lands in ∞¯ℂ=ℙ⁡(E∨). Then effective invariants are given by

Remark 2.1.38.

In the vertex case, we have the relation

Before we continue, we will consider the geometry of ℛg,c→⁢(∞,β) when X=ℙN. Consider the diagram

This induces a stable map C→ℙN, so in fact it is equivalent to the data of s:C→ℙN and an isomorphism

which is equivalent to an isomorphism

Therefore, when N=0, ℛg,𝒞⁢(∞,β) is the moduli of canonical divisors with specified zero orders, and for general N, we have the moduli of d-spin linear series of rank N.

We can now take a root of the log target ∞, which is given by the diagram

On virtual cycles, we obtain

where ⋆ is either “vir” or “red.”

Geometrically, we must have a balancing condition

Example 2.1.39.

Let X=ℙN and suppose E=𝒪⁡(d) is a line bundle. Then we obtain

This implies that

so β must be one of 0,1,2,…,⌊2⁢g−2d⌋.

When g=0, the upper bound is negative, so there are no genus zero reduced invariants. When g=1, then the upper bound is zero, so the only reduced invariant is in the case β=0. This implies that

so because the contact orders are all negative, they must equal −1.

Example 2.1.40.

We will consider legs of contact order −1 in the ℙ0 case. Let C be smooth of genus g and η∈H0⁢(ωC). We now choose a marking p∈ℂ, we obtain η∈H0⁢(ωlog). The stable limit as we scale η to infinity is a genus g component mapping entirely to ∞ and a genus 0 component with a zero at p. On ℙ1, we see that

This implies the contact order of the genus 0 component with ∞ is 1, so the genus g component has contact order −1.

Remark 2.1.41.

The legs with contact order −1 are created by adding compact type markings outside of ∞ and should be viewed as the unit, denoted by 𝟏.

2.1.8 Log GLSM axioms

Consider the diagram

where π:𝒞∘→𝒰g,c→⁢(∞,β) is the universal punctured curve.

Theorem 2.1.42.

We have the equations

and

for any D∈H2⁢(∞¯). In addition, if we consider the diagram

then we have

Example 2.1.43.

If we consider Zd⊂ℙN and let g=1, then we only need to compute one non-ambient invariant.

The reduced virtual dimension of 𝒰g,c→⁢(∞,β) is given by the formula

which also happens to equal

Therefore, when the virtual dimension of the stable map moduli is negative, then the reduced virtual dimension is zero.

Example 2.1.44.

Consider Z5⊂ℙ4. The reduced dimension of 𝒰g,𝒞⁢(∞,β) is simply

We can remove all legs of contact order −1, so we reduce to the case when c⁡(h)=−2 for all h. Therefore, there are exactly ⌊2⁢g−25⌋+1 reduced invariants.

Example 2.1.45.

For a general semi-positive hypersurface in ℙN, the reduced virtual dimension when g≥2 is

Using the balancing condition, we obtain an upper bound of

If 2−N+2⁢N+2d is negative, then the reduced virtual dimension is structly less than

After removing all legs of contact order −1, all effective cycles vanish! For example, we have this negativity whenever d=3 and N≥9, when d=4 and N≥6, and when d≥5 and N≥5. For complete intersections in other targets, we can run the same arguments, and they are governed by birational invariants.

2.1.9 Uniform minimal degeneracy

Definition 2.1.46.

A punctured R-map has uniform minimal degeneracy if there exists a unique

Sometimes, we will also need to consider disconnected graphs, so if we enforce uniform minimal degeneracy, we will obtain a cartesian diagram

where both vertical arrows are log blowups and hence log étale and projective. These introduce two new tautological classes on 𝒰⁡(∞,𝝉⋎), which are

coming from emax and

coming from emin. The class ψmin is a key ingredient for the tropical decomposition and is needed for the virtual localization formula.

2.1.10 Some examples

Here, we will give a few exmples.

Example 2.1.47 (Quintic).

Here, the target is X5⊂ℙ4. We will consider the log target

with the infinity part

Example 2.1.48 (Double cubic).

Here, the target is X3,3⊂ℙ5. We will consider the log target

with the infinity part

Example 2.1.49 (Quintic FJRW).

We will consider the log target

with the infinity part

2.1.11 ℂω× action

Our goal is now to proceed towards a virtual localization formula for log GLSM. Here, the two structural formulae are related as in Figure 2.5.

[𝒰]red GW/FJRW [𝒰⁡(𝔓)]vir [𝒰⁡(∞)]red effective twisted GW/FJRW tropical decompositionlocalization
Figure 2.5: Relation between structural formulae and virtual cycles in log GLSM

There are two ways to think about this ℂω×-action. If we have an R-map

this was a morphism of stacks, so there is a 2-morphism making the diagram commute. Abstractly, the action simply scales the 2-morphism, which is an isomorphism of line bundles.

More concretely, consider the quintic example. Then an R-map is equivalent to the data of a stable map f:C→ℙ4 and a section of the projective bundle

Using this description, the ℂω×-action simply scales the p-field, where we scale the first summand and not the second.

We now have a ℂω× action on all moduli spaces that we considered previously, for example ℛ⁡(𝔓,β) or 𝒰⁡(𝔓,β).

Proposition 2.1.50.

The perfect obstruction theories 𝔼ℛ and 𝔼𝒰red are ℂω×-equivariant.

Remark 2.1.51.

A key input to this result is that the superpotential is ℂω×-equivariant. If it was in fact invariant, then there would be no need to develop the theory of log GLSM.

We may now apply the virtual localization theorem, proved in increasing strength by Graber-Pandharipande, Chang-Kiem-Li, and Aranha-Khan-Latyntsev-Park-Ravi [GP99; CKL17; Ara+24] to decompose the reduced virtual cycle as

2.1.12 Fixed loci

There are several kinds of fixed loci:

  • •

    There is a fixed locus ℛg,c→⁢(0𝔓,β) of all maps going into the zero-section.

  • •

    There is another fixed locus ℛg,c→⁢(∞𝔓,β) of maps going into infinity.

  • •

    More general fixed loci may be described by decorated bipartite graphs, with vertices at either 0 or ∞ being decorated by a genus and curve class and edges being decorated by the contact order (which implies the degree of the edge).

From a bipartite graph, the R-maps which appear in the corresponding fixed locus arise from taking scaling limits of R-maps of the corresponding tropical type. From the fixed curves, we can obtain other curves of the same tropical type by smoothing nodes which appear at 0.

The moduli spaces corresponding to stable vertices are given as follows. When v∈V∞, the vertex moduli space is

where all contact orders are negative. When v∈V0, then we have

where X was the ambient space.

2.1.13 Virtual localization formula

Example 2.1.52.

In the case of the quintic, note that ∞≅0. Then we have a stabilization morphism st:𝒰→ℳ¯g,n⁢(X,β). Then we will describe the virtual localization formula on a moduli space which is related to the fixed locus. It is given by the Cartesian diagram

We then obtain the virtual localization formula

where t is the equivariant parameter and ch is the contact order. The 0 part gives the twisted GW theory of ℙ4 with special insertions and the ∞ part gives a more general version of effective invariants.

Example 2.1.53.

In the case of the double cubic, note that ∞ℂ≅ℙ5×ℙ1→ℙ5=0ℂ. Then we have a stabilization morphism st:𝒰→ℳ¯g,n⁢(ℙ5,β). For v∈V∞, we have

Then we will describe the virtual localization formula on a moduli space which is related to the fixed locus. It is given by the Cartesian diagram

Here, Δ is composition the diagonal and deleting the ℙ1 in the first factor. We then obtain the virtual localization formula

where t is the equivariant parameter, ch is the contact order, and H∞ is the hyperplane class on the ℙ1 factor. The 0 part gives the twisted GW theory of ℙ5 with special insertions and the ∞ part gives a more general version of effective invariants.

Example 2.1.54.

In the case of the FJRW theory of the quintic, note that ∞C=(ℙ4,𝒪⁡(1))5→0ℂ=B⁢μ5. Then we have a stabilization morphism st:𝒰→ℳ¯g,𝒞15. We now set

Then we will describe the virtual localization formula on a moduli space which is related to the fixed locus. It is given by the Cartesian diagram

We then obtain the virtual localization formula

where t is the equivariant parameter and ch is the contact order. The 0 part gives the twisted GW theory of ℙ4 with special insertions and the ∞ part gives a more general version of effective invariants.

2.2 Applications to Gromov-Witten theory

2.2.1 Genus two calculations

The goal is to prove the formula

for the quintic threefold.

We will first use the localization formula to compute

There are many localization graphs, which may be obtained as modifications of the ones in Figure 2.4. The most important ones are displayed in Figure 2.6.

g=2 β g=1 β g=0 β g=0 β g=0 β1 g=0 β2 g=1 β=0 g=0 β=0 g=0 β=0 g=1 β=0 g=1 β=0 ABCDE1111122
Figure 2.6: Localization graphs with nonzero contribution for g=2.

The contributions will now be given. For graph A, the contribution is

which is the 𝒪⁡(5)-twisted Gromov-Witten potential of ℙ4. Graph B contributes

Graph C contributes Equation 2.2, and graph D contributes Equation 2.3. Graph E will contribute

where ceff is the effective invariant.

To obtain a formula with five terms from all of our localization graphs, we will apply the shift μ=(1−I0)⁢ψ+I1⁢H which comes from quasimap wall-crossing. Under the shift, the contribution of graph E becomes the constant F2⁢(Q=0), while the contributions from all other graphs vanish.

2.2.2 Genus two mirror theorem

For convenience, we will set t=5⁢s. This will make the equivariant parameters the same for both the Gromov-Witten and FJRW calculations. This gives us

Because the twisted theory is semisimple, we can compute it using the Givental-Teleman reconstruction theorem [Giv01b; Tel12]. For a graph Γ∈{A,B,C,D,E}, the first step is to write

using the descendant-ancestor correspondence of Kontsevich-Manin [KM98]. The second step is to use the Givental-Teleman reconstruction theorem to compute the ancestor invariants by a stable graph sum.

For simple reasons, we can rewrite [S⁡(ψ¯)⁢α⁡(ψ¯)]+=S⁡(5⁢s−5⁢H)⁢α1⁢(ψ¯). Both S⁡(5⁢s−5⁢H) and

have entries in a ring

where the generators are defined using genus-zero invariants. Using the fact that

the mirror theorem, and S∗⁢1=z−1⁢J, we see that S∗⁢1|z=5⁢s−5⁢H=1. To compute the other entries, we use the quantum connection

Quantum multiplication by τ˙ is given by the matrix

where the ∗ denote nonzero entries. However, the entries of A and their derivatives lie in

Using a computer, we then get the exact formulae for

A miracle implies that

If we recall that the physicists predicted that Fg is a polynomial in five generators (which are exactly the same as the five generators here).

Remark 2.2.1.

A similar calculation for FJRW theory yields similar formulae. For example, graph B contributes

and graph D contributes

2.2.3 LG/CY correspondence for the quintic

Write

Similarly, we can write

We can then construct the generators for the FJRW theory of the quintic similarly to the generators for the GW theory of the quintic.

Theorem 2.2.2.

Write

Then after identification of the generators, we have

The proof of this result is a direct matching of contributions from graphs in the virtual localization formula. It relies on the fact that A𝖢𝖸=A𝖫𝖦 under a suitable choice of basis, which also matches the identification of the generators. The tropical input from log GLSM is that the effective invariants are equal in both the CY and LG phases. Finally, we need to identify the insertions in the two phases, and this follows from a direct computation of the specialized S-matrix.

2.2.4 Geometry of LG/CY correspondence

Let ∙∈{𝖫𝖦,𝖢𝖸}. Recall that

and ∞𝖢𝖸,ℂ=ℙ4. The log target for the LG phase is 𝔓𝖫𝖦,ℂ=[ℙ5/μ5], but the infinity-section is

Remark 2.2.3.

There is a modification of the targets such that ∞𝖢𝖸≅∞𝖫𝖦.

The first step is to match the virtual localization formulae. In the LG phase, there are no curve classes at 0, so we compensate for this by using insertions. Also, note that we need to consider the inertia stack

and in fact the contact orders determine the twisted sectors at the nodes. Then there is a bijection between the 𝖫𝖦 and 𝖢𝖸 decorated graphs. For example, we match the graphs in Figure 2.7.

g=4 β=1 g=0 β1 g=0 β2 g=4 β∞=1 g=0g=0212135452515𝖢𝖸𝖫𝖦
Figure 2.7: Corresponding graphs in CY and LG phases.

Now let

where the state spaces are given by

Defining

we define

We may also define the contribution ContΓ0,τ:(ℋ∙⁢⟦ψ¯⟧)|E|→ℚ of 0 by

Finally, for any contact order c∈{1,2}, let

be defined by

in the CY case and

in the LG case.

Putting these ingredients together, we obtain the following formula:

Theorem 2.2.4.

We have the identity

The proof of this theorem follows the following steps:

  1. 1.

    Virtual localization;

  2. 2.

    Apply the shift coming from quasimap wall-crossing;

  3. 3.

    Keep track of signs and powers of 5.

It now remains to match the contributions. Here, the contributions ContΓ,∙∞ match because of the geometric isomorphism ∞𝖢𝖸≅∞𝖫𝖦. The matching of the 𝒮c,τ∙ follows from a direct computation.

2.2.5 Twisted theories

Recall that the 𝒪⁡(5)-twisted GW theory of ℙ4 is given by integrals against

This is very hard to compute with, so we add a torus action of (ℂ×)5 on ℙ4. We will denote the equivariant parameters on the base by λ0,…,λ4. This gives us a general 𝒪⁡(5)-twisted theory.

Unfortunately, this is extremely complicated, but we consider the formal quintic theory, which is given by setting s=0 and

Therefore, we have a theory with only one equivariant parameter λ. This theory (also called the λ-twisted theory) has several advantages:

  • •

    It has been studied before by Zinger, Kim-Lho, Lho-Pandharipande, and various other authors [Zin09; KL18; LP18];

  • •

    The λ-twisted I-function

    is much nicer than the s-twisted I-function

    In particular, the appearance of 5⁢s in the numerator of each term of Is is the reason for the appearance of extra generators for the s-twisted theory.

Our goal now is to rewrite log GLSM invariants using the λ-twisted theory. Using the tropical decomposition formula, we obtain

Because the canonical virtual cycle does not depend on the equation of the quintic, we can compute

Using the log GLSM localization formula, we now obtain

We will now apply operations to graphs as in Figure 2.8.

g=2g=1g=1g=1g=1g=1g=1g=2g=1Γ𝗍𝗋𝗈𝗉Γ𝗅𝗈𝖼Γ𝗍𝗋𝗂Γ𝖻𝗂⟹⟹
Figure 2.8: Tropical graph, localization graph, and tripartite graph

We may then combine the two decompositions to form a tripartite graph. Using this, we then obtain

Here, the second equality comes from combining the middle and upper layers of a tripartite graph.

Remark 2.2.5.
  1. 1.

    It is not a priori clear that the specialization s=0,λj=ζj⁢λ makes sense. However, we note that H5=λ5, so H−1=H4λ5, so the specialization

    makes sense.

  2. 2.

    The graph Γ𝖻𝗂 may have contact orders c≥3.

2.2.6 Combinatorial structure theorem

Recall that the R-matrix action was found by Givental as follows:

  • •

    We will stabilize a localization graph to a stable graph. This involves contracting rational tails and chains:

  • •

    The tails become the translation action;

  • •

    The chains become edge contributions V⁡(z,w);

  • •

    The ends become leg contributions R⁢(z)−1.

We will consider something similar in log GLSM. We will begin by stabilizing graphs Γ𝖻𝗂 as in Figure 2.9.

0∞g=0g=1g=1g=0g=1g=1g=1g=1⇒
Figure 2.9: Stabilization of Γ𝖻𝗂

Applying Givental-Teleman to the vertices at level 0, we see that there are ∞⁢∞ edges, 0⁢∞ edges, and 00 edges. There are also both 0 and ∞ legs. This implies that

where ℛ is generalized from the Givental formalism. It is given by

Also, the matrices R and S are determined from the quantum differential equation, which in the λ-twisted case are given by

2.2.7 GW theory of the quintic

We are now able to apply the combinatorial structure theorem.

  • •

    We have a stable graph sum formula involving ℛ;

  • •

    The entries of ℛ are polynomials in the five generators, which are determined by A;

  • •

    The generalized tail contribution

    is a polynomial in the holomorphic generator X=11−55⁢q.

To prove the holomorphic anomaly equation, consider differentiating by S∈{A,B1,B2,B3}. We look directly at the stable graph localization formula. Differentiation comes from cutting edges as in Figure 2.10.

∂∂S⟹+++
Figure 2.10: Cutting of edges.

If a matrix R satisfies the equation

then the V-tensor satisfies the equation

where

Therefore, when we compute ∂∂S⁢Fg, we place ΔS at the nodes, and this will match the holomorphic anomaly equation.

Remark 2.2.6.

This method also proves that for any extra generator ℰ, we have

This implies that Fg is independent of ℰ, so we reduce the nine generators to five generators.

Remark 2.2.7.

In fact, the holomorphic anomaly equations hold for arbitrary values of the effective invariants. In particular, if we define Fgc by the formal theory and effective invariants, then

  1. 1.

    The generating function Fgc satisfies the holomorphic anomaly equations;

  2. 2.

    Fgc has orbifold regularity.

Orbifold regularity is proved by showing that every graph which appears in the graph sum for Fg satisfies orbifold regularity.

The meaning of orbifold regularity is that if we replace X,A1,B1,B2,B3 by the LG versions X𝗈𝗋𝖻,A1𝗈𝗋𝖻,B1𝗈𝗋𝖻,B2𝗈𝗋𝖻,B3𝗈𝗋𝖻, then Fg is a regular function near q=∞.

Conjecture 2.2.8 (Conifold gap).

At the conifold point q=5−5, there is the conifold gap condition

where Δ is the local flat coordinate near the conifold point.

The conifold gap condition has been checked for g≤5 by computer. Unfortunately, a systematic way of proving the conjecture is not known.

3 See BCOV from the A-side: MSP fields

Using the master space construction introduced by Thaddeus and Dolgachev-Hu [Tha96; DH98] in the context of variation of GIT, Chang-Li-Li-Liu [CLLL19; CLLL22] defined the theory of Mixed-Spin-P fields, which connects the Gromov-Witten and FJRW theories of the quintic threefold. The moduli space has a torus action, and one type of fixed point gives the Gromov-Witten theory of the quintic, which led to hope that the theory could be used to compute GW invariants of the quintic. However, the calculations quickly proved intractable, and an insight of Guo, which led to the creation of N-Mixed-Spin-P fields [CGLL21], was required to resolve the difficulties. This led to the proof of the finite generation conjecture and holomorphic anomaly equations for the quintic threefold by Chang-Guo-Li [CGL21; CGL19], which was generalized to hypersurfaces in weighted projective space by the author [Lei24b; Lei24a]. More recently, a new stability condition for GLSMs was introduced by Chang-Guo-Li-Li-Zhou [Cha+23b], and this will enable the construction of MSP moduli spaces for other targets.

3.1 Geometry of MSP moduli spaces

The genus-zero Gromov-Witten invariants of the quintic were computed by Givental, Lian-Liu-Yau, and various other authors [Giv96; LLY97]. Their computation uses the fact that the moduli space

is smooth and the obstruction sheaf corresponding to 𝒪⁡(5) is a vector bundle. This fails, however in higher genus, and in genus one, the work of Zinger, Vakil-Zinger, and Li-Zinger [Zin09; VZ08; LZ09] on reduced Gromov-Witten invariants desingularizes the moduli spaces and enables computation of the genus one invariants. This result was also recovered by Kim-Loh [KL18] using quasimap wall-crossing [CK20; Zho22].

There is also the approach of Maulik-Pandharipande [MP06] using the degeneration formula, which in principle computes all of the invariants but is impossible to calculate with in practice. Finally, there is the approach of Fan-Lee [FL19].

3.1.1 P-field reformulation of GW and FJRW theory

Let F=x1N+⋯+xrN be an arbitrary Fermat polynomial. The setup works in full generality, but the calculations can only be conducted for the quintic threefold. The first step is to construct the moduli space of P-fields following the work of Chang-Li [CL12a], which comes from the work of Guffin-Sharpe [GS09] in physics.

Consider the vector space ℂ5×ℂ with the action of ℂ× given by

The GIT quotient with the positive stability chamber is Kℙ4. We will consider the superpotential

A map C→Kℙ4 is equivalent to the data

This is just a stable map, and the P-field is given by modifying ρ by ωC to obtain

We will also require that (φ1,…,φ5) is nonvanishing, so it gives a stable map to ℙ4.

The moduli space ℳp of such ξ is noncompact. However, we can follow Kiem-Li [KL13] and use cosection localization for a cosection

The degeneracy locus is in fact

where Z5⊂ℙ4 is the quintic threefold. This gives a virtual cycle1

FJRW invariants were originally defined using analysis. There are algebraic reformulations by Polishchuk-Vaintrob and by Chiodo [PV16; Chi06]. We once again consider ℂ5×ℂ with an action of ℂ× given by

However, we use the negative stability chamber to obtain the quotient [ℂ5/μ5]. The objects in our moduli space are given by

such that ρ is nowhere zero. This implies that ℒ5≅ωC. We will consider the same superpotential

The moduli space ℳ15,5⁢p of such η is still noncompact, but cosection localization again provides a virtual cycle

recovers the FJRW virtual cycle.

3.1.2 Master space and MSP fields

Following the work of Thaddeus [Tha96] and various other authors, we are interested in a geometric lift of a path connecting two GIT quotients. Consider the space ℂ5×ℂ×ℙ1 with the action of ℂ× given by

Now define the master space

Because ℙ1 is a GIT quotient, we can reformulate the GLSM for this target as being given by the data of

such that (φ1,…,φ5,μ) is nowhere zero, (ρ,ν) is nowhere zero, and (μ,ν) is nowhere zero. We also require that Aut⁡ϑ is finite.

Remark 3.1.1.

Because the target is a Deligne-Mumford stack, we will need to require the source curve to be a twisted curve. For orbifold markings, we require that ℒ|Σ is a nontrivial automorphism of Aut⁡Σ. For scheme markings we will consider either (1,ρ) markings where we force ρ=0 or (1,φ) markings were we force φ=0. These are required to construct the cosection. These markings are called narrow, and all other markings are called broad.

3.1.3 Cosection localization and virtual cycle

The numerical data is g=g⁡(C), d0=deg⁡ℒ⊗𝒩, and d∞=deg⁡𝒩. This gives the moduli space

with a relative perfect obstruction theory

Here, π:𝒞→𝒲g,d→ is the universal curve.

Fiberwise, the cosection is given by the following construction. Let2

Then the cosection is given by

and its image lies in H1⁢(ωC)≅ℂ (this isomorphism is canonical).

The degeneracy locus σ−1⁢(0) is given by

The first difficult step is to prove the properness of the degeneracy locus.

Theorem 3.1.2.

The degeneracy locus σ−1⁢(0) is proper in the narrow setting.

3.1.4 Virtual localization

There is a ℂ×-action on 𝒲 which actually preserves σ−1⁢(0) given by

This allows us to compute MSP invariants using virtual localization.

We will now relate the fixed loci to the previous moduli spaces.

  • •

    Let ϑ∈𝒲. If μ=0, the 𝒩=𝒪C by the condition that (μ,ν) is nowhere vanishing. This reduces the fields to (φ,ρ), and we need (φ1,…,φ5) to be nowhere vanishing, so we obtain P-fields. We will call this level 0

  • •

    If we let ν=0, then the information of μ disappears and ρ must be nowhere zero. This reduces us to FJRW theory. This is called level ∞.

  • •

    Finally, if we set φ=ρ=0, then both μ,ν are required to be nowhere vanishing, so we only have the data of a curve. When we perform virtual localization, this will give us Hodge integrals. This is called level 1.

Fixed loci F⊂𝒲ℂ× will be labelled by graphs. The problem is that there may be a situation of two edges v0→v1→v∞ where one edge has degree d and the other edge has degree −d. While there is no geometric smoothing of the node (it is banned by the stability condition), there are nontrivial infinitesimal deformations, so we cannot decompose the virtual cycle. Fortunately, there is a significant simplification, which was first discovered by Chang-Li [CL20a] for the quintic and should hold for all other Calabi-Yau threefolds.

Theorem 3.1.3 ([CL20a]).

These graphs contribute 0 to the MSP virtual localization formula.

Let Λ be the set of regular graphs. Then the virtual localization theorem for cosection localized virtual cycles due to Chang-Kiem-Li [CKL17] gives

Because of irregular vanishing, we can decompose [𝒲Γ]locvir along the vertices.

  • •

    At a level 0 vertex, we have

    where Z is the quintic.

  • •

    At a level ∞ vertex, we have

  • •

    At a level 1 vertex, we simply have

3.1.5 NMSP fields

The setup can calculate the genus 1 Gromov-Witten invariants of the quintic, which was done by Chang-Guo-Li-Zhou [CGLZ20], but the calculations are too complicated to perform more computations. However, an insight of Guo is to replace μ with

and consider an action of (ℂ×)N scaling μ diagonally.

3.1.6 Reformulation of stability condition

Recall that the original MSP master space is given by a GIT quotient of ℂ5×ℂ×ℂ×ℂ with coordinates x1,…,x5,p,u,v by (ℂ×)2 acting with weights

and the polarization t1⁢t22. This is a change of coordinates from the standard presentation

and polarization t2.

Recall that a stable MSP field

consists of the data of

  • •

    A pointed twisted curve 𝒞 – the nodes are required to be balanced, which means that they locally look like

  • •

    ℒ,𝒩∈Pic⁡𝒞 are line bundles such that ℒ⊕𝒩 is representable;

  • •

    φ1,…,φ5∈H0⁢(ℒ), ρ∈H0⁢(ℒ−5⊗ω𝒞log), μ∈H0⁢(ℒ⊗𝒩), and ν∈H0⁢(𝒩) are sections such that

    • –

      (φ,μ) is nonvanishing;

    • –

      (ρ,ν) is nonvanishing;

    • –

      (μ,ν) is nonvanishing;

    • –

      |Aut⁡(ξ)|<∞.

    We denote the nonvanishing condition by MSP-Stab-I and the finiteness of automorphisms by MSP-Stab-II.

There is a superpotential

with critical locus (see Figure 3.1)

The semistable locus is

Kℙ4[ℂ5/μ5]ℙ⁡(1,5)𝖢𝗈𝗇𝖾⁡(X5)ℙ4
Figure 3.1: Critical locus 𝖢𝗈𝗇𝖾⁡(X5)∪ℙ⁡(1,5)

The degeneracy locus of the MSP moduli space (with reduced stack structure) consists of all ξ such that 𝒞 is mapped entirely into the critical locus. In other words, we require that

The upshot of this is that the critical locus is proper.

Theorem 3.1.4 ([CLLL19]).

Fixing the discrete data, the moduli of stable MSP fields is a separated DM stack of finite type and the degeneracy locus of the cosection is proper.

Our goal is now to generalize the stability condition:

  • •

    Stability should come from GIT with the R-charge (which gives the ωlog);

  • •

    Properness should follow from properness of the target (or at least the critical locus).

Lemma 3.1.5.

Under MSP-Stab-I, the condition that |Aut⁡ξ|<∞ is equivalent to the condition that

as A→15+0+.

Now note that the unstable locus is cut out by the functions

This allows to reformulate MSP-Stab-I as the condition that the sections

have no common zeros. This reformulation allows us to give easier proofs of basic facts about MSP fields.

Corollary 3.1.6.

Fixing the discrete data, stable MSP fields are bounded.

Proof.

We need to prove that

  • •

    There are finitely many irreducible components;

  • •

    The degrees deg⁡ℒ𝒞′ and deg⁡𝒩|𝒞′ are uniformly bounded for irreducible components 𝒞′⊂𝒞.

The first condition follows from positivity of the line bundle

For the second condition, suppose for simplicity that 𝒞′=𝒞. Then we vary the polarization to obtain the inequalities

This gives a bounded region, and the denominators are bounded, so we are done. □

Remark 3.1.7.

In the nonabelian case, boundedness is much harder. See [Cha+23a] for more details.

3.1.7 General setup

The setup consists of the following data:

  • •

    An extension

    of reductive groups;

  • •

    An affine scheme V (think a vector space or critical locus) with an action of Γ such that Vs⁢(θ)=Vss⁢(θ)≠∅;

Remark 3.1.8.

The lift ϑ of θ is not essential.

Definition 3.1.9.

An LG-quaasimap is a tuple

where

  1. 1.

    𝒞 is a pointed twisted curve;

  2. 2.

    u:𝒞→[V/Γ] is representable such that the base locus u−1⁢([Vun⁢(θ)/Γ]) is discrete and disjoint from the special points;

  3. 3.

    κ:u∗⁢Lϖ→≅ω𝒞log is an isomorphism of line bundles.

Example 3.1.10.

In the MSP case, either V=ℂ8 or Crit⁡(p⁢∑xi5). We also have Γ=(ℂ×)2×ℂR×.

  • •

    A map u:𝒞→[V/Γ] is the data of ℒ,𝒩,ℛ∈Pic⁡(𝒞) and fields φ→,ρ,μ,ν. Here, note that ρ∈H0⁢(ℒ−5⊗ℛ).

  • •

    The data of κ is an isomorphism u∗⁢Lϖ=ℛ≅ω𝒞log.

A fancier way to write an LG-quasimap is as a commutative diagram

3.1.8 Ω-stability

Define

Then recall that we have a morphism

Lemma 3.1.11 ([FJR18]).

Each Rk is Γ−i⁢n⁢v⁢a⁢r⁢i⁢a⁢n⁢t.

Set

where

Definition 3.1.12.

An element f∈R∙=⨁k≥0Rk is homogeneous if 0≠f∈Rk,c for some k and c. We then define the θ-weight of f by k≕wt⁡(f), the R-charge of f to be c, and the slope of f to be ck.

Example 3.1.13.

In the original MSP example, weights are given in Table 3.1. In this example, we have

Table 3.1: Weights.
xi⁢v2 u⁢v u10⁢p
Weight 1 1 5
R-charge 0 0 1.

In general, for f∈Rk,c and a map u:𝒞→[V/Γ], we have

Now choose a package Ω=(S,A,ϑ), where

  • •

    S⊂R+ is a finite set of homogeneous elements;

  • •

    A>maxf∈S⁡{slope⁡(f)}.

Definition 3.1.14.

A prestable LG-quasimap ξ is Ω-stable if

Ω-1

The base locus u−1⁢([V⁡(S)/Γ]) is discrete and disjoint from the markings and nodes;

Ω-2

For all non-special points x∈𝒞sm, we have

Ω-3

We have the ampleness condition

Remark 3.1.15.
  1. 1.

    The A here corresponds roughly to 1ε in the previous literature about quasimaps. Roughly, decreasing A means attempting to force more of the curve to go into the stable locus;

  2. 2.

    V⁡(S) may be larger than the unstable locus Vun=V⁡(R+);

  3. 3.

    If any irreducible component 𝒞′⊂𝒞 satisfies ω𝒞log|𝒞′>0, then Ω-3 is implied by either Ω-1 or Ω-2;

  4. 4.

    In some way, Ω-stability is independent of ϑ (up to changing the other data).

Theorem 3.1.16.

Fixing deg⁡u∗⁢Lϑ=d, the genus g, and the number k of marked points, the stack

is a separated DM stack of finite type.

Theorem 3.1.17.

Suppose that X⫽θG is projective and S is full (which means that set-theoretically V⁡(S)=Vun⁢(θ)). Then LGQg,kΩ⁡(X,d) is proper.

Example 3.1.18.

In the original MSP example, S is full. The element u10⁢p∈S has slope 15, so we need A>15. For any smooth point q∈𝒞sm, the conditions Ω-1 and Ω-2 are equivalent. In particular, they mean that at least one of the following conditions is satisfied:

As A→15+0+, we see that the first inequality holds if and only if ordq⁡φi=ordq⁡ν=0, which is equivalent to nonvanishing of φi⁢ν2. This recovers the original nonvanishing condition MSP-Stab-I.

3.1.9 More examples

Example 3.1.19.

Consider ℙN with an R-charge. This means that we take Γ=ℂ××ℂR× and V=ℂN+1. We will have Γ act with weights

For simplicity, assume that c0≤⋯≤cN. In this model, an LG quasimap is a package

For the stability condition, we choose S={x0,…,xN} and A>cN.

The condition Ω-2 becomes the condition

In general, we see that ordx⁡(φi) is allowed to be positive. The condition Ω-3 becomes the positivity condition

For example, if N=1, c0=0, and c1=1, then as A→1+, Ω-2 becomes the condition that either φ1⁢(x)≠0 or ordx⁡(φ0)≤1 at any non-special point x∈𝒞. In particular, it is impossible to completely remove basepoints.

Example 3.1.20.

We will now consider the LG phase for a complete intersection

We will consider V=ℂN+1×ℂ×ℂ and Γ=ℂ××ℂR× acting on V with weights

We will choose the character (t1,t2)↦t1−1, and the corresponding GIT quotient is

An LG quasimap is given by the data

We choose S={p1,p2}, and assuming that d1≤d2, we need to choose A>1d1.

The stability condition Ω-2 becomes

while Ω-3 becomes the positivity condition

As A→(1d1)+, the condition Ω-2 becomes the condition that either ρ1⁢(x)≠0 or ordx⁡(ρ2)≤d2d1−1. Therefore, if d2<2⁢d1, then ρ1 and ρ2 have no common zeroes.

Remark 3.1.21.

If d1=d2, this hybrid model admits a good lift in the sense of Fan-Jarvis-Ruan [FJR18] and the theory was studied by Clader [Cla17].

Example 3.1.22.

We will now consider MSP theory for X3,3⊂ℙ2×ℙ2. The Calabi-Yau phase is given by Kℙ2×ℙ2, which is a GIT quotient of ℂ3×ℂ3×ℂ by ℂ××ℂ× with weights

There are three phases for this model, which are given in Figure 3.2.

Kℙ2×ℙ2[Tot⁡(𝒪ℙ2⁢(−1)⊕3)/μ3][Tot⁡(𝒪ℙ2⁢(−1)⊕3)/μ3]
Figure 3.2: Phases of X3,3⊂ℙ2×ℙ2.

To construct the MSP moduli space involving all phases, consider the data

We will choose the set

The maximum slope is 23, so we will choose A=23+0+. In particular, part of Ω-2 becomes the condition

which is equivalent to

This is equivalent to the condition that u1 and u2 are nonzero and ord⁡(ρ)≤3. Putting all of the conditions together, we obtain the following nonvanishing conditions:

  • •

    (u1,u2,v)≠0;

  • •

    (y,u2)≠0;

  • •

    (x,u1)≠0;

  • •

    Either v≠0 or ord⁡(ρ)≤3.

Unfortunately, it is not possible to remove the basepoints.

Consider the action of ℂ××ℂ× given by scaling μ. A schematic diaagram of the the types of fixed loci is given in Figure 3.3.

Kℙ2×ℙ2ℙ2ℙ2pt[Tot⁡(𝒪ℙ2⁢(−1)⊕3)/μ3][Tot⁡(𝒪ℙ2⁢(−1)⊕3)/μ3]ℒ1≅ℒ2≅𝒩−1ρ∈H0⁢(𝒩6⊗ω𝒞log)ordx⁡(ρ)≤3
Figure 3.3: Diagram of fixed loci for X3,3∈ℙ2×ℙ2.
Remark 3.1.23.

There is a wall-and-chamber structure as we vary A. If we consider A→∞, the condition Ω-2 is always true, and Ω-3 bans all rational tails. This gives a stability condition similar to stable quasimaps which was already studied by Fan-Jarvis-Ruan. In addition, there should be a wall-crossing structure as we vary A (which is desirable because we want to separate the different kinds of fixed points).

Remark 3.1.24.

Ω-stability exhibits a phenomenon which is close to being independent of S. If V is reduced and V⫽θG is projective, then whenever S is full, we have

  • •

    The quantity

    is independent of S;

  • •

    Fixing A and ϑ, Ω=(S,A,ϑ)-stability is independent of S.

3.1.10 Proof of key properties of the moduli space

The idea of the proof is as follows:

  • •

    First prove the ℙN case;

  • •

    Compare the general case to ℙN.

To warm up, we will review the proof of properness for stable maps to ℙ1. A stable map to ℙ1 is equivalent to the data

such that there are no common zeroes and ℒ|𝒞′>0 whenever ω𝒞log|𝒞′<0. We will use the valuative criteria.

Consider a family over a disk Δt. We have two divisors (φ1=0) and (φ2=0), but they may meet on the central fiber transversely with vanishing order 1. We can blow up the intersection point to separate the two divisors as in Figure 3.4. Denote the blowup morphism by π:𝒞~→𝒞. Then we actually have

where E is the exceptional divisor. Then the stable limit is given by

If A>1, we simply keep the base point (and disallow the rational tail).

↓t=0Δtφ2=0φ1=0𝒞↓t=0ΔtE𝒞~π
Figure 3.4: Stable limit of family of stable maps.

If we introduce an R-charge 0,c, an LG quasimap is given by

Now π∗⁢φ2 vanishes to order 1+c on E. Blowing up again, the vanishing orders now become (1,2⁢c)∼(0,2⁢c−1), which seems worse. If instead (φ1=0) has multiplicity k>c, then on E the vanishing orders are (k,1+c)∼(k−1−c,0). Then we blow up the node, which does not change ω𝒞log. Therefore, we cannot improve a length 1 zero of φ1, but we can improve a length >k zero of φ1.

↓t=0Δtφ2=0φ1=0𝒞↓t=0ΔtE𝒞~π↓t=0Δt
Figure 3.5: Stable limit of family of LG quasimaps.

We will now study the general case. We will assume that S={f0,…,fN} all of weight 1. Then if we consider

This induces a cartesian diagram

Properness reduces to proving properness of the bottom morphism. Using the valuative criterion, consider a curve 𝒞∘→Δ∘ and an LG quasimap

If we have an extension of Φ⁡(ξ∘) to

we need a unique extension ξ of ξ∘ such that Φ⁡(ξ)=ξ¯.

Because the line bundle ω𝒞log comes from the coarse moduli space, we can trivialize it locally. Working locally, any LG quasimap becomes a usual quasimap. First assume that there are no base points (we will deal with base points using Hartogs’s theorem). We already have an extension to the coarse moduli space X¯.

Lemma 3.1.25 ([AV02]).

Locally, given 𝒞¯→X¯, there exists a unique commutative diagram

This completes the proof of properness.

3.2 Applications to Gromov-Witten theory

3.2.1 MSP invariants

We will consider MSP moduli spaces 𝒲g,n,d with d0=d, d∞=0, only (1,ρ) insertions, and arbitrary values of N. We first note that the MSP virtual localization formula is given by

where tα are the equivariant variables and V1α denotes those vertices at level 1 where the curve satisfies μα≠0 and μβ≠α=0. Also, 𝔞∈𝔞∞ are connected components of the level ∞ part of Θ and Ae is the contribution of an edge. In particular, define

These are the top degree part of the contribution to the virtual localization formula coming from a vertex v. We will denote the full contribution at level 1 by [ℳ¯g,n]α,tw. From now on, we will specialize our equivariant variables to roots of unity as tα=−ζNα⁢t. For convenience, we will also specialize t such that tN=−1.

We may define MSP invariants using virtual localization. Note that by the condition that ρ vanishes at the marked points, we have evaluation morphisms

which restrict to

where 𝒲Γ− is the degeneracy locus of 𝒲Γ. Therefore, we may define MSP invariants with insertions from the state space

Using the vertex contributions to the virtual normal bundle, we define the pairing

The state space has several bases, which we will discuss now.

  • •

    Let p=c1⁢(𝒪ℙ4+N⁢(1)) be the equivariant ambient hyperplane class. Then we have the basis ϕi=pi for i=0,…,N+3;

  • •

    There is the basis {𝟏Z5,H,H2,H3}∪{𝟏α}α=1N;

The last kind of MSP invariant we need to define is the MSP [0,1] invariant. Here, we simply consider the class

where Λ[0,1] denotes the set of all graphs without any level ∞ vertices.

3.2.2 Genus zero MSP theory

In genus zero, the full MSP and the [0,1] theory are equal. This follows from the following lemma:

Lemma 3.2.1.

We have

and an equality

of virtual cycles.

The lemma tells us that the genus-zero MSP invariants are the same as the GW invariants of a degree 5 hypersurface in ℙ4+N, which is in particular Fano. In particular, the MSP I-function is given by the formula

This automatically implies the following result.

Lemma 3.2.2.

We have

whenever N≥2.

The main result we need to know about genus zero MSP theory is the explicit form of the quantum connection. Let D≔q⁢ddq.

Lemma 3.2.3.

The MSP S-matrix satisfies the differential equation

where AM is given by the matrix

in the basis {ϕi} for N>5.

3.2.3 MSP [0,1] CohFT

Define the MSP R-matrix by the Birkhoff factorization

where

is defined using the quantum Riemann-Roch theorem. Here, we need to shift SZ to the point τZ5=I1I0⁢H, and

where

Here, L⁡(q)≔(1−55⁢q)1N.

Theorem 3.2.4.

The MSP [0,1] invariants come from a CohFT Ω[0,1], which is defined by the formula

Remark 3.2.5.

The normalized tail contribution at the isolated points is given by

In addition, when N≫3⁢g−3+n, there is no tail contribution at level 0.

3.2.4 Degree bound for MSP theory

In order to compute the invariants of a Calabi-Yau threefold using MSP theory, we need to control the MSP invariants. Our goal will be to control the MSP [0,1] invariants, but these are defined as a mysterious sum of virtual localization contributions. First, we will control the full MSP invariants.

Lemma 3.2.6.

The full MSP correlator

is a polynomial in q of degree at most

This follows from the fact that the virtual dimension of the MSP moduli space is N⁡(d+1−g)+n. To obtain the same degree bound for the [0,1] correlators, we will need a decomposition formula for the full MSP theory in terms of the [0,1] theory and the remaining contributions. First, we will construct bipartite graphs from localization graphs, where vertices are either [0,1] vertices or ∞ vertices. For an example of this procedure, see Figure 3.6.

214021005⟹349
Figure 3.6: Obtaining a bipartite graph from a localization graph.
Lemma 3.2.7.

We have the MSP decomposition formula

Here, the contribution Cont[v]∞ of a vertex v at level ∞ is a generating series of FJRW-like invariants, which is a polynomial in q of degree at most

In addition, Λ𝖻𝗂𝗉𝖺𝗋𝗍𝗂𝗍𝖾 is the set of stable bipartite graphs, Lv∘ is the set of legs which get contracted to v after stabilization, and c⁡(i) is the stable vertex that i gets contracted to after stabilization. For an example of stabilization, see Figure 3.7.

0v2,50v1,20v3,6ℓ1ℓ2ℓ3e1e2⟹v2,5v1,2v3,6ℓ1ℓ2ℓ3
Figure 3.7: Stabilization of a bipartite graph. In this example, we have c⁡(ℓ1)=(v1,e1), c⁡(ℓ2)=(e2,v2), and c⁡(ℓ3)=ℓ3.

This lemma is proved by directly applying the virtual localization formula and then analyzing the following two situations:

  • •

    What happens at a vertex at level ∞;

  • •

    What happens when we split a graph at a vertex at level 1.

By using the decomposition formula and a careful degree-counting argument, we obtain the following degree bound for the [0,1] theory.

Lemma 3.2.8.

The MSP [0,1] correlator

is a polynomial in q of degree at most

3.2.5 Polynomiality

We first introduce the ring of five generators. Let

and define the following generators:

Here, recall that I11=1+D⁡(I1I0).

Lemma 3.2.9 ([YY04]).

The ring

contains all Ak and Bk.

Theorem 3.2.10.

Introduce the series

Then Pg,n∈ℛ for all g,n such that 2⁢g−2+n>0.

If we want to prove this result using the results we have already proved, then we need to prove a polynomiality result for the the entries of the R-matrix. At level 0, we use the equation

and the explicit forms of the MSP quantum connection and the quantum connection for the quintic to obtain

To simplify what follows, define the normalized basis

where I22 was defined previously and I33=I11. If we define

then the recursive formula

where Cb=D⁢log⁡(I0⁢⋯⁢Ib⁢b)∈ℛ and cj=(0,…,0,120,770,1345,770), yields the following result:

Lemma 3.2.11.

If j≢b+k(modN), then (Rk)kb=0. Otherwise, we have (Rk)b+kb∈ℛ and Y⁢(Rk)b+N+kn∈ℛ.

At level 1, define the normalized basis 𝟏¯α=L−N+32⁢𝟏α. Then define

where Lα=−tα⁢L.

Lemma 3.2.12.

The quantity (Rk)jα is independent of α and is a polynomial in Y of degree at most k+⌊jN⌋.

The lemma is proved as follows:

  • •

    Fix the case when j=0 by using the Picard-Fuchs equation

    and an oscillating integral. Solving the Picard-Fuchs equation expresses Rk as an antiderivative of some polynomial in Rm<k and their derivatives. To prove that the constants of integration are zero, we use the fact that R⁡(z)⁢𝟏|ptα can be computed from the asymptotic expansion of the oscillating integral

    where

    is the mirror superpotential and γα is the Lefschetz thimble near the critical point of W corresponding to ptα.

  • •

    Use the recursion

    to induct on j.

Proof of Theorem 3.2.10.

First, note that we have the base cases P0,3=1 due to Zagier-Zinger [ZZ08] and

due to Zinger [Zin09]. The relation

implies that we only need to prove Pg≥2∈ℛ.

Consider the correlator (5⁢Y)g−1⁢⟨⟩g,0[0,1], which is a polynomial in Y of degree at most g−1. By the stable graph sum formula, we have

For all non-leading graphs, we use the relation ∑v(gv−1)+|E|=g−1 to assign powers of Y to all of the edges. Then the contributions from vertices are given as follows:

  • •

    At a level 0 vertex, the contributions are simply

    which reduces to Pgv,m by the string and dilaton equations.

  • •

    At a level 1 vertex, the contribution is

    After summing over all α, we see that this is nonzero only if the total power of tα is a multiple of N (here, we may want N to be a prime number).

Using the fact that the contribution from an edge between two level 1 vertices satisfies a balancing condition, the total factor of the Lα for the various α becomes 1. This implies that ContΓ∈ℛ for any non-leading Γ, so we must have Pg∈ℛ. □

Remark 3.2.13.

We can recover the genus one mirror theorem very quickly using the results we have already proved. If we consider the correlator

there are only two stable graphs. The contribution of the stable graph with a genus 1 vertex at the quintic is given by

The other graph contributes

at level 0. Finally, we can prove that the total contribution from level 1 is a degree 1 polynomial in Y, so using the known values of N1,1 and ⟨H⟩1,1,0Z fixes the two coefficients of Y.

Remark 3.2.14.

The formula

is expected to hold for all of the examples in Table 1.2, where we denote the Calabi-Yau threefold by Z. It has been proven for hypersurfaces in weighted projective space by the author [Lei24b].

3.2.6 Mirror symmetry picture

Recall that mirror symmetry is for families

On the A-side, we have bases ϕi∈H1,1⁢(Z) mirror to ei∈H2,1⁢(Z∨). On the B-side, near the maximal unipotent monodromy point, we have periods

Then we will have the mirror map given by

so we see that

Using the B-model basis, we define the Yukawa coupling

which is a rational function over ℳ¯.

Remark 3.2.15.

Physicists have more predictions about the Yukawa couplings, including pole orders.

In the case of the quintic Z5⊂ℙ4, there is only one Kähler parameter, and the Yukawa coupling is simply

For Z3,3⊂ℙ2×ℙ2, the Yukawa couplings are symmetric under permuting 1 and 2 and are given by

where the discriminant Δ is given by

In higher genus, we define

Conjecture 3.2.16 ([BCOV94]).

Define

where the lower order terms come from a sum over stable graphs. This fg,a→ is a rational function in q and after normalization by some power of the Yukawa couplings is a polynomial of degree at most 3⁢g−3+n.

Example 3.2.17.

In the case when g=2, the contributions to f2 come from the graps in Figure 3.8.

21111000000
Figure 3.8: Stable graphs of genus 2 with no marked points.

We put Si⁢j on each edge and at each vertex, we place

where we replace all e0 by ψ¯ and there is a correction term in genus 1 with only ancestor insertions.

This Feynman graph sum can be realized using geometric quantization. In particular, consider the matrix

which is explicitly defined using BCOV’s propagators [BCOV94]. This is symplectic with respect to the pairing

Therefore, we may define its geometric quantization by

where Q is a quadratic form given by

We will view R∈End⁡HB, where

It naturally acts on Givental’s symplectic vector space, but we can restrict it to a finite-dimensional vector space HA (which is still larger than HB). We will call it RA when acting on HA and RB when acting on HB.

Remark 3.2.18.

Note that Givental’s graph sum contains more information than BCOV’s graph sum coming from the extra basis elements.

Theorem 3.2.19.

Here, we restrict to the case of the quintic. Let ∙ denote either A or B. Denote

where we define

Then we have the identity

In other words, there is the commutative diagram

3.2.7 MSP realization of the Feynman rule

Recall that MSP for the quintic is given by the charges

Virtual localization gives us the fixed loci

where W=(x15+⋯+x55=0)⊂ℙ4+N. At the level of CohFTs, recall that we had an identity

One advantage of increasing N is that it makes the Dijkgraaf-Witten map τ (which sends I⁡(−z) onto the slice of ℒloc coming from H∗⁢(Z5)⊕H∗⁢(N⁢pt)) very nice. The actual values of τ on components were given in Section 3.2.3. If we compute the edge contribution V00 connecting two vertices at level 0, this will be exactly the propagators as N→∞.

Our goal is to use the polynomiality of Ω[0,1] to deduce polynomiality of the CohFT

We will consider the factorization

which factorizes

It now suffices to prove that 𝔾 preserves polynomiality, and in fact we see that the entries of 𝔾 lie in ℚ⁡[Y] with explicit degree bounds.

3.2.8 FJRW theory of the quintic

The unfortunate issue with the MSP moduli space of the quintic is that there are N vertices at level ∞. Instead, we will consider N copies of the field ν. To calculate FJRW invariants, we will set d0=0 and d∞=d. One nice feature of this is that the moduli space has no vertices at level 0, so we do not need to prove irregular vanishing.

The ambient space of the target is Tot⁡(𝒪ℙ⁡(1,…,1,5)⁢(−1)⊕5). Virtual localization will give N isolated points and the FJRW theory as fixed loci. When computing using MSP, there is a factorization R=Rloc⁢Δ, where Rloc comes from virtual localization and Δ comes from Grothendieck-Riemann-Roch. In this model, neither step preserves the unit, but their composition does preserve the unit, and therefore we obtain

3.2.9 Multi-parameter models

We will consider only a 1-parameter deformation of the GLSM for Z3,3⊂ℙ2×ℙ2. This is given by the charges

There is of course another deformation given by exchanging the roles of ℒ1 and ℒ2 with respect to the field μ. The fixed loci are now 𝖫𝖦, ℙ2, and Z3,3. We then have the identities

Note that Ω1[0,1] is polynomial only in q1 and Ω2[0,1] is polynomial only in q2. We then find that R1A=R2A, and so in fact we have

where 𝔾i is polynomial in qi (with no bound in the other Kähler parameter). The first identity gives polynomiality in q1 and the second gives polynomiality in q2.

d1d23⁢g−3
Figure 3.9: MSP bounds (gray) and sharp bounds (blue) for Z3,3⊂ℙ2×ℙ2.
Remark 3.2.20.

The bound obtained using MSP for this example is not sharp, for example see Figure 3.9.

Bibliography

  • [ACGS24]
    Dan Abramovich, Qile Chen, Mark Gross, and Bernd Siebert. Punctured logarithmic maps. 2024. arXiv: 2009.07720.
  • [AFKP23]
    Sergei Alexandrov, Soheyla Feyzbakhsh, Albrecht Klemm, and Boris Pioline. Quantum geometry and mock modularity. 2023. arXiv: 2312.12629.
  • [AL07]
    Murad Alim and Jean Dominique Länge. “Polynomial structure of the (open) topological string partition function”. In: J. High Energy Phys. 10 (2007), pp. 045, 13. issn: 1126-6708,1029-8479. doi: 10.1088/1126-6708/2007/10/045.
  • [Ale+24]
    Sergei Alexandrov, Soheyla Feyzbakhsh, Albrecht Klemm, Boris Pioline, and Thorsten Schimannek. “Quantum geometry, stability and modularity”. In: Commun. Number Theory Phys. 18.1 (2024), pp. 49–151. issn: 1931-4523,1931-4531. doi: 10.4310/cntp.2024.v18.n1.a2.
  • [Ara+24]
    Dhyan Aranha, Adeel A. Khan, Alexei Latyntsev, Hyeonjun Park, and Charanya Ravi. Virtual localization revisited. 2024. arXiv: 2207.01652.
  • [AV02]
    Dan Abramovich and Angelo Vistoli. “Compactifying the space of stable maps”. In: J. Amer. Math. Soc. 15.1 (2002), pp. 27–75. issn: 0894-0347,1088-6834. doi: 10.1090/S0894-0347-01-00380-0.
  • [BCOV94]
    M. Bershadsky, S. Cecotti, H. Ooguri, and C. Vafa. “Kodaira-Spencer theory of gravity and exact results for quantum string amplitudes”. In: Comm. Math. Phys. 165.2 (1994), pp. 311–427. issn: 0010-3616,1432-0916.
  • [BG83]
    Robert L. Bryant and Phillip A. Griffiths. “Some observations on the infinitesimal period relations for regular threefolds with trivial canonical bundle”. In: Arithmetic and geometry, Vol. II. Vol. 36. Progr. Math. Birkhäuser Boston, Boston, MA, 1983, pp. 77–102. isbn: 3-7643-3133-X.
  • [BMT14]
    Arend Bayer, Emanuele Macrì, and Yukinobu Toda. “Bridgeland stability conditions on threefolds I: Bogomolov-Gieseker type inequalities”. In: J. Algebraic Geom. 23.1 (2014), pp. 117–163. issn: 1056-3911,1534-7486. doi: 10.1090/S1056-3911-2013-00617-7.
  • [CCLT09]
    Tom Coates, Alessio Corti, Yuan-Pin Lee, and Hsian-Hua Tseng. “The quantum orbifold cohomology of weighted projective spaces”. In: Acta Math. 202.2 (2009), pp. 139–193. issn: 0001-5962,1871-2509. doi: 10.1007/s11511-009-0035-x.
  • [CG07]
    Tom Coates and Alexander Givental. “Quantum Riemann-Roch, Lefschetz and Serre”. In: Ann. of Math. (2) 165.1 (2007), pp. 15–53. issn: 0003-486X,1939-8980. doi: 10.4007/annals.2007.165.15.
  • [CGL19]
    Huai-Liang Chang, Shuai Guo, and Jun Li. BCOV’s Feynman rule of quintic 3-folds. 2019. arXiv: 1810.00394.
  • [CGL21]
    Huai-Liang Chang, Shuai Guo, and Jun Li. “Polynomial structure of Gromov-Witten potential of quintic 3-folds”. In: Ann. of Math. (2) 194.3 (2021), pp. 585–645. issn: 0003-486X,1939-8980. doi: 10.4007/annals.2021.194.3.1.
  • [CGLL21]
    Huai-Liang Chang, Shuai Guo, Jun Li, and Wei-Ping Li. “The theory of N-mixed-spin-P fields”. In: Geom. Topol. 25.2 (2021), pp. 775–811. issn: 1465-3060,1364-0380. doi: 10.2140/gt.2021.25.775.
  • [CGLZ20]
    Huai-Liang Chang, Shuai Guo, Wei-Ping Li, and Jie Zhou. “Genus-one Gromov-Witten invariants of quintic three-folds via MSP localization”. In: Int. Math. Res. Not. IMRN 19 (2020), pp. 6347–6390. issn: 1073-7928,1687-0247. doi: 10.1093/imrn/rny201.
  • [Cha+23a]
    Huai-Liang Chang, Shuai Guo, Jun Li, Wei-Ping Li, and Yang Zhou. A boundedness theorem for principal bundles on curves. 2023. arXiv: 2312.11197.
  • [Cha+23b]
    Huai-Liang Chang, Shuai Guo, Jun Li, Wei-Ping Li, and Yang Zhou. Mixed-Spin-P fields for GIT quotients. 2023. arXiv: 2312.11174.
  • [Chi06]
    Alessandro Chiodo. “The Witten top Chern class via K-theory”. In: J. Algebraic Geom. 15.4 (2006), pp. 681–707. issn: 1056-3911,1534-7486. doi: 10.1090/S1056-3911-06-00444-9.
  • [CJRa]
    Qile Chen, Felix Janda, and Yongbin Ruan. “Structural formulae in logarithmic Gauged Linear Sigma Models I: The tropical decomposition formula”. In preparation.
  • [CJRb]
    Qile Chen, Felix Janda, and Yongbin Ruan. “Structural formulae in logarithmic Gauged Linear Sigma Models III: The localization formula”. In preparation.
  • [CJR21]
    Qile Chen, Felix Janda, and Yongbin Ruan. “The logarithmic gauged linear sigma model”. In: Invent. Math. 225.3 (2021), pp. 1077–1154. issn: 0020-9910,1432-1297. doi: 10.1007/s00222-021-01044-2.
  • [CJR23]
    Qile Chen, Felix Janda, and Yongbin Ruan. Punctured logarithmic R-maps. 2023. arXiv: 2208.04519.
  • [CJW21]
    Qile Chen, Felix Janda, and Rachel Webb. “Virtual cycles of stable (quasi-)maps with fields”. In: Adv. Math. 385 (2021), Paper No. 107781, 49. issn: 0001-8708,1090-2082. doi: 10.1016/j.aim.2021.107781.
  • [CK20]
    Ionuţ Ciocan-Fontanine and Bumsig Kim. “Quasimap wall-crossings and mirror symmetry”. In: Publ. Math. Inst. Hautes Études Sci. 131 (2020), pp. 201–260. issn: 0073-8301,1618-1913. doi: 10.1007/s10240-020-00114-0.
  • [CKL17]
    Huai-Liang Chang, Young-Hoon Kiem, and Jun Li. “Torus localization and wall crossing for cosection localized virtual cycles”. In: Adv. Math. 308 (2017), pp. 964–986. issn: 0001-8708,1090-2082. doi: 10.1016/j.aim.2016.12.019.
  • [CL12a]
    Huai-Liang Chang and Jun Li. “Gromov-Witten invariants of stable maps with fields”. In: Int. Math. Res. Not. IMRN 18 (2012), pp. 4163–4217. issn: 1073-7928,1687-0247. doi: 10.1093/imrn/rnr186.
  • [CL12b]
    Kevin J. Costello and Si Li. Quantum BCOV theory on Calabi-Yau manifolds and the higher genus B-model. 2012. arXiv: 1201.4501 [math.QA].
  • [CL20a]
    Huai-Liang Chang and Jun Li. “A vanishing associated with irregular MSP fields”. In: Int. Math. Res. Not. IMRN 20 (2020), pp. 7347–7396. issn: 1073-7928,1687-0247. doi: 10.1093/imrn/rnaa049.
  • [CL20b]
    Huai-Liang Chang and Mu-Lin Li. “Invariants of stable quasimaps with fields”. In: Trans. Amer. Math. Soc. 373.5 (2020), pp. 3669–3691. issn: 0002-9947,1088-6850. doi: 10.1090/tran/8011.
  • [Cla17]
    Emily Clader. “Landau-Ginzburg/Calabi-Yau correspondence for the complete intersections X3,3 and X2,2,2,2”. In: Adv. Math. 307 (2017), pp. 1–52. issn: 0001-8708,1090-2082. doi: 10.1016/j.aim.2016.11.010.
  • [CLL15]
    Huai-Liang Chang, Jun Li, and Wei-Ping Li. “Witten’s top Chern class via cosection localization”. In: Invent. Math. 200.3 (2015), pp. 1015–1063. issn: 0020-9910,1432-1297. doi: 10.1007/s00222-014-0549-5.
  • [CLLL19]
    Huai-Liang Chang, Jun Li, Wei-Ping Li, and Melissa Chiu-Chu Liu. “Mixed-Spin-P fields of Fermat polynomials”. In: Cambridge Journal of Mathematics 7.3 (2019), pp. 319–364. issn: 2168-0930. doi: 10.4310/CJM.2019.v7.n3.a3.
  • [CLLL22]
    Huai-Liang Chang, Jun Li, Wei-Ping Li, and Chiu-Chu Melissa Liu. “An effective theory of GW and FJRW invariants of quintic Calabi-Yau manifolds”. In: J. Differential Geom. 120.2 (2022), pp. 251–306. issn: 0022-040X,1945-743X. doi: 10.4310/jdg/1645207466.
  • [COGP92]
    Philip Candelas, Xenia C. de la Ossa, Paul S. Green, and Linda Parkes. “A pair of Calabi-Yau manifolds as an exactly soluble superconformal theory”. In: Essays on mirror manifolds. Int. Press, Hong Kong, 1992, pp. 31–95. isbn: 962-7670-01-4.
  • [CT24]
    Andrei Caldararu and Junwu Tu. Effective Categorical Enumerative Invariants. 2024. arXiv: 2404.01499 [math.AG]. url: https://arxiv.org/abs/2404.01499.
  • [Del71]
    Pierre Deligne. “Théorie de Hodge. II”. In: Inst. Hautes Études Sci. Publ. Math. 40 (1971), pp. 5–57. issn: 0073-8301,1618-1913.
  • [DH98]
    Igor V. Dolgachev and Yi Hu. “Variation of geometric invariant theory quotients”. In: Inst. Hautes Études Sci. Publ. Math. 87 (1998). With an appendix by Nicolas Ressayre, pp. 5–56. issn: 0073-8301,1618-1913.
  • [Dub96]
    Boris Dubrovin. “Geometry of 2D topological field theories”. In: Integrable systems and quantum groups (Montecatini Terme, 1993). Vol. 1620. Lecture Notes in Math. Springer, Berlin, 1996, pp. 120–348. isbn: 3-540-60542-8. doi: 10.1007/BFb0094793.
  • [ES96]
    Geir Ellingsrud and Stein Arild Strømme. “Bott’s formula and enumerative geometry”. In: J. Amer. Math. Soc. 9.1 (1996), pp. 175–193. issn: 0894-0347,1088-6834. doi: 10.1090/S0894-0347-96-00189-0.
  • [FJR18]
    Huijun Fan, Tyler J. Jarvis, and Yongbin Ruan. “A mathematical theory of the gauged linear sigma model”. In: Geom. Topol. 22.1 (2018), pp. 235–303. issn: 1465-3060,1364-0380. doi: 10.2140/gt.2018.22.235.
  • [FK21]
    David Favero and Bumsig Kim. General GLSM Invariants and Their Cohomological Field Theories. 2021. arXiv: 2006.12182.
  • [FL19]
    Honglu Fan and Yuan-Pin Lee. “Towards a quantum Lefschetz hyperplane theorem in all genera”. In: Geom. Topol. 23.1 (2019), pp. 493–512. issn: 1465-3060,1364-0380. doi: 10.2140/gt.2019.23.493.
  • [FLY08]
    Hao Fang, Zhiqin Lu, and Ken-Ichi Yoshikawa. “Analytic torsion for Calabi-Yau threefolds”. In: J. Differential Geom. 80.2 (2008), pp. 175–259. issn: 0022-040X,1945-743X.
  • [Giv01a]
    Alexander B. Givental. “Gromov-Witten invariants and quantization of quadratic Hamiltonians”. In: vol. 1. 4. Dedicated to the memory of I. G. Petrovskii on the occasion of his 100th anniversary. 2001, pp. 551–568, 645. doi: 10.17323/1609-4514-2001-1-4-551-568.
  • [Giv01b]
    Alexander B. Givental. “Semisimple Frobenius structures at higher genus”. In: Internat. Math. Res. Notices 23 (2001), pp. 1265–1286. issn: 1073-7928,1687-0247. doi: 10.1155/S1073792801000605.
  • [Giv04]
    Alexander B. Givental. “Symplectic geometry of Frobenius structures”. In: Frobenius manifolds. Vol. E36. Aspects Math. Friedr. Vieweg, Wiesbaden, 2004, pp. 91–112. isbn: 3-528-03206-5.
  • [Giv96]
    Alexander B. Givental. “Equivariant Gromov-Witten invariants”. In: Internat. Math. Res. Notices 13 (1996), pp. 613–663. issn: 1073-7928,1687-0247. doi: 10.1155/S1073792896000414.
  • [Giv98]
    Alexander Givental. “Elliptic Gromov-Witten invariants and the generalized mirror conjecture”. In: Integrable systems and algebraic geometry (Kobe/Kyoto, 1997). World Sci. Publ., River Edge, NJ, 1998, pp. 107–155. isbn: 981-02-3266-7.
  • [GJR17]
    Shuai Guo, Felix Janda, and Yongbin Ruan. A mirror theorem for genus two Gromov-Witten invariants of quintic threefolds. 2017. arXiv: 1709.07392.
  • [GJR18]
    Shuai Guo, Felix Janda, and Yongbin Ruan. Structure of Higher Genus Gromov-Witten Invariants of Quintic 3-folds. 2018. arXiv: 1812.11908.
  • [GKMW07]
    Thomas W. Grimm, Albrecht Klemm, Marcos Mariño, and Marlene Weiss. “Direct integration of the topological string”. In: J. High Energy Phys. 8 (2007), pp. 058, 78. issn: 1126-6708,1029-8479. doi: 10.1088/1126-6708/2007/08/058.
  • [GP99]
    T. Graber and R. Pandharipande. “Localization of virtual classes”. In: Invent. Math. 135.2 (1999), pp. 487–518. issn: 0020-9910,1432-1297. doi: 10.1007/s002220050293.
  • [GR19]
    Shuai Guo and Dustin Ross. “Genus-one mirror symmetry in the Landau-Ginzburg model”. In: Algebr. Geom. 6.3 (2019), pp. 260–301. issn: 2313-1691,2214-2584. doi: 10.1007/s00013-019-01351-4.
  • [Gri68a]
    Phillip A. Griffiths. “Periods of Integrals on Algebraic Manifolds, I. (Construction and Properties of the Modular Varieties)”. In: American Journal of Mathematics 90.2 (1968), pp. 568–626. issn: 00029327, 10806377. (Visited on 01/16/2025).
  • [Gri68b]
    Phillip A. Griffiths. “Periods of Integrals on Algebraic Manifolds, II: (Local Study of the Period Mapping)”. In: American Journal of Mathematics 90.3 (1968), pp. 805–865. issn: 00029327, 10806377. (Visited on 01/16/2025).
  • [GS09]
    Josh Guffin and Eric Sharpe. “A-twisted Landau–Ginzburg models”. In: Journal of Geometry and Physics 59.12 (2009), pp. 1547–1580. issn: 0393-0440. doi: https://doi.org/10.1016/j.geomphys.2009.07.014.
  • [HKQ09]
    M.-x. Huang, A. Klemm, and S. Quackenbush. “Topological string theory on compact Calabi-Yau: modularity and boundary conditions”. In: Homological mirror symmetry. Vol. 757. Lecture Notes in Phys. Springer, Berlin, 2009, pp. 45–102. isbn: 978-3-540-86374-8.
  • [HKTY95]
    S. Hosono, A. Klemm, S. Theisen, and S.-T. Yau. “Mirror symmetry, mirror map and applications to Calabi-Yau hypersurfaces”. In: Comm. Math. Phys. 167.2 (1995), pp. 301–350. issn: 0010-3616,1432-0916.
  • [Kat86]
    Sheldon Katz. “On the finiteness of rational curves on quintic threefolds”. In: Compositio Math. 60.2 (1986), pp. 151–162. issn: 0010-437X,1570-5846.
  • [KL13]
    Young-Hoon Kiem and Jun Li. “Localizing virtual cycles by cosections”. In: J. Amer. Math. Soc. 26.4 (2013), pp. 1025–1050. issn: 0894-0347,1088-6834. doi: 10.1090/S0894-0347-2013-00768-7.
  • [KL18]
    Bumsig Kim and Hyenho Lho. “Mirror theorem for elliptic quasimap invariants”. In: Geom. Topol. 22.3 (2018), pp. 1459–1481. issn: 1465-3060,1364-0380. doi: 10.2140/gt.2018.22.1459.
  • [KM94]
    M. Kontsevich and Yu. Manin. “Gromov-Witten classes, quantum cohomology, and enumerative geometry”. In: Comm. Math. Phys. 164.3 (1994), pp. 525–562. issn: 0010-3616,1432-0916.
  • [KM98]
    M. Kontsevich and Yu. Manin. “Relations between the correlators of the topological sigma-model coupled to gravity”. In: Comm. Math. Phys. 196.2 (1998), pp. 385–398. issn: 0010-3616,1432-0916. doi: 10.1007/s002200050426.
  • [KO22]
    Bumsig Kim and Jeongseok Oh. “Localized Chern characters for 2-periodic complexes”. In: Selecta Math. (N.S.) 28.2 (2022), Paper No. 23, 26. issn: 1022-1824,1420-9020. doi: 10.1007/s00029-021-00743-1. url: https://doi.org/10.1007/s00029-021-00743-1.
  • [Kon92]
    Maxim Kontsevich. “Intersection theory on the moduli spaces of curves and the matrix Airy function”. In: Commun. Math. Phys. 147.1 (1992), pp. 1–23.
  • [Kon95]
    Maxim Kontsevich. “Enumeration of rational curves via torus actions”. In: The moduli space of curves (Texel Island, 1994). Vol. 129. Progr. Math. Birkhäuser Boston, Boston, MA, 1995, pp. 335–368. isbn: 0-8176-3784-2. doi: 10.1007/978-1-4612-4264-2_12.
  • [Lei24a]
    Patrick Lei. Higher genus Gromov-Witten theory of one-parameter Calabi-Yau threefolds II: Feynman rule and anomaly equations. 2024. arXiv: 2412.06527.
  • [Lei24b]
    Patrick Lei. Higher-genus Gromov-Witten theory of one-parameter Calabi-Yau threefolds I: Polynomiality. 2024. arXiv: 2409.11659.
  • [Li01]
    Jun Li. “Stable morphisms to singular schemes and relative stable morphisms”. In: J. Differential Geom. 57.3 (2001), pp. 509–578. issn: 0022-040X,1945-743X.
  • [Li11]
    Si Li. BCOV theory on the elliptic curve and higher genus mirror symmetry. 2011. arXiv: 1112.4063 [math.QA].
  • [Liu24]
    Zhiyu Liu. Castelnuovo bound for curves in projective 3-folds. 2024. arXiv: 2407.20161.
  • [LLY97]
    Bong H. Lian, Kefeng Liu, and Shing-Tung Yau. “Mirror principle. I”. In: Asian J. Math. 1.4 (1997), pp. 729–763. issn: 1093-6106,1945-0036. doi: 10.4310/AJM.1997.v1.n4.a5.
  • [LP18]
    Hyenho Lho and Rahul Pandharipande. “Stable quotients and the holomorphic anomaly equation”. In: Adv. Math. 332 (2018), pp. 349–402. issn: 0001-8708,1090-2082. doi: 10.1016/j.aim.2018.05.020.
  • [LR22]
    Zhiyu Liu and Yongbin Ruan. Castelnuovo bound and higher genus Gromov-Witten invariants of quintic 3-folds. 2022. arXiv: 2210.13411.
  • [LZ09]
    Jun Li and Aleksey Zinger. “On the genus-one Gromov-Witten invariants of complete intersections”. In: J. Differential Geom. 82.3 (2009), pp. 641–690. issn: 0022-040X,1945-743X.
  • [MNOP06a]
    Davesh Maulik, Nikita Nekrasov, Andrei Okounkov, and Rahul Pandharipande. “Gromov-Witten theory and Donaldson-Thomas theory, I”. In: Compos. Math. 142.5 (2006), pp. 1263–1285.
  • [MNOP06b]
    Davesh Maulik, Nikita Nekrasov, Andrei Okounkov, and Rahul Pandharipande. “Gromov-Witten theory and Donaldson-Thomas theory, II”. In: Compos. Math. 142.5 (2006), pp. 1286–1304.
  • [MP06]
    D. Maulik and R. Pandharipande. “A topological view of Gromov-Witten theory”. In: Topology 45.5 (2006), pp. 887–918. issn: 0040-9383. doi: 10.1016/j.top.2006.06.002.
  • [OP10]
    A. Okounkov and R. Pandharipande. “The local Donaldson-Thomas theory of curves”. In: Geom. Topol. 14.3 (2010), pp. 1503–1567. issn: 1465-3060,1364-0380. doi: 10.2140/gt.2010.14.1503.
  • [Par24]
    John Pardon. Universally counting curves in Calabi–Yau threefolds. 2024. arXiv: 2308.02948.
  • [Pic21]
    Renata Picciotto. Moduli of stable maps with fields. 2021. arXiv: 2009.04385.
  • [Pop13]
    Alexandra Popa. “The genus one Gromov-Witten invariants of Calabi-Yau complete intersections”. In: Trans. Amer. Math. Soc. 365.3 (2013), pp. 1149–1181. issn: 0002-9947,1088-6850. doi: 10.1090/S0002-9947-2012-05550-4.
  • [PP17]
    Rahul Pandharipande and Aaron Pixton. “Gromov-Witten/pairs correspondence for the quintic 3-fold”. In: J. Amer. Math. Soc. 30 (2017), pp. 389–449.
  • [PPZ15]
    Rahul Pandharipande, Aaron Pixton, and Dimitri Zvonkine. “Relations on ℳ¯g,n via 3-spin structures”. In: J. Amer. Math. Soc. 28.1 (2015), pp. 279–309. issn: 0894-0347,1088-6834. doi: 10.1090/S0894-0347-2014-00808-0.
  • [PV16]
    Alexander Polishchuk and Arkady Vaintrob. “Matrix factorizations and cohomological field theories”. In: J. Reine Angew. Math. 714 (2016), pp. 1–122. issn: 0075-4102,1435-5345. doi: 10.1515/crelle-2014-0024.
  • [Tel12]
    Constantin Teleman. “The structure of 2D semi-simple field theories”. In: Invent. Math. 188 (2012), pp. 525–588.
  • [Tha96]
    Michael Thaddeus. “Geometric invariant theory and flips”. In: J. Amer. Math. Soc. 9.3 (1996), pp. 691–723. issn: 0894-0347,1088-6834. doi: 10.1090/S0894-0347-96-00204-4.
  • [Tia87]
    Gang Tian. “Smoothness of the universal deformation space of compact Calabi-Yau manifolds and its Petersson-Weil metric”. In: Mathematical aspects of string theory (San Diego, Calif., 1986). Vol. 1. Adv. Ser. Math. Phys. World Sci. Publishing, Singapore, 1987, pp. 629–646. isbn: 9971-50-273-9; 9971-50-274-7.
  • [Tod89]
    Andrey N. Todorov. “The Weil-Petersson geometry of the moduli space of S⁢U⁢(n≥3) (Calabi-Yau) manifolds. I”. In: Comm. Math. Phys. 126.2 (1989), pp. 325–346. issn: 0010-3616,1432-0916.
  • [VZ08]
    Ravi Vakil and Aleksey Zinger. “A desingularization of the main component of the moduli space of genus-one stable maps into ℙn”. In: Geom. Topol. 12.1 (2008), pp. 1–95. issn: 1465-3060,1364-0380. doi: 10.2140/gt.2008.12.1.
  • [Wan20]
    Jun Wang. A mirror theorem for Gromov-Witten theory without convexity. 2020. arXiv: 1910.14440.
  • [Wit92]
    Edward Witten. “Mirror manifolds and topological field theory”. In: Essays on mirror manifolds. Int. Press, Hong Kong, 1992, pp. 120–158. isbn: 962-7670-01-4.
  • [Yau78]
    Shing Tung Yau. “On the Ricci curvature of a compact Kähler manifold and the complex Monge-Ampère equation. I”. In: Comm. Pure Appl. Math. 31.3 (1978), pp. 339–411. issn: 0010-3640,1097-0312. doi: 10.1002/cpa.3160310304.
  • [YY04]
    Satoshi Yamaguchi and Shing-Tung Yau. “Topological string partition functions as polynomials”. In: J. High Energy Phys. 7 (2004), pp. 047, 20. issn: 1126-6708,1029-8479. doi: 10.1088/1126-6708/2004/07/047.
  • [Zho22]
    Yang Zhou. “Quasimap wall-crossing for GIT quotients”. In: Invent. Math. 227.2 (2022), pp. 581–660. issn: 0020-9910,1432-1297. doi: 10.1007/s00222-021-01071-z.
  • [Zin09]
    Aleksey Zinger. “The reduced genus 1 Gromov-Witten invariants of Calabi-Yau hypersurfaces”. In: J. Amer. Math. Soc. 22.3 (2009), pp. 691–737. issn: 0894-0347,1088-6834. doi: 10.1090/S0894-0347-08-00625-5.
  • [Zon16]
    Zhengyu Zong. Equivariant Gromov-Witten Theory of GKM Orbifolds. 2016. arXiv: 1604.07270.
  • [ZZ08]
    Don Zagier and Aleksey Zinger. “Some properties of hypergeometric series associated with mirror symmetry”. In: Modular forms and string duality. Vol. 54. Fields Inst. Commun. Amer. Math. Soc., Providence, RI, 2008, pp. 163–177. isbn: 978-0-8218-4484-7.