## Abstract

We find at least 527 new four-dimensional Fano manifolds, each of which is a complete intersection in a smooth toric Fano manifold.

### 1. Introduction

Fano manifolds are the basic building blocks of algebraic geometry, both in the sense of the Minimal Model Program [1–4] and as the ultimate source of most explicit examples and constructions. There are finitely many deformation families of Fano manifolds in each dimension [5]. There is precisely 1 one-dimensional Fano manifold: the line; there are 10 deformation families of two-dimensional Fano manifolds: the del Pezzo surfaces and there are 105 deformation families of three-dimensional Fano manifolds [6–13]. Very little is known about the classification of Fano manifolds in higher dimensions.

In this paper, we begin to explore the geography of Fano manifolds in dimension 4. Four-dimensional Fano manifolds of higher Fano index have been classified [6,14–22]—there are 35 in total—but the most interesting case, where the Fano variety has index 1, is wide open. We use computer algebra to find many four-dimensional Fano manifolds that arise as complete intersections in toric Fano manifolds in codimension at most 4. We find at least 738 examples, 717 of which have Fano index 1 and 527 of which are new.

Suppose that *Y* is a toric Fano manifold and that *L*_{1},…,*L*_{c} are nef line bundles on *Y* such that −*K*_{Y}−*Λ* is ample, where *Λ*=*c*_{1}(*L*_{1})+⋯+*c*_{1}(*L*_{c}). Let *X*⊂*Y* be a smooth complete intersection defined by a regular section of ⊕_{i}*L*_{i}. The Adjunction Formula gives that

*X*is Fano. We find all four-dimensional Fano manifolds

*X*of this form such that the codimension

*c*is at most 4.

Our interest in this problem is motivated by a program to classify Fano manifolds in higher dimensions using mirror symmetry [23]. For each four-dimensional Fano manifold *X* as above, therefore, we compute the essential ingredients for this program: the quantum period and regularized quantum differential equation associated to *X*, and a Laurent polynomial *f* that corresponds to *X* under mirror symmetry; we also calculate basic geometric data about *X*, the ambient space *Y* and *f*. The results of our computations in machine-readable form, together with full details of our implementation and all source code used, can be found in the electronic supplementary material.

### 2. Finding four-dimensional fano toric complete intersections

Our method is as follows. Toric Fano manifolds *Y* are classified up to dimension 8 by Batyrev *et al.*^{1} [24–29]. For each toric Fano manifold *Y* of dimension *d*=4+*c*, we:

(i) compute the nef cone of | |||||

(ii) find all $\mathit{\Lambda}\in {H}^{2}(Y;\mathbb{Z})$ such that | |||||

(iii) decompose |

Each such decomposition determines a four-dimensional Fano manifold *X*⊂*Y* , defined as the zero locus of a regular section of the vector bundle ⊕_{i}*L*_{i}. To compute the nef cone in step (i), we consider dual exact sequences

where the map *ρ* is defined by the *N* rays of a fan *Σ* for *Y* . There are canonical identifications ${\mathbb{L}}^{\star}\cong {H}^{2}(Y;\mathbb{Z})\cong \mathrm{Pic}(Y)$, and the nef cone of *Y* is the intersection of cones

*D*

_{i}is the image under

*D*of the

*i*th standard basis vector in ${({\mathbb{Z}}^{N})}^{\star}$ [30, proposition 15.1.3]. The classes

*Λ*in step (ii) are the lattice points in the polyhedron

*P*=NC(

*Y*)∩(−

*K*

_{Y}−NC(

*Y*)) such that −

*K*

_{Y}−

*Λ*lies in the interior of NC(

*Y*). Since NC(

*Y*) is a strictly convex cone,

*P*is compact and the number of lattice points in

*P*is finite. We implement step (iii) by first expressing

*Λ*as a sum of Hilbert basis elements in NC(

*Y*) in all possible ways

*b*

_{i}may be repeated; this is a knapsack-style problem. We then, for each decomposition (2.1), partition the

*b*

_{i}into

*c*subsets

*S*

_{1},…,

*S*

_{c}in all possible ways and define the line bundle

*L*

_{i}to be the sum of the classes in

*S*

_{i}.

We found 117 173 distinct triples (*X*;*Y* ;*L*_{1},…,*L*_{c}), with a total of 17 934 distinct ambient toric varieties *Y* . Note that the representation of a given Fano manifold *X* as a toric complete intersection is far from unique: for example, if *X* is a complete intersection in *Y* given by a section of *L*_{1}⊕⋯⊕*L*_{c} then it is also a complete intersection in $Y\times {\mathbb{P}}^{1}$ given by a section of ${\pi}_{1}^{\star}{L}_{1}\oplus \cdots \oplus {\pi}_{1}^{\star}{L}_{c}\oplus {\pi}_{2}^{\star}{\mathcal{O}}_{{\mathbb{P}}^{1}}(1)$. Thus, we have found far fewer than 117 173 distinct four-dimensional Fano manifolds. We show below, by calculating quantum periods of the Fano manifolds *X*, that we find at least 738 non-isomorphic Fano manifolds. Since the quantum period is a very strong invariant—indeed no examples of distinct Fano manifolds *X*≇*X*′ with the same quantum period *G*_{X}=*G*_{X′} are known—we believe that we found precisely 738 non-isomorphic Fano manifolds. Eliminating the quantum periods found in [31], we see that at least 527 of our examples are new.

### Remark 2.1

There exist Fano manifolds which do not occur as complete intersections in toric Fano manifolds. But in low dimensions, most Fano manifolds arise this way: 8 of the 10 del Pezzo surfaces, and at least 78 of the 105 smooth three-dimensional Fano manifolds, are complete intersections in toric Fano manifolds [32].

### Remark 2.2

It may be the case that any *d*-dimensional Fano manifold which occurs as a toric complete intersection in fact occurs as a toric complete intersection in codimension *d*; we know of no counterexamples. But even if this holds in dimension 4, our search will probably not find all four-dimensional Fano manifolds which occur as toric complete intersections. This is because, if one of the line bundles *L*_{i} involved is nef but not ample, then the Kähler cone for *X* can be strictly bigger than the Kähler cone for *Y* . In other words, it is possible for −*K*_{X} to be ample on *X* even if −*K*_{Y}−*Λ* is not ample on *Y* . For an explicit example of this in dimension 3, see [32, §55].

### 3. Quantum periods and mirror laurent polynomials

The quantum period *G*_{X} of a Fano manifold *X* is a generating function

*c*

_{d}of

*X*which plays an important role in mirror symmetry. A precise definition can be found in [32, §B], but roughly speaking one can think of

*c*

_{d}as the ‘virtual number’ of rational curves

*C*in

*X*that pass through a given point, satisfy certain constraints on their complex structure, and satisfy 〈−

*K*

_{X},

*C*〉=

*d*. The quantum period is discussed in detail in [23,32]; for us what will be important is that the regularized quantum period

*regularized quantum differential equation*of

*X*:

*p*

_{m}are polynomials and

*D*=

*t*(d/d

*t*).

It has been proposed that Fano manifolds should correspond under mirror symmetry to Laurent polynomials which are *extremal* or of *low ramification* [23], in the sense discussed in §4. An *n*-dimensional Fano manifold *X* is said to be *mirror-dual* to a Laurent polynomial $f\in \mathbb{C}[{x}_{1}^{\pm 1},\dots ,{x}_{n}^{\pm 1}]$ if the regularized quantum period of *X* coincides with the classical period of *f*:

*X*is mirror-dual to the Laurent polynomial

*f*then the regularized quantum differential equation (3.3) for

*X*coincides with the Picard–Fuchs differential equation satisfied by

*π*

_{f}. The correspondence between Fano manifolds and Laurent polynomials is not one-to-one, but it is expected that any two Laurent polynomials

*f*,

*g*that are mirror-dual to the same Fano manifold are related by a birational transformation $\phi :{({\mathbb{C}}^{\times})}^{n}\u21e2{({\mathbb{C}}^{\times})}^{n}$ called a

*mutation*or a

*symplectomorphism of cluster type*[33–35]:

*φ*

^{⋆}

*f*=

*g*. We will write such a mutation as $f\stackrel{\phi}{\u21e2}g$. Mutations are known to preserve the classical period [33]: if $f\stackrel{\phi}{\u21e2}g$ then

*π*

_{f}=

*π*

_{g}.

### Remark 3.1

In the paragraphs above we discuss *the* regularized quantum differential equation and *the* Picard–Fuchs differential equation. This involves choices of normalization. Our conventions are that the regularized quantum differential operator is the operator *L*_{X} as in (3.3) such that:

(i) the order, | |||||

(ii) the degree of | |||||

(iii) the leading coefficient of | |||||

(iv) the coefficients of the polynomials |

The Picard–Fuchs differential operator is the differential operator *L*_{f} such that:

*P*

_{m}are polynomials and

*D*=

*t*(d/d

*t*), and that the analogues of conditions (i)–(iv) above hold.

We determined the quantum period *G*_{X}, for each of the triples (*X*;*Y* ;*L*_{1},…,*L*_{c}) from §2, as follows. For each such triple we found, using the Mirror Theorem for toric complete intersections [36] and a generalization of a technique due to V. Przyjalkowski, a Laurent polynomial *f* that is mirror-dual to *X*. This process is described in detail in §5. We then computed, for each triple, the first 20 terms of the power series expansion of ${\hat{G}}_{X}={\pi}_{f}$ using the Taylor expansion

*α*

_{d}is the coefficient of the unit monomial in

*f*

^{d}. We divided the 117 173 triples into 738 ‘buckets’, according to the value of the first 20 terms of the power series expansion of ${\hat{G}}_{X}={\pi}_{f}$, and then proved that any two Fano manifolds

*X*,

*X*′ in the same bucket have the same quantum period by exhibiting a chain of mutations $f\stackrel{{\phi}_{0}}{\u21e2}{f}_{1}\stackrel{{\phi}_{1}}{\u21e2}\cdots \stackrel{{\phi}_{n-1}}{\u21e2}{f}_{n}\stackrel{{\phi}_{n}}{\u21e2}g$ that connects the Laurent polynomials

*f*and

*g*mirror-dual to

*X*and

*X*′.

For each quantum period *G*_{X}, we computed the quantum differential operator *L*_{X} directly from the mirror Laurent polynomial *f* chosen above, using Lairez’s generalized Griffiths–Dwork algorithm [37]. The output from Lairez’s algorithm is a differential operator $L=\sum _{m=0}^{N}{P}_{m}(t){D}^{m}$ with ${P}_{0},\dots ,{P}_{N}\in \mathbb{Q}[t]$ such that, with very high probability, *Lπ*_{f}≡0. Such an operator *L* gives a recurrence relation for the Taylor coefficients *α*_{0},*α*_{1},*α*_{2}… of *π*_{f}; using this recurrence relation and the first 20 Taylor coefficients computed above, we solved for the first 2000 Taylor coefficients *α*_{k}. We then consider an operator

*π*

_{f}give 2000 linear equations for the coefficients of the polynomials ${\overline{P}}_{m}$ and, provided that $(\overline{N}+1)(\overline{R}+1)\ll 2000$, this linear system is highly over-determined. Since we are looking for the Picard–Fuchs differential operator (see remark 3.1), we may assume that $(\overline{N},\overline{R})$ is lexicographically less than $(N,\mathrm{deg}{p}_{N})$. We searched systematically for such differential operators with $(\overline{N}+1)(\overline{R}+1)\ll 2000$, looking for the operator $\overline{L}$ with lexicographically minimal $(\overline{N},\overline{R})$ and clearing denominators so that the analogues of conditions (iii) and (iv) in remark 3.1 holds. We can say with high confidence that this operator $\overline{L}$ is in fact the Picard–Fuchs operator

*L*

_{f}, although this is not proven—partly because Lairez’s algorithm relies on a randomized interpolation scheme that is not guaranteed to produce an operator annihilating

*π*

_{f}, and partly because if

*L*

_{f}were to involve polynomials

*P*

_{m}of extremely large degree, 2000 terms of the Taylor expansion of

*π*

_{f}will not be enough to detect

*L*

_{f}. The operators $\overline{L}$ that we found satisfy a number of delicate conditions that act as consistency checks: for example they are of Fuchsian type (which is true for

*L*

_{f}, as

*L*

_{f}arises geometrically from a variation of Hodge structure). Thus, we are confident that $\overline{L}={L}_{f}$ in every case.

^{2}Since ${\hat{G}}_{X}={\pi}_{f}$ and

*L*

_{X}=

*L*

_{f}by construction, this determines, with high confidence, the quantum period

*G*

_{X}and the regularized quantum differential operator

*L*

_{X}.

### Remark 3.2

The use of Laurent polynomials and Lairez’s algorithm is essential here. There is a closed formula [32, corollary D.5] for the quantum period of the Fano manifolds that we consider, and one could in principle use this together with the linear algebra calculation described above to compute (a good candidate for) the regularized quantum differential operator *L*_{X}. In practice, however, for many of the examples that we treat here, it is impossible to determine enough Taylor coefficients from the formula: the computations involved are well beyond the reach of current hardware, both in terms of memory consumption and runtime. By contrast, our approach using mirror symmetry and Lairez’s algorithm will run easily on a desktop PC.

### Remark 3.3

The regularized quantum differential equation for *X* coincides with the (unregularized) quantum differential equation for an anticanonical Calabi–Yau manifold *Z*⊂*X*. The study of the regularized quantum period from this point of view was pioneered by Batyrev *et al.* [39,40] and an extensive study of fourth-order Calabi–Yau differential operators was made in [41]. We found 26 quantum differential operators with *N*=4; these coincide with or are equivalent to the fourth-order Calabi–Yau differential operators with AESZ IDs 1, 3, 4, 5, 6, 15, 16, 17, 18, 19, 20, 21, 22, 23, 34, 369, 370 and 424 in the Calabi–Yau Operators Database [42], together with one new fourth-order Calabi–Yau differential operator (which corresponds to our period sequence with ID 469).

### 4. Ramification data

Consider now one of our regularized quantum differential operators

### Definition 4.1 ([23])

Let $S\subset {\mathbb{P}}^{1}$ be a finite set and $\mathbb{V}\to {\mathbb{P}}^{1}\setminus S$ a local system. Fix a basepoint $x\in {\mathbb{P}}^{1}\setminus S$. For *s*∈*S*, choose a small loop that winds once anticlockwise around *s* and connect it to *x* via a path, thereby making a loop *γ*_{s} about *s* based at *x*. Let ${T}_{s}:{\mathbb{V}}_{x}\to {\mathbb{V}}_{x}$ denote the monodromy of $\mathbb{V}$ along *γ*_{s}. The *ramification* of $\mathbb{V}$ is

The ramification $\mathrm{rf}(\mathbb{V})$ is independent of the choices of basepoint *x* and of small loops *γ*_{s}. A non-trivial, irreducible local system $\mathbb{V}\to {\mathbb{P}}^{1}\setminus S$ has $\mathrm{rf}(\mathbb{V})\ge 2\mathrm{rk}(\mathbb{V})$: see [23, §2].

### Definition 4.2

Let $\mathbb{V}\to {\mathbb{P}}^{1}\setminus S$ be a local system as above. The *ramification defect* of $\mathbb{V}$ is the quantity $\mathrm{rf}(\mathbb{V})-2\mathrm{rk}(\mathbb{V})$. A local system of ramification defect zero is called *extremal*.

### Definition 4.3

The *ramification* (respectively, *ramification defect*) of a differential operator *L*_{X} is the ramification (respectively, ramification defect) of the local system of solutions *L*_{X}*f*≡0.

To compute the ramification of *L*_{X}, we proceed as in [31]. One can compute Jordan normal forms of the local $\mathrm{log}$-monodromies $\{\mathrm{log}{T}_{s}:s\in S\}$ using linear algebra over a splitting field *k* for *p*_{N}(*t*). (Every singular point of *L*_{X} is defined over *k*.) This is classical, going back to Birkhoff [43], as corrected by Gantmacher [44, vol. 2, §10] and Turrittin [45]; a very convenient presentation can be found in the book of Kedlaya [46, §7.3]. In practice, we use the symbolic implementation of $\overline{\mathbb{Q}}$ provided by the computational algebra system Magma [47,48]. We computed ramification data for 575 of the 738 regularized quantum differential operators, finding ramification defects as shown in table 1; this lends some support to the conjecture, due to Golyshev [23], that a Laurent polynomial *f* which is mirror-dual to a Fano manifold should have a Picard–Fuchs operator *L*_{f} that is extremal or of low ramification. For the remaining 163 regularized quantum differential operators, the symbol *p*_{N}(*t*) contains a factor of extremely high degree. This makes the computation of ramification data prohibitively expensive.

ramification defect | 0 | 1 | 2 | 3 |

number of occurrences | 92 | 290 | 167 | 26 |

### 5. The przyjalkowski method

We now explain, given complete intersection data (*X*;*Y* ;*L*_{1},…,*L*_{c}) as in §2, how to find a Laurent polynomial *f* that is mirror-dual to *X*. This is a slight generalization of a technique that we learned from V. Przyjalkowski^{3} [50,51], and which is based on the mirror theorems for toric complete intersections due to Givental [36] and Hori *et al.* [52]. Recall the exact sequence

from §2 and the elements ${D}_{i}\in {\mathbb{L}}^{\star}$, 1≤*i*≤*N*, defined by the standard basis elements of ${({\mathbb{Z}}^{N})}^{\star}$. Recall further that ${\mathbb{L}}^{\star}\cong \mathrm{Pic}(Y)$, so that each line bundle *L*_{m} defines a class in ${\mathbb{L}}^{\star}$. Suppose that there exists a choice of disjoint subsets *E*,*S*_{1},…,*S*_{c} of {1,2,…,*N*} such that:

— { | |||||

— each | |||||

— $\sum _{k\in {S}_{m}}{D}_{k}={L}_{m}$ for each |

and choose distinguished elements *s*_{m}∈*S*_{m}, 1≤*m*≤*c*. Set ${S}_{m}^{\circ}={S}_{m}\setminus \{{s}_{m}\}$. Writing the map *D* in terms of the standard basis for ${({\mathbb{Z}}^{N})}^{\star}$ and the basis {*D*_{j}:*j*∈*E*} for ${\mathbb{L}}^{\star}$ defines an (*N*−*d*)×*N* matrix (*m*_{ji}) of integers. Let (*x*_{1},…,*x*_{N}) denote the standard coordinates on ${({\mathbb{C}}^{\times})}^{N}$, let *r*=*N*−*d*, and define *q*_{1},…,*q*_{r} and *F*_{1},…,*F*_{c} by

*et al.*[52] have shown that

*W*=

*x*

_{1}+⋯+

*x*

_{N}and

*Γ*is a certain cycle in the submanifold of ${({\mathbb{C}}^{\times})}^{N}$ defined by

*y*

_{i}for $i\in \bigcup _{m=1}^{c}{S}_{m}^{\circ}$, setting

*q*

_{1}=⋯=

*q*

_{r}=1 to eliminate the variables

*x*

_{j},

*j*∈

*E*, allows us to write

*W*−

*c*as a Laurent polynomial

*f*in the variables

*f*is mirror-dual to

*X*.

The Laurent polynomial *f* produced by Przyjalkowski’s method depends on our choices of *E*,*S*_{1},…,*S*_{c}, and *s*_{1},…,*s*_{c}, but up to mutation this is not the case:

### Theorem 5.1 ([53])

*Let Y be a toric Fano manifold and let L*_{1}*,…,L*_{c} *be nef line bundles on Y such that −K*_{Y}*−Λ is ample, where Λ=c*_{1}(*L*_{1})+⋯+*c*_{1}(*L*_{c}). *Let X⊂Y be a smooth complete intersection defined by a regular section of ⊕*_{i}*L*_{i}*. Let f and g be Laurent polynomial mirrors to X obtained by applying Przyjalkowski’s method to* (*X;Y ;L*_{1}*,…,L*_{c}) *as above, but with possibly different choices for the subsets E,S*_{1}*,…,S*_{c} *and the elements s*_{1}*,…,s*_{c}*. Then there exists a mutation φ such that* $f\stackrel{\phi}{\u21e2}g$.

### Example 5.2

Let *Y* be the projectivization of the vector bundle ${\mathcal{O}}^{\oplus 2}\oplus \mathcal{O}{(1)}^{\oplus 2}$ over ${\mathbb{P}}^{2}$. Choose a basis for the two-dimensional lattice ${\mathbb{L}}^{\star}$ such that the matrix (*m*_{ji}) of the map *D* is

*L*

_{1}→

*Y*defined by the element $(2,1)\in {\mathbb{L}}^{\star}$, and the Fano hypersurface

*X*⊂

*Y*defined by a regular section of

*L*

_{1}. Applying Przyjalkowski’s method to the triple (

*X*;

*Y*;

*L*

_{1}) with

*E*={3,4},

*S*

_{1}={1,2,5} and

*s*

_{1}=1 yields the Laurent polynomial

*X*. Applying the method with

*E*={3,4},

*S*

_{1}={1,6} and

*s*

_{1}=1 yields

### Remark 5.3

Observe that, for a complete intersection of dimension *n* and codimension *c*, Przyjalkowski’s method requires partitioning *n*+*c* variables into *c* disjoint subsets. If (*n*+*c*)/*c*<2 then at least one of the subsets must have size one and so the corresponding variable, *x*_{j} say, is eliminated from the Laurent polynomial via the equation *x*_{j}=1. One could therefore have obtained the resulting Laurent polynomial from a complete intersection with smaller codimension: new Laurent polynomials are found only when (*n*+*c*)/*c*≥2, that is, when the codimension is at most the dimension. In particular, all possible mirrors to four-dimensional Fano toric complete intersections given by the Przyjalkowski method occur for complete intersections in toric manifolds of dimension at most 8.

### 6. Examples

#### (a) The cubic fourfold

Let *X* be the cubic fourfold. This arises in our classification from the complete intersection data (*X*;*Y* ;*L*) with $Y={\mathbb{P}}^{5}$ and $L={\mathcal{O}}_{{\mathbb{P}}^{5}}(3)$. The Przyjalkowski method yields [54, §2.1] a Laurent polynomial

*X*, and elementary calculation gives

*L*

_{X}

*g*≡0 are

*L*

_{X}is extremal.

#### (b) A (3,3) complete intersection in ${\mathbb{P}}^{6}$

Let *X* be a complete intersection in $Y={\mathbb{P}}^{6}$ of type (3,3). This arises in our classification from the complete intersection data (*X*;*Y* ;*L*_{1},*L*_{2}) with ${L}_{1}={L}_{2}={\mathcal{O}}_{{\mathbb{P}}^{6}}(3)$. The Przyjalkowski method yields a Laurent polynomial

*X*, and [32, corollary D.5] gives

*L*

_{X}is

*L*

_{X}

*g*≡0 are

*L*

_{X}is extremal.

### 7. Results and analysis

We close by indicating how basic numerical invariants—degree and size of cohomology—vary across the 738 families of Fano manifolds that we have found. The degree (−*K*_{X})^{4} varies from 5 to 800, as shown in figures 1 and 2.

We do not have direct access to the size of the cohomology algebra of our Fano manifolds *X*, as many of the line bundles occurring in the complete intersection data (*X*;*Y* ;*L*_{1},…,*L*_{c}) are not ample and so the Lefschetz theorem need not apply. But the order *N* of the regularized quantum differential operator is a good proxy for the size of the cohomology. *N* is the rank of a certain local system—an irreducible piece of the Fourier–Laplace transform of the restriction of the Dubrovin connection (in the Frobenius manifold given by the quantum cohomology of *X*) to the line in *H*^{•(X)} spanned by −*K*_{X}—and in the case where this local system is irreducible, which is typical, *N* will coincide with the dimension of *H*^{•}(*X*). For our examples, *N* lies in the set {4,6,8,10,12}. Figure 3 shows how *N* varies with the degree (−*K*_{X})^{4}, with darker grays indicating a larger number of examples with that *N* and degree.

The isolated example on the right of figure 3, with *N*=6 and degree 800, is the blow up of $\mathbb{P}(1,1,1,1,3)$ at a point. Figure 4 again shows how *N* varies with the degree (−*K*_{X})^{4}, but this time with toric Fano manifolds highlighted in red. Figure 5 shows how the Euler number *χ*(*T*_{X}) varies with the degree (−*K*_{X})^{4}, with darker grays indicating a larger number of examples with that Euler number and degree. The three examples with the largest Euler number *χ* are a quintic hypersurface in ${\mathbb{P}}^{5}$, with *χ*=825; a complete intersection of type (2,4) in ${\mathbb{P}}^{6}$, with *χ*=552; and a complete intersection of type (3,3) in ${\mathbb{P}}^{6}$, with *χ*=369. The three examples with the most negative Euler number are ${\mathbb{P}}^{1}\times {V}_{4}^{3}$ where ${V}_{4}^{3}$ is a quartic hypersurface in ${\mathbb{P}}^{4}$, with *χ*=−112; ${\mathbb{P}}^{1}\times {V}_{6}^{3}$ where ${V}_{6}^{3}$ is a complete intersection of type (2,3) in ${\mathbb{P}}^{5}$, with *χ*=−72; and ${\mathbb{P}}^{1}\times {V}_{8}^{3}$ where ${V}_{8}^{3}$ is a complete intersection of type (2,2,2) in ${\mathbb{P}}^{6}$, with *χ*=−48.

### Ethics statement

This research did not involve human or animal subjects.

### Data accessibility

The Electronic Supplementary Material contains the results of our computations, in machine readable form, together with full source code written in Magma [47]. See the files called README.txt for details. The source code and data, but not the text of this paper, are released under a Creative Commons CC0 license (*Creative Commons CC0 license*, https://creativecommons.org/publicdomain/zero/1.0/ and https://creativecommons.org/publicdomain/zero/1.0/legalcode.): see the files called COPYING.txt for details. If you make use of the source code or data in an academic or commercial context, you should acknowledge this by including a reference or citation to this paper.

## Acknowledgements

The computations underlying this work were performed using the Imperial College High Performance Computing Service and the compute cluster at the Department of Mathematics, Imperial College London. We thank Simon Burbidge, Matt Harvey and Andy Thomas for valuable technical assistance. We thank John Cannon and the Computational Algebra Group at the University of Sydney for providing licenses for the computer algebra system Magma. We thank Alessio Corti for a number of very useful conversations, Pierre Lairez for explaining his generalized Griffiths–Dwork algorithm and sharing his code with us, and Duco van Straten for his analysis of the regularized quantum differential operators with *N*=4.

### Author contributions

T.C., T.P. and A.K. performed the research and wrote the paper.

### Funding statement

This research was supported by a

### Competing interests

We have no competing interests.

## Footnotes

1 Øbro’s method, which applies in all dimensions, starts with the observation that without loss of generality one cone in the fan for *Y* is the positive orthant and the sum of the primitive generators *v*_{1},…,*v*_{N} for the rays of the fan lies in this cone; he then uses the facts that *Y* is smooth and Fano to give bounds on the coordinates of the *v*_{i}, and defines an order on such fans with the property that if *F*_{1}<*F*_{2} then the toric Fano varieties defined by *F*_{1} and *F*_{2} are non-isomorphic. By constructing the fans in increasing order, one never needs to check for fan isomorphism, and thus one never needs to refer back to fans already found.

3 Przyjalkowski informs us that he learned this, for the case of the cubic threefold, from Katzarkov [49] and Orlov (V. Przyjalkowski 2014, personal communication).

### References

- 1
Reid M . 2002 Update on 3-folds.**In Proc. Int. Congress of Mathematicians**, Beijing, 2002, vol., pp. 513–524 Beijing, China: Higher Ed. Press. Google Scholar**II** - 2
Corti A (ed.) 2007**Flips for 3-folds and 4-folds**. Oxford Lecture Series in Mathematics and its Applications, vol.. Oxford, UK: Oxford University Press. Crossref, Google Scholar**35** - 3
Birkar C, Cascini P, Hacon CD& McKernan J . 2010 Existence of minimal models for varieties of log general type.**J. Am. Math. Soc.**, 405–468. (doi:10.1090/S0894-0347-09-00649-3). Crossref, ISI, Google Scholar**23** - 4
Hacon CD& McKernan J . 2010 Existence of minimal models for varieties of log general type. II.**J. Am. Math. Soc.**, 469–490. (doi:10.1090/S0894-0347-09-00651-1). Crossref, Google Scholar**23** - 5
Kollár J, Miyaoka Y& Mori S . 1992 Rational connectedness and boundedness of Fano manifolds.**J. Differ. Geom.**, 765–779. Crossref, ISI, Google Scholar**36** - 6
- 7
- 8
Iskovskih VA . 1979**Anticanonical models of three-dimensional algebraic varieties**. Current problems in mathematics, vol. 12, pp. 59–157. (Russian) Moscow, Russia: VINITI. (239) (loose errata). Google Scholar - 9
Mori S& Mukai S . 1981/82 Classification of Fano 3-folds with B_{2}≥ 2.**Manuscr. Math.**, 147–162. (doi:10.1007/BF01170131). Crossref, ISI, Google Scholar**36** - 10
Mori S& Mukai S . 1983 On Fano 3-folds with B_{2}≥ 2.**Algebraic varieties and analytic varieties (Tokyo, 1981), Adv. Stud. Pure Math.**, vol., pp. 101–129. Amsterdam, The Netherlands: North-Holland. Google Scholar**1** - 11
Mori S& Mukai S . 1986 Classification of Fano 3-folds with B_{2}≥ 2.**Algebraic and topological theories (Kinosaki, 1984)**pp. 496–545. Tokyo, Japan: Kinokuniya. Google Scholar - 12
Mori S& Mukai S . 2003 Erratum: ‘Classification of Fano 3-folds with B2 ≥ 2’ (Manuscr. Math. 36 (1981/82), no. 2, 147-162).**Manuscr. Math**., 407. (doi:10.1007/s00229-002-0336-2). Crossref, ISI, Google Scholar**110** - 13
Mori S& Mukai S . 2004**Extremal rays and Fano 3-folds**pp. 37–50. The Fano Conf., University Torino, Turin. Google Scholar - 14
Shokurov VV . 1985 A nonvanishing theorem.**Izv. Akad. Nauk SSSR Ser. Mat.**, 635–651. Google Scholar**49** - 15
Kobayashi S& Ochiai T . 1973 Characterizations of complex projective spaces and hyperquadrics.**J. Math. Kyoto Univ.**, 31–47. Crossref, Google Scholar**13** - 16
Kollár Y . 1981 Higher-dimensional Fano varieties of large index.**Vestnik Moskov. Univ. Ser. I Mat. Mekh.**, 80–81. Google Scholar**31–34** - 17
Serpico ME . 1980 Fano varieties of dimensions n ≥ 4 and of index r ≥ n-1.**Rend. Sem. Mat. Univ. Padova**,, 295–308. Google Scholar**62** - 18
Fujita T . 1980 On the structure of polarized manifolds with total deficiency one.**I. J. Math. Soc. Japan**, 709–725. (doi:10.2969/jmsj/03240709). Crossref, ISI, Google Scholar**32** - 19
Fujita T . 1981 On the structure of polarized manifolds with total deficiency one.**II. J. Math. Soc. Japan**, 415–434. (doi:10.2969/jmsj/03330415). Crossref, Google Scholar**33** - 20
Fujita T . 1984 On the structure of polarized manifolds with total deficiency one.**III J. Math. Soc. Japan**, 75–89. (doi:10.2969/jmsj/03610075). Crossref, ISI, Google Scholar**36** - 21
Fujita T . 1990**Classification theories of polarized varieties**. London Mathematical Society Lecture Note Series, vol.. Cambridge, UK: Cambridge University Press. Crossref, Google Scholar**155** - 22
Iskovskikh VA& Prokhorov YuG . 1999**Fano varieties**Algebraic geometry, V, Encyclopaedia*Math. Sci.*, vol., pp. 1–247. Berlin, Germany: Springer. Google Scholar**47** - 23
Coates T, Corti A, Galkin S, Golyshev V& Kasprzyk AM . 2014 Mirror symmetry and Fano manifolds.**European Congress of Mathematics Kraków, 2–7 July 2012**285–300. Google Scholar - 24
Batyrev VV . 1981 Toric Fano threefolds.**Izv. Akad. Nauk SSSR Ser. Mat.**, 704–717. 927. Google Scholar**45** - 25
Watanabe K& Watanabe M . 1982 The classification of Fano 3-folds with torus embeddings.**Tokyo J. Math.**, 37–48. (doi:10.3836/tjm/1270215033). Crossref, Google Scholar**5** - 26
Batyrev VV . 1999 On the classification of toric Fano 4-folds.**J. Math. Sci. (New York)**, 1021–1050. (doi:10.1007/BF02367245). Crossref, Google Scholar**94** - 27
Sato H . 2000 Toward the classification of higher-dimensional toric Fano varieties.**Tohoku Math. J.**, 383–413. (doi:10.2748/tmj/1178207820). Crossref, ISI, Google Scholar**52** - 28
Kreuzer M& Nill B . 2009 Classification of toric Fano 5-folds.**Adv. Geom.**, 85–97. (doi:10.1515/ADVGEOM.2009.005). Crossref, ISI, Google Scholar**9** - 29
Øbro M . 2007 An algorithm for the classification of smooth Fano polytopes. (http://arxiv.org/abs/0704.0049). Google Scholar - 30
Cox DA, Little JB& Schenck HK . 2011**Toric varieties**Graduate Studies in Mathematics, vol.. Providence, RI: American Mathematical Society. Crossref, Google Scholar**124** - 31
Coates T, Galkin S, Kasprzyk AM& Strangeway A . 2014 Quantum periods for certain four-dimensional Fano manifolds. (http://arxiv.org/abs/1406.4891). Google Scholar - 32
Coates T, Corti A, Galkin S& Kasprzyk AM . 2013 Quantum periods for 3-dimensional Fano manifolds. (http://arxiv.org/abs/1303.3288). Google Scholar - 33
Akhtar M, Coates T, Galkin S& Kasprzyk AM . 2012 Minkowski polynomials and mutations.**SIGMA Symmetry Integrability Geom. Methods Appl.**, Paper 094, 17. (doi:10.3842/SIGMA.2012.094). Google Scholar**8** - 34
- 35
Katzarkov L& Przyjalkowski V . 2012 Landau-Ginzburg models—old and new.**Proc. Gökova Geometry-Topology Conf. 2011**, pp. 97–124. Somerville, MA: Int. Press. Google Scholar - 36
Givental A . 1998 A mirror theorem for toric complete intersections.**Topological field theory, primitive forms and related topics (Kyoto, 1996)**Progr. Math., vol., pp. 141–175. Boston, MA: Birkhäuser Boston. Google Scholar**160** - 37
Lairez P . 2014 Computing periods of rational integrals. (http://arxiv.org/abs/1404.5069) Source code available from https://github.com/lairez/periods. Google Scholar - 38
van Hoeij M . 1997 Factorization of differential operators with rational functions coefficients.**J. Symb. Comput.**, 537–561. (doi:10.1006/jsco.1997.0151). Crossref, ISI, Google Scholar**24** - 39
Batyrev VV, Ciocan-Fontanine I, Kim B& van Straten D . 1998 Conifold transitions and mirror symmetry for Calabi-Yau complete intersections in Grassmannians.**Nuclear Phys. B**, 640–666. (doi:10.1016/S0550-3213(98)00020-0). Crossref, ISI, Google Scholar**514** - 40
Batyrev VV, Ciocan-Fontanine I, Kim B& van Straten D . 2000 Mirror symmetry and toric degenerations of partial flag manifolds.**Acta Math.**, 1–39. (doi:10.1007/BF02392780). Crossref, ISI, Google Scholar**184** - 41
Almkvist G, van Enckevort C, van Straten D& Zudilin W . 2005 Tables of Calabi–Yau equations. (arXiv:math/0507430 [math.AG]). Google Scholar - 42
van Straten D . 2005**Calabi-Yau Operators Database**online, access via http://www.mathematik.uni-mainz.de/CYequations/db/. Google Scholar - 43
Birkhoff GD . 1913 Equivalent singular points of ordinary linear differential equations.**Math. Ann.**, 134–139. (doi:10.1007/BF01455347). Crossref, Google Scholar**74** - 44
Gantmacher FR . 1959**The theory of matrices**vols. 1. 2 (translated by KA Hirsch). New York, NY: Chelsea Publishing Co. Google Scholar - 45
Turrittin HL . 1955 Convergent solutions of ordinary linear homogeneous differential equations in the neighborhood of an irregular singular point.**Acta Math.**, 27–66. (doi:10.1007/BF02392519). Crossref, ISI, Google Scholar**93** - 46
Kedlaya KS . 2010**p-adic differential equations**Cambridge Studies in Advanced Mathematics, vol.. Cambridge, UK: Cambridge University Press. Crossref, Google Scholar**125** - 47
Bosma W, Cannon J& Playoust C . 1997 The Magma algebra system. I. The user language.**J. Symb. Comput.**, 235–265. (doi:10.1006/jsco.1996.0125). Crossref, ISI, Google Scholar**24** - 48
Steel AK . 2010 Computing with algebraically closed fields.**J. Symb. Comput.**, 342–372. (doi:10.1016/j.jsc.2009.09.005). Crossref, ISI, Google Scholar**45** - 49
Katzarkov L . 2009 Homological mirror symmetry and algebraic cycles.**Homological mirror symmetry**Lecture Notes in Phys., vol., pp. 125–152. Berlin, Germany: Springer. Google Scholar**757** - 50
Przyjalkowski V . 2011 Hori-Vafa mirror models for complete intersections in weighted projective spaces and weak Landau-Ginzburg models.**Cent. Eur. J. Math.**, 972–977. (doi:10.2478/s11533-011-0070-7). Crossref, Google Scholar**9** - 51
Przhiyalkovskiĭ VV . 2013 Weak Landau-Ginzburg models of smooth Fano threefolds.**Izv. Ross. Akad. Nauk Ser. Mat.**, 135–160. Google Scholar**77** - 52
Hori K, Katz S, Klemm A, Pandharipande R, Thomas R, Vakil R, Vafa C& Zaslow E . 2003**Mirror symmetry**. Clay Mathematics Monographs, vol.. Providence, RI: American Mathematical Society; Clay Mathematics Institute, Cambridge, MA. With a preface by Vafa. Google Scholar**1** - 53
- 54
Ilten NO, Lewis J& Przyjalkowski V . 2013 Toric degenerations of Fano threefolds giving weak Landau-Ginzburg models.**J. Algebra**, 104–121. (doi:10.1016/j.jalgebra.2012.11.002). Crossref, ISI, Google Scholar**374**