1: \documentclass[a4paper]{article}
2:
3: \usepackage[latin1]{inputenc}
4: \usepackage{amsmath}
5: \usepackage{rotating}
6: \usepackage{calc}
7: \usepackage{graphics}
8: \usepackage{color}
9:
10: \usepackage[pdftitle={Monodromy calculatons of fourth order equations
11: of Calabi-Yau type}, pdfauthor={Christian van Enckevort and Duco
12: van Straten},ps2pdf]{hyperref}
13: \newcommand{\hlink}[1]{\href{#1}{\texttt{}#1}}
14: \newcommand{\dburl}{http://enriques.mathematik.uni-mainz.de/enckevort/db}
15: \newcommand{\eqnno}[2][4]{\href{\dburl/summary.php?order=#1\&eqnno=#2}{#2}}
16:
17: \usepackage{array}
18: \newcolumntype{C}{>{$}c<{$}}
19: \newcolumntype{L}{>{$}l<{$}}
20: \newcolumntype{R}{>{$}r<{$}}
21:
22: \usepackage{myref}
23: \usepackage{mysymb}
24:
25: \newcommand{\tcdot}{\text{$\cdot$}}
26: \newcommand{\eqnbox}[1]{\parbox{0.7\textwidth}{\vspace{0.3\baselineskip}%
27: #1\vspace{0.3\baselineskip}}}
28:
29: \newcommand{\Td}{\mathrm{Td}}
30: \newcommand{\Id}{\mathrm{Id}}
31: \newcommand{\Grass}{\mathrm{Grass}}
32: \newcommand{\LGrass}{\mathrm{LGrass}}
33: \renewcommand{\P}{\mathbb{P}}
34: \renewcommand{\L}{{\mathbb{L}}}
35: \newcommand{\lra}{{\longrightarrow}}
36: \newcommand{\Auteq}{\mathop{\mathrm{Auteq}}}
37:
38: \newtheorem{hyp}{Hypothesis}
39:
40: %\renewcommand{\H}{\mathbf{H}}
41: %\newcommand{\Proj}{\mathbb{P}}
42: %\newcommand{\Aut}{\mathop{\mathrm{Aut}}}
43: %\newcommand{\Pic}{\mathrm{Pic}}
44:
45: \begin{document}
46: \title{Monodromy calculations of fourth order equations of Calabi-Yau
47: type}
48: \author{Christian van Enckevort and Duco van Straten}
49: \maketitle
50:
51: {\renewcommand{\thefootnote}{}\footnotetext{AMS classification: 14J32
52: (Primary) 32S40, 81T30 (Secondary)}}
53:
54: \begin{abstract}
55: This paper contains a preliminary study of the monodromy of certain
56: fourth order differential equations, that were called of Calabi-Yau
57: type in \cite{AZ}. Some of these equations can be interpreted as
58: the Picard-Fuchs equations of a Calabi-Yau manifold with one complex
59: modulus, which links up the observed integrality to the conjectured
60: integrality of the Gopakumar-Vafa invariants. A natural question is
61: if in the other cases such a geometrical interpretation is also
62: possible. Our investigations of the monodromies are intended as a
63: first step in answering this question. We use a numerical approach
64: combined with some ideas from homological mirror symmetry to
65: determine the monodromy for some further one-parameter models.
66: Furthermore, we present a conjectural identification of the
67: Picard-Fuchs equation for 5 new examples from Borceas list and
68: conjecture the existence of some new Calabi-Yau three folds. The
69: paper does not contain any theorems or proofs but is, we think,
70: nevertheless of interest.
71: \end{abstract}
72:
73: \section{Introduction}
74: A differential operator of order $n$ on $\P^1$ has the form
75: \begin{equation}
76: \label{eq:L}
77: L:= a_n(z) \frac{d^n}{dz^n} + a_{n-1}(z)\frac{d^{n-1}}{dz^{n-1}} +
78: \dots + a_0(z),
79: \end{equation}
80: where the $a_i(z)$ are polynomials. The set $\Sigma \subset \P^1$ of
81: singular points is given by the zeros of $a_n(z)$ and possibly
82: $z=\infty$. The solutions to the equation $L y=0$ can be considered
83: as a $\C$-local system $\L$ of rank $n$ on $S:=\P^1\setminus \Sigma$.
84: After the choice of a base point $s \in \P^1 \setminus \Sigma$, the
85: information of $\L$ is given by the monodromy representation
86: \[
87: \pi_1(S,s) \lra \Aut(\L_s) = \mathrm{Gl}_n(\C)
88: \]
89: A power series $y_0(x) \in \Z[[x]]$ that satisfies a homogeneous
90: linear differential equation as above is a G-function and a folklore
91: conjecture that goes back to Bombieri and Dwork states that all such
92: power series and differential operators have a \emph{geometrical
93: origin} (see \cite{KZ}). This means that the operator should occur
94: as a \emph{factor} of a Picard-Fuchs operator describing the variation
95: of a cohomology of a family $\rho: \mathcal{Y} \lra \P^1$, with
96: singular fibres over $\Sigma$ and defined over a number field. The
97: local system $\L$ should then be a summand of a local system
98: $\L_{\C}:=R^d\rho_*(\C_{\mathcal{Y}})_{|S}$, where $d$ is the complex
99: dimension of the fibres of $\rho$. It follows among other things that
100: the equation has regular singularities with all exponents rational. It
101: can be shown that the set of power series of geometric origin in this
102: sense is closed under the ordinary product of power series and under
103: the coefficientwise Hadamard product of series. On the level of local
104: systems, the Cauchy product corresponds to the tensor product, whereas
105: the Hadamard product correspond to the convolution of local systems.
106: We refer to the books \cite{An} and \cite{Ka} for details.
107:
108: The fourth order equations in \cite{AZ} were collected with a stricter
109: notion of geometrical origin in mind: by requiring that the operator
110: admits an invariant symplectic form and gives rise to integral
111: instanton numbers, it starts making sense asking for the existence of
112: a one-parameter family $\mathcal{Y} \lra \P^1$ of \emph{Calabi-Yau
113: three folds}, whose associated Picard-Fuchs operator for
114: $\Hgrp^3(Y_s)$ is the given one. The instanton numbers then should
115: have the interpretation of counting curves on a mirror manifold $X$
116: with Picard number one. The first $14$ equations in the list are in
117: fact the much studied hypergeometric cases (see \cite{CdOGP},
118: \cite{M1}, \cite{KT}, \cite{BvS}, \cite{V}, \cite{DM}). Mirror pairs of
119: Calabi-Yau threefolds obtained from Batyrev's polar duality of
120: reflexive polytopes \cite{Ba} yield a plethora of examples but usually
121: with high Picard number (see \cite{CYH}). By taking restrictions to
122: carefully chosen one-dimensional sub-loci these examples sometimes
123: give rise to equations of Calabi-Yau type, but the instanton numbers
124: computed in this way represent \emph{sums over different homology
125: classes} and there will not exist a Calabi-Yau three fold $X$ with
126: Picard number one with the given instanton numbers. Case $15$ is an
127: example of this phenomenon: it is the equation belonging to the
128: diagonal restriction of Calabi-Yau family in $\P^3 \times \P^3$ (see
129: \cite{BvS}). The list contains many more of such examples. The question
130: is how can one see this from the differential equation alone.
131:
132: In order to find the cases that are potentially of strict geometric
133: origin, we remark that a geometrical local system $\L_{\C}$ carries a
134: integral lattice $\L_{\Z} =R^d\rho_*(\Z_{\mathcal{Y}})|_S$ and that
135: Poincaré-duality provides it with a \emph{unimodular} pairing
136: $\langle\cdot,\cdot\rangle$, which in our case is alternating. Hence
137: the monodromy representation is in the symplectic group $\Sp(4,\Z)$.
138: For differential equations of hypergeometric type, the monodromy
139: representation is explicitly known, essentially because the associated
140: local system is rigid (Levelt's theorem, \cite{BH},\cite{Ka}). This
141: leads to the $14$ hypergeometric cases mentioned above. For equations
142: with three singular points which are not of hypergeometric type or for
143: equations with more than three singular points, the monodromy
144: representation is in general not determined by local data alone and we
145: have the problem of accesory parameters. We do not know of any general
146: method to determine the monodromy representation in such cases. We
147: use a brute force numerical approach combined with ideas from
148: homological mirror symmetry to conjecturally determine the monodromy
149: for some further one-parameter models.
150:
151: \textbf{Acknowledgements}: We would like to thank G.~Almkvist,
152: C.~Doran, A.~Klemm, and W.~Zudilin for their interest in the project.
153: In particular we thank C.~Doran for his explanation of the integral
154: basis and A.~Klemm for the suggestion of using the genus one instanton
155: numbers as an extra integrality check and for explanations on higher
156: genus computations.
157:
158: \section{Sketch of Homological Mirror Symmetry}
159: According to Kontsevich \cite{Ko1}, the phenomenon of mirror symmetry
160: between Calabi-Yau spaces $X$ and $Y$ should be formulated in terms of
161: equivalence of categories. To a Calabi-Yau space $X$ one can associate
162: two triangulated categories, namely the derived category of coherent
163: sheaves $D^b(X)$ and a derived Fukaya-category $D\mathcal{F}(X)$ of
164: lagrangian cycles (graded, with local systems on them) in $X$
165: (see~\cite{FOOO}). The first category depends only on the holomorphic
166: moduli, the second only on the symplectic (or Kähler) moduli.
167: Mirror symmetry between Calabi-Yau spaces $X$ and $Y$ is then
168: expressed as equivalences of categories.
169: \[
170: \mathrm{Mir}: D^b(X) \stackrel{\approx}{\lra}
171: D\mathcal{F}(Y),\quad D^b(Y) \stackrel{\approx}{\lra}
172: D\mathcal{F}(X)
173: \]
174: These equivalences induce isomorphisms between the corresponding
175: $K$-groups. Via the Chern character they descend to cohomology:
176: \[
177: \mathrm{mir}: \Hgrp^{\text{ev}}(X,\Q) \stackrel{\approx}{\lra}
178: \Hgrp^d(Y,\Q),\quad \Hgrp^{\text{ev}}(Y,\Q)
179: \stackrel{\approx}{\lra} \Hgrp^d(X,\Q),
180: \]
181: where $d = \dim_C X = \dim_C Y$. This also induces an isomorphism
182: between the Kähler moduli $\Hgrp^{1,1}(X)$ of $X$ and the complex
183: moduli of $\Hgrp^{d-1,1}(Y)$ of $Y$. In the Strominger-Yau-Zaslow
184: picture of mirror symmetry (see \cite{SYZ}, \cite{Gr}) $X$ and $Y$ are
185: represented as (real) singular torus fibration over a common base $B$.
186: The fibres are dual tori and mirror symmetry should correspond to
187: fibrewise T-duality. From this one can get some intuitive
188: understanding of the mirror transformation on objects. In particular,
189: the structure sheaf $\mathcal{O}_p$ of a point $p \in X$ gets mapped
190: to a SYZ-fibre $\mathbf{T}$ (with a local system on it) in $Y$ and the
191: structure sheaf $\mathcal{O}_X$ should map to the image $\mathbf{S}$
192: of a section $\sigma: B \lra Y$ of the fibration.
193:
194: For any pair $(\mathcal{E},\mathcal{F})$ of objects of $D^b(X)$ the Euler
195: bilinear form is defined by
196: \[
197: \langle\mathcal{E}, \mathcal{F}\rangle := \chi(\mathcal{E},\mathcal{F}) =
198: \sum_i (-1)^i \dim \Hom(\mathcal{E}, \mathcal{F}[i]).
199: \]
200: which by Serre duality and triviality of the canonical bundle is
201: $(-1)^d$ symmetric. It descends via the Chern-character to a bilinear
202: form $\langle \cdot, \cdot \rangle$ on the cohomology
203: $\Hgrp^{\text{ev}}(X,\Q)$ of $X$, which by Riemann-Roch is given by
204: \[
205: \langle \alpha,\beta \rangle = \int_X \tilde{\alpha} \cup \beta \cup \td(X),
206: \]
207: where $\tilde{\alpha} = (-1)^k \alpha$ for $\alpha \in
208: \Hgrp^{2k}(X,\Q)$.
209:
210: Under the mirror transformation the form $\langle \cdot, \cdot \rangle$ should
211: correspond to the intersection form $\langle \cdot, \cdot \rangle$ of the
212: corresponding lagrangians. One instance of this can easily be checked
213: \[
214: \langle\mathcal{O}_p, \mathcal{O}_X\rangle = 1
215: = \langle \mathbf{T}, \mathbf{S} \rangle.
216: \]
217:
218: \section{Monodromy in one-parameter models}
219: From now on we assume that $X$ and $Y$ are strict Calabi-Yau
220: three-folds and furthermore that they satisfy $h^{2,1}(Y) = 1 =
221: h^{1,1}(X)$. This is the case of so called one-parameter models: $Y$
222: varies in a one-dimensional moduli space and $X$ has one Kähler
223: modulus, i.e., $\Pic(X)=\Z$. In such a case one has $\dim \Hgrp^3(Y)
224: = 4 = \dim \Hgrp^{\text{ev}}(X)$.
225:
226: To be specific, we assume that we have a proper map $\rho: \mathcal{Y}
227: \lra \P^1$, smooth outside singular fibres that sit over points from
228: $\Sigma \subset \P^1$ and furthermore that $Y$ is the fibre over a
229: base-point $s \in \P^1 \setminus \Sigma=:S$. As the geometrical
230: monodromy along a path $\gamma \in \pi_1(S,s)$ can be realised as a
231: symplectic map $M(\gamma): Y_s \lra Y_s$, $M(\gamma)$ induces an
232: autoequivalence of its symplectic invariant $D\mathcal{F}(Y)$, thus
233: setting up a homomorphism
234: \[
235: \pi_1(S,s) \lra \Auteq(D\mathcal{F}(Y))
236: \]
237: which is a refined version of the ordinary monodromy representation of
238: $\pi_1(S,s)$ on $\Hgrp^{\text{odd}}(Y)$. The group $\pi_1(S,s)$ is
239: generated by paths that encircle one of the singular fibres of the
240: family. The induced transformation is determined by the specific
241: properties of the singular fibre. If the fibre aquires the simplest
242: type of singularity, namely an $A_1$-singularity (`conifold'), there
243: is a vanishing lagrangian $3$-sphere. The geometrical monodromy is
244: then a Dehn-twist along this sphere and its effect on homology is given
245: by the classical Picard-Lefschetz transformation \cite{Le},
246: \cite{AGV}, \cite{Lo}:
247: \[
248: \alpha \mapsto S_{\delta}(\alpha) := \alpha - \langle \delta, \alpha
249: \rangle \delta
250: \]
251: where $\delta$ is the homology class of the vanishing cycle.
252: In the situation of mirror symmetry there also will be a point of
253: degeneration with maximal unipotent monodromy. The fibre will typically
254: have normal crossing singularities and there will be a `vanishing
255: $n$-torus', invariant under the monodromy.
256:
257: Using the mirror equivalence $\mathrm{Mir}$ we get a representation
258: \[
259: \pi_1(S,s) \lra \Auteq(D^b(X))
260: \]
261: and one may ask what sort of autoequivalences correspond to
262: specific types of degenerations of $Y$.
263:
264: In \cite{ST00} Seidel and Thomas described a type of autoequivalence
265: in $D^b(X)$ to mirror a symplectic Dehn-twist. It is the
266: Seidel-Thomas twist $T_{\mathcal{E}}$ by a so called spherical object
267: $\mathcal{E}$ of $D^b(X)$, which has the property that
268: $\dim(\textup{Ext}^*(\mathcal{E},\mathcal{E})) = \dim
269: \Hgrp^*(\mathbf{S})$ and is given by the triangle
270: \[
271: \lra (\mathcal{E}, \mathcal{F})\otimes \mathcal{E} \lra \mathcal{F} \lra
272: T_{\mathcal{E}}(\mathcal{F})\stackrel{+1}{\lra}
273: \]
274: The structure sheaf $\mathcal{O}_X$ is the basic spherical object in
275: $D^b(X)$, but also each line bundle $L \in \Pic(X)$ is spherical.
276: Another particularly simple type of autoequivalence is the operation
277: $\otimes L$ of tensoring with a line bundle $L$. Note that
278: $\mathcal{O}_p \otimes L = \mathcal{O}_p$. This fits on the mirror
279: side to the monodromy tranformation around a point of maximal
280: unipotent monodromy, with invariant vanishing torus $\mathbf{T}$.
281:
282: Let us write out these transformations on the level of cohomology.
283: Let $L = \mathcal{O}(H)$ be the ample generator of $\Pic(X)$.
284: The powers $1,H,H^2,H^3$ form a basis for $\Hgrp^{\text{ev}}(X,\Q)$.
285: With respect to this basis, the matrix $T$ of tensoring with
286: $L$ is given by
287: \begin{equation}
288: \label{eq:DM0}
289: T = \begin{pmatrix}
290: 1 & 0 & 0 & 0 \\
291: 1 & 1 & 0 & 0 \\
292: \frac{1}{2} & 1 & 1 & 0 \\
293: \frac{1}{6} & \frac{1}{2} & 1 & 1
294: \end{pmatrix},\quad
295: \end{equation}
296: as easily follows from $\ch(L \otimes \mathcal{E}) = \ch(L) \cup
297: \ch(\mathcal{E}) = e^H \cup \ch(\mathcal{E})$.
298:
299: The twist $T_{\mathcal{O}_X}$ on the level of cohomology is given by
300: $\gamma \mapsto \gamma -\int_X\gamma \cup td(X) \cdot 1$ and hence its
301: matrix is given by
302: \begin{equation}
303: \label{eq:DMcon}
304: S = \begin{pmatrix}
305: 1 & -c & 0 & -d \\
306: 0 & 1 & 0 & 0 \\
307: 0 & 0 & 1 & 0 \\
308: 0 & 0 & 0 & 1
309: \end{pmatrix}
310: \end{equation}
311: where
312: \[
313: d:=H^3,\qquad c:=c_2\cdot H/12.
314: \]
315: The matrix $Q$ representing the bilinear form $\langle \cdot, \cdot
316: \rangle$ in this basis is given by
317: \[
318: Q = \begin{pmatrix}
319: 0 & c & 0 & d \\
320: -c & 0 & -d & 0 \\
321: 0 & d & 0 & 0 \\
322: -d & 0 & 0 & 0
323: \end{pmatrix}
324: \]
325: Now Kontsevich \cite{Ko2} observed the miracle that for the quintic
326: and its mirror the matrices $T$ and $S$ indeed correspond to
327: monodromy matrices of the Picard-Fuchs operator
328: \[
329: \theta^4 - 5^5 z \bigl (\theta+\tfrac{1}{5}\bigr )
330: \bigl (\theta+\tfrac{2}{5}\bigr ) \bigl (\theta+\tfrac{3}{5}\bigr )
331: \bigl (\theta+\tfrac{4}{5} \bigr ).
332: \]
333: It has $0$, $1/5^5$ and $\infty$ as singular points. In an appropriate base,
334: the monodromy around $0$ is given by $T$ and around $1/5^5$ by $S$.
335:
336: We see that apparently the following happens: there is a point of
337: maximal unipotent monodromy, corresponding to $\otimes
338: \mathcal{O}(H)$ in $\Auteq(D^b(X))$ and there is a conifold point,
339: corresponding to the twist along $\mathcal{O}_X$.
340:
341: Similar things occur in all the 14 hypergeometric cases. As there are
342: only three singular points in these cases, these two monodromies generate
343: the monodromy group. We refer to \cite{H} for a generalisation to
344: Calabi-Yaus in more general toric manifolds.
345:
346: Calabi-Yau spaces with Picard number one seem to be rather scarse. Apart
347: from the 14 hypergeometric cases there is there is a list (not
348: claiming completeness in any sense) by Borcea \cite{Bo} containing $11$
349: further cases. The examples are ramified covers and complete
350: intersections in Fano-varieties with Picard-number one. We know of a
351: few other cases. Basic invariants for such $X$ are the \emph{degree}
352: $d:=H^3$, the \emph{second Chern class} $c_2 \cdot H$ and the Euler
353: number $c_3=\chi_{top}$, of which the first two can be read off from
354: the matrix $S$.
355:
356: It is sometimes more convenient to work with a different
357: representation based on the one used by C.~Doran and J.~Morgan (see
358: \cite{DM}). That basis can be obtained from the one above using the
359: coordinate transformation given by the matrix
360: \[
361: W = \begin{pmatrix}
362: 0 & 0 & 0 & 1 \\
363: 0 & 0 & 1 & -1 \\
364: 0 & \frac{1}{d} & -\frac{1}{2} &
365: \frac{1}{3}-\frac{ c }{ d} \\
366: \frac{1}{d} & 0 & -\frac{c }{ d} &
367: \frac{c}{ d}
368: \end{pmatrix}
369: \]
370: where $c$ and $d$ are as above. This yields the following
371: representation:
372: \begin{gather*}
373: T_{\text{DM}} = W^{-1} T W
374: = \begin{pmatrix}
375: 1 & 1 & 0 & 0 \\
376: 0 & 1 & d & 0 \\
377: 0 & 0 & 1 & 1 \\
378: 0 & 0 & 0 & 1
379: \end{pmatrix},\quad
380: S_{\text{DM}} = W^{-1} S W
381: = \begin{pmatrix}
382: 1 & 0 & 0 & 0 \\
383: -k & 1 & 0 & 0 \\
384: -1 & 0 & 1 & 0 \\
385: -1 & 0 & 0 & 1
386: \end{pmatrix} \\
387: Q_{\text{DM}} = W^t Q W
388: = \begin{pmatrix}
389: 0 & 0 & 0 & 1 \\
390: 0 & 0 & -1 & 1 \\
391: 0 & 1 & 0 & -k \\
392: -1 & -1 & k & 0
393: \end{pmatrix}
394: \end{gather*}
395: Here $d=H^3$ and $k=\frac{c_2 \cdot H}{12}+\frac{H^3}{6}$. This last
396: number has a simple interpretation as the dimension
397: $\textup{dim}(H^0(X,{\cal O}(H))$ of the linear system $|H|$.
398:
399: \section{Computation of the monodromy}
400: Our starting point for the computation of the monodromy is the
401: following working hypothesis
402: \begin{hyp}
403: Any differential equation of Calabi-Yau type which is strictly
404: geometrical and for which the instanton numbers have an
405: interpretation as the numbers of curves on a mirror manifold, the
406: monodromy should satisfy the following conditions:
407: \begin{itemize}
408: \item[(H1)] There is a point of maximal unipotent monodromy,
409: correspronding to $\otimes \mathcal{O}(H)$ in $\Auteq(D^b(X))$.
410: \item[(H2)] There is a conifold point, corresponding to the twist
411: along $\mathcal{O}_X$.
412: \end{itemize}
413: \end{hyp}
414: By construction all the equations in the list from \cite{AESZ} have a
415: point of maximal unipotent monodromy at $z=0$. The non-obvious part is
416: to find a conifold point. We observed that in the cases where we know
417: the conifold point the \emph{spectrum}, i.e., the set of zeros of the
418: indicial equation at that point, was $\{0,1,1,2\}$. This is also
419: suggested by Hodge theory. Therefore as a practical selection
420: criterium, we computed the indicial equations at the singular points
421: of all equations and found the equations with at least one singular
422: point with spectrum $\{0,1,1,2\}$. As of the time of writing of this
423: article there were 178 such equations in our database. In many cases
424: there are several such points, but there are also some notable
425: exceptions, where no such singular point exists. An example is
426: equation \eqnno{32}, which is related to $\zeta(4)$ (see \cite{AZ}).
427: For the moment, we are unable to find integral or even just rational
428: lattices for these cases.
429:
430: % read("analyzespectra.maple");
431: % nops(set0112);
432: % (using db.m.20041221-article)
433: For all the 178 equations that do have at least one singular point
434: with spectrum $\{0,1,1,2\}$ we computed high precision numerical
435: approximations for a set of generators of the monodromy group. These
436: computations were done in \texttt{Maple}. The first step was to
437: determine the critical points $z_1,\dots,z_\ell$ and to choose a
438: reference point $p$. Next for each of the critical points $z_i$
439: except the point $z=\infty$ we choose a piecewise linear loop starting
440: and ending at the reference point $p$ and enclosing only one
441: critical point, namely $z_i$ (see \sref{fig:loops}).
442:
443: \begin{figure}
444: \begin{center}
445: \input{loops.pstex_t}
446: \end{center}
447: \caption{Piecewise linear loops around the critical points $z_i$}
448: \label{fig:loops}
449: \end{figure}
450:
451: Using the \texttt{Maple}-function \texttt{dsolve} we can numerically
452: integrate the differential equation along these paths. It turns out to
453: be a bit tricky to obtain the precision needed for the next steps. We
454: used the following options: method=gear, relerr=$10^{-15}$,
455: abserr=$10^{-15}$ and also increased \texttt{Digits} to 100. This
456: yielded the monodromy matrices with respect to an arbitrary basis and
457: produces fully filled $4 \times 4$-matrices with seemingly random
458: complex entries.
459:
460: At this point there is a simple consistency check that we can do. If
461: there exists an integral lattice, the characteristic polynomial of
462: each of the monodromies should be a polynomial with integral
463: coefficients. As a further check, the roots of the indicial equations
464: at the corresponding singular points, should be logarithms of the
465: roots of the characteristic polynomial. For the MUM-point and the
466: points with spectrum $\{0,1,1,2\}$ the characteristic polynomial
467: should be $(1-\lambda)^4$. This provides an indication of the
468: precision we have achieved.
469:
470: The next step is to try and find a simultaneous base change that makes
471: all matrices integral, i.e., to find a monodromy invariant lattice
472: $\Lambda$. The crucial observation is the following. The monodromy $S$
473: around an $A_1$-singularity has the property that $\rk(S-\Id)=1$. The
474: one-dimensional image of $S-\Id$ is the span of the vanishing cycle.
475: Now choose one of the singular points with spectrum $\{0,1,1,2\}$ and
476: call the monodromy around the loop enclosing this singular point $S$.
477: As we are working with numerical approximations we cannot expect
478: $S-\Id$ to have rank $1$, but we can hope that the columns of the
479: matrix $S-\Id$ are nearly proportional. In that case we can pick an
480: arbitrary vector and apply $S-\Id$ to it. In this way we find a vector
481: $v_0$ that should be a good approximation to a lattice vector.
482:
483: Further lattice vectors $v_1,\dots,v_k$ can be obtained by applying
484: words in the numerically computed monodromy matrices to $v_0$. By
485: picking $n$ independent vectors among the ones found in this way, we
486: should find a basis for $\Lambda \otimes \Q$. When we transform the
487: monodromy matrices to this basis, the resulting matrices should have
488: rational entries. Of course this will not be exact, but we can try to
489: find rational matrices close to the matrices that we do find. For
490: this we used continued fractions. It may happen that we get very
491: large denominators or that the rational approximation is not very
492: accurate. In that case we can try another set of $n$ independent
493: vectors among the $v_i$. If that is not successful, we can try another
494: point with spectrum $\{0,1,1,2\}$, if there is any. As a consistency
495: check, we can compute the characteristic polynomials of these rational
496: matrices and check that they have integral coefficients. As noticed
497: above, at the MUM-point and the conifold point the characteristic
498: polynomial should be $(1-\lambda)^4$, which we can also check. If any
499: of these checks fails, we have to try again with a different basis or
500: a different singular point with spectrum $\{0,1,1,2\}$. However, it
501: can and does happen that we try all potential conifold points and
502: several choices of a basis in each case, but do not find a rational
503: basis. We did find a rational basis in 143 of the 178 investigated
504: cases.
505:
506: The rational basis found in this way is still rather arbitrary.
507: However, a major advantage is that at this point we expect to be
508: working with the \emph{exact} monodromy matrices. This allows us to do
509: linear algebra without worrying about the extra complications of
510: working with non exact numerical approximations. Provided that the
511: monodromy matrices around the MUM-point and the conifold point have
512: the right Jordan structure, we can find a new basis such that with
513: respect to this basis they have the standard form \pref{eq:DM0} and
514: \pref{eq:DMcon}. In a geometrical situation we expect the transformed
515: matrices to be integral. This happens in 64 cases. When we have the
516: monodromies around the MUM-point and the conifold point in the
517: standard form, we can read off the invariants $H^3$ and $c_2\cdot H$
518: and try to match the invariants with those of known Calabi-Yau spaces.
519:
520: Despite our efforts to identify equivalent Calabi-Yau equations our
521: list probably still includes some Calabi-Yau equations that correspond
522: to the same geometrical situation. Transformations in the parameter
523: $z$ are a way of constructing seemingly different equations that
524: actually describe the same geometrical situation. In a geometrical
525: language this corresponds to pullback under a map $f: \P^1 \rightarrow
526: \P^1$. If the map $f$ is not injective, this may increase the number
527: of singular points. As long as the map $f$ is unramified around the
528: MUM-point and the conifold point it does not change the monodromies
529: and we ought to find the same $H^3$ and $c_2 \cdot H$. So as practical
530: way of trying to group together the equations that correspond to the
531: same geometry, we sort the 64 integral equations we found according to
532: $H^3$ and $c_2 \cdot H$. If we find several equations with the same
533: $H^3$ and $c_2 \cdot H$, it turns out that the genus zero instanton
534: numbers also coincide. This is a strong indication that these equations
535: are equivalent.
536:
537: \section{Conifold-period and Euler characteristic}
538: If $\Omega$ is a family of holomorphic three forms on $Y_s$
539: (that is, a section of ${\cal L}:=\rho_*(\omega_{\cal Y}/S)$)
540: and $\Gamma$ is a horizontal family of cycles, then the
541: \emph{periods}
542: \[
543: \int_{\Gamma} \Omega
544: \]
545: are the solutions of the associated Picard-Fuchs equation. In our
546: situation we identified two cycles, namely the torus $\mathbf{T}$ near
547: the MUM-point, and the vanishing sphere $\mathbf{S}$ near the conifold
548: point $z_c$. Correspondingly we have the fundamental period
549: $\int_{\mathbf{T}} \Omega$ which is the unique holomorphic solution
550: near the MUM-point. Equally important is the period $\int_{\mathbf{S}}
551: \Omega$ which we call the \emph{conifold-period} and which was called
552: $z_2(t)$ in the paper \cite{CdOGP}. As the local monodromy around the
553: conifold point is supposed to be a symplectic reflection in $S$, this
554: special period can be determined directly from the differential
555: equation as follows. At such a conifold point there exists a basis of
556: solutions to the Calabi-Yau-equation around this point that consists
557: of three power series solutions and one solution of the form
558: \[
559: y(z) = f(z) \log(z-z_c) + g(z).
560: \]
561: Going around $z_c$ once $y$ is replaced by $y(z) + 2\pi i f(z)$. The
562: power series $f(z)$ represents a special solution around $z=z_c$ that
563: is determined up to a multiplicative scalar and which we call the
564: \emph{conifold-period}. This function can be continued analytically
565: around an arbitrary path which avoids the singularities of the
566: differential equation. It is a remarkable fact that in all (but one,
567: namely nr.~\eqnno{224}) of the examples we know, the point $z_c$ is the
568: singular point that is \emph{closest to the origin}. So there is a
569: preferred path from $z_c$ to $0$ by going along a straight line and we
570: consider the analytic continuation along this path. In \cite{CdOGP}
571: the expansion of the conifold period around $0$ is derived for the
572: quintic. It has the form
573: \begin{equation}
574: \label{eq:z2t}
575: z_2(t) = \frac{H^3}{6} t^3 + \frac{c_2 \cdot H}{24} t +
576: \frac{c_3}{(2\pi i)^3} \zeta(3) + O(q).
577: \end{equation}
578: The term $O(q)$ stands for any terms containing $q=e^{2\pi i t}$ and
579: $t = \frac{1}{2\pi i} \frac{y_1(z)}{y_0(z)} = \frac{1}{2\pi i} \log z
580: + \frac{1}{2\pi i} \frac{f_1(z)}{f_0(z)}$ instead of \pref{eq:tcoord}.
581: Remarkable here is the `constant term' $\frac{c_3}{(2\pi i)^3}
582: \zeta(3)$. This term is related to the four-loop correction to
583: the free energy $F_0$ introduced in \cite{CdOGP}.\footnote{To be
584: precise, one has an expansion (see \cite{KKV}):
585: \[
586: F_0=H^3 \frac{t^3}{3!} + (c_2 \cdot H)t + \frac{\chi}{2}\zeta(3) +
587: \sum_{d=1}^{\infty} n_d^0 Li_3(q^d)
588: \]
589: where $\mathrm{Li}_3(x):=\sum_{k=1}^{\infty} k^{-3} x^k$ is the classical
590: trilogarithm.}
591:
592: One can conjecture this expansion to hold in all cases, which leads to
593: the following algorithm to determine $c_3$. One can easily compute an
594: expansion of $y(z)$ to an arbitrary number of terms, e.g., using the
595: \texttt{Maple}-function \texttt{formal\_sol} from the
596: \texttt{DEtools}-package, as we did. This allows us to find $f(z)$ as
597: the coefficient of $\log(z-z_c)$. Around the MUM-point $z=0$ we can
598: compute expansions of the elements $y_i(z)$ of the Frobenius basis
599: (see \pref{app:inst}). We suppose that the domains of convergence of
600: the solutions around $z=0$ and those around $z=z_c$ overlap. That
601: enables us to pick some point $z_*$ where both expansions converge.
602: Computing numerically $f^{(k)}(z_*)$ ($k=0,\dots,3$) and
603: $y^{(k)}_i(z_*)$ ($k,i=0,\dots,3$), we can consider the equations
604: \[
605: f^{(k)}(z_*) = \sum_{i=0}^3 c_i y_i^{(k)}(z_*).
606: \]
607: These equations can be solved for the $c_i$ and determine the analytic
608: continuation $z_2$ around $0$ of $f(z)$ as a linear combination of the
609: $y_i(z)$
610: \[
611: z_2 = \sum_{i=0}^3 c_i y_i(z).
612: \]
613: From this we can readily read of the expansion of $z_2$ in $t$. At
614: this point we can already check that the coefficient of $t^2$
615: vanishes. As the conifold-period $f(z)$ was only determined up to a
616: constant, of course $z_2(t)$ ist determined up to a constant. If we
617: suppose that $H^3$ is known, then one can multiply the expansion for
618: $z_2(t)$ by a constant such that the coefficient of $t^3$ is
619: $\frac{H^3}{6}$. We can then read off $c_2 \cdot H$ and $c_3$. In
620: praxis we find $H^3$ as discussed above from the monodromy generators.
621: This also yields $c_2 \cdot H$, so we have one more consistency check.
622: It is remarkable that in all cases we indeed find an integral value of
623: $c_3$!
624:
625: \begin{table}
626: \vspace{-3cm}
627:
628: \caption{\textbf{Calabi-Yau equations with integral monodromy}}
629: \begin{center}
630: \makebox[\textwidth]{%
631: \input{table.tex}}
632: \end{center}
633: \label{tab:moni}
634: \end{table}
635:
636: \section{Comments on the table of Calabi-Yau-equations}
637: In \sref{tab:moni} the heading Sings denotes the number of singular
638: point of the (first mentioned) differential equation. An additional
639: $*$ indicates, that an apparent singularity is present, around which
640: there is no monodromy. The notation $X(\dots)$ denotes a complete
641: intersection of the indicated degrees in the indicated manifold.
642: Apart from the familiar $13$ hypergeometric cases and the cases from
643: complete intersection in Grassmanians that were studied in in
644: \cite{BCKvS}, one finds a few notable further cases. First there is
645: the elusive $14$th hypergeometric case, observed in \cite{Al} and
646: \cite{DM}. Any complete intersection $X(2,12)$ inside
647: $\P(1,1,1,1,4,6)$ has a singular point of type $A_1/(\Z/2)$, which
648: does not admit a crepant resolution. The case $X \stackrel{2:1}{\lra}
649: B_5$ is the Calabi-Yau double cover of the Fano-threefold $B_5$, which
650: is nothing but the three-dimensional section of $\Grass(2,5)$, which
651: is no.~14 in the list of Borcea. We found a fit with the equation
652: \eqnno{51} from \cite{AESZ}. A mirror for this Calabi-Yau is not
653: known, but we conjecture the Picard-Fuchs equation to be the indicated
654: one. We find similar fits for
655: \begin{description}
656: \item[\mathversion{bold}$X(1,1,1,1,1,1,2) \subset X_{10}$:] Here
657: $X_{10} \subset \P^{15}$ is the celebrated 10-dimensional spinor
658: variety of isotropic $4$-planes in the 8-dimensional quadric.
659: \item[\mathversion{bold}$X(1,1,2) \subset \LGrass(3,6)$:]
660: $\LGrass(3,6)=\mathrm{Sp}(3,\C)/P(\alpha_3) \subset \Grass(3,6)$ is
661: the Lagrangian Grassmanian.
662: \item[\mathversion{bold}$X(1,2) \subset X_5$:] Here
663: $X_5=G_2/P(\alpha_{\text{long}}) \subset \Grass(5,7)$ is the space of
664: 5-di\-men\-sio\-nal subspaces isotropic for a 4-form on a 7-dimensional
665: space.
666: \end{description}
667: These are complete intersections inside homogeneous spaces. In
668: principle one can calculate the Picard-Fuchs equation for the
669: instanton numbers for these cases and verify our conjecture. The first
670: method consist in computing the quantum cohomology of these
671: homogeneous examples (for example by fixed point localisation) and then
672: use the quantum Lefschetz hyperplane principle. A second method
673: consists of finding a toric degeneration and then using polar duality.
674: Such toric degenerations have been constructed for all spherical
675: varieties in \cite{AB}. Both methods were used in \cite{BCKvS} for the
676: case of complete intersections in Grassmannians.
677:
678: In his thesis \cite{To}, F. Tonoli considers Calabi-Yau varieties in
679: $\P^6$ of degree $12$ up to $17$. The first one is the complete
680: intersection $X(2,2,3)$, the second one the $5 \times 5$-Pfaffian, for
681: which we found a fit with the data from equation \eqnno{99}. The $7
682: \times 7$-Pfaffian was considered in \cite{Ro98}. The remaining three
683: case are new Calabi-Yau threefolds for which we have not yet found
684: corresponding Picard-Fuchs equations.
685:
686: The column for the Euler characteristic $c_3$ was determined using the
687: expansion of the conifold-period around the MUM-point. It is a miracle
688: that we found integral values in all cases (except \eqnno{224}). This
689: checked with the known Euler number in those cases where a geometrical
690: interpretation was known. However, there are two notable cases where
691: we get a \emph{positive} value for $c_3$, which excludes an
692: interpretation as a Calabi-Yau space with Picard number one.
693: Furthermore, the conjectural integrality of elliptic intanton numbers
694: implies some congruence property on $c_3$. In most cases this was
695: satisfied, giving a strong indication that a Calabi-Yau threefold with
696: the indicated invariants should exist. In some cases however, we
697: found non-integral in this way $n_d^1$. This is indicated with a $*$
698: after the value for $c_3$.
699:
700: In the database column we indicate the number of the equation in the
701: electronic database of Calabi-Yau equations that can be found at the
702: web address
703: \[
704: \text{\hlink{http://enriques.mathematik.uni-mainz.de/enckevort/db}}
705: \]
706: Up to 180 these numbers coincide with the ones used in \cite{AESZ}.
707: For higher numbers one should check the source field in the electronic
708: database. If it contains Almkvist[$n$] the corresponding number in
709: \cite{AESZ} is $n$.
710:
711: \input{examples.tex}
712:
713: \section{Open problems}
714: The work described in this paper is no more than a start and there are
715: many open problems left. We have found quite a few Calabi-Yau
716: equations that look in every respect like the Picard-Fuchs equation of
717: a Calabi-Yau manifold, but for which we do not know if a Calabi-Yau
718: manifold exists. We know the degree, the second Chern class, the Euler
719: characteristic and the instanton numbers.
720:
721: To determine an integral lattice we need to single out two singular
722: points, where we bring the monodromies into the the standard forms
723: $T_{\text{DM}}$ and $S_{\text{DM}}$. When there are no singular points
724: with spectrum $\{0,1,1,2\}$ we do not have a good candidate for
725: $S_{\text{DM}}$ and cannot even start our procedure for determining an
726: integral lattice. It would be interesting to see what can be done in
727: such cases. We also did not look for other integral lattices as in
728: \cite{DM}.
729:
730: The conjectural appearance of the constant term $c_3\zeta(3)/(2\pi
731: i)^3$ in the expansion of the conifold period (and the free energy) is
732: very intriguing. Is this a mathematical theorem?
733:
734: The key obstacle to computing the elliptic instanton numbers is
735: finding the holomorphic function $f(z)$ in \pref{eq:F1}. Our ansatz in
736: combination with our recipe for determining the exponents works a many
737: cases, but it is no more than an educated guess. A better understanding of
738: the genus one computation in terms of the BCOV-torsion as in \cite{FL}
739: will probably be helpful.
740:
741: Many of the equations from the list in \cite{AESZ} come from Hadamard
742: products. The singular points of a Hadamard product are given by
743: products of the singular points of the factors. Maybe it is also
744: possible to determine the monodromies of the Hadamard product in terms
745: of the monodromies of the factors.
746:
747: %Diagonal families???
748:
749: \appendix
750: \section{Orbifolds of $A_1$}
751: In many examples one encounters monodromy tranformations that
752: are not described by the usual Picard-Lefschetz formula, but
753: rather are \emph{powers} of such operations. We offer
754: a possible explanation of this phenomonen, which is only visible
755: on the integral level.
756:
757: Consider a lattice $\Lambda$ with bilinear form $\langle \cdot, \cdot
758: \rangle$. For $\beta \in \Lambda$ and $\lambda \in \Z$ consider the
759: the transformation
760: \begin{equation}
761: \label{eq:Slb}
762: S_{\lambda,\beta}: \Lambda \lra \Lambda,\quad
763: S_{\lambda,\beta}(\alpha) = \alpha - \lambda \langle \beta,
764: \alpha \rangle \beta.
765: \end{equation}
766: The transformation $S_{\lambda,\beta}$ preserves $\langle \cdot, \cdot
767: \rangle$ in the symmetric case only when $\lambda=2/Q(\beta,\beta)$
768: (or $\lambda=0$). In that case $S_{\lambda,\beta}$ has order two and is
769: a reflection. When $\langle \cdot, \cdot \rangle$ is antisymmetric,
770: there is no restriction on $\lambda$ and $S_{\lambda,\beta} \circ
771: S_{\lambda',\beta} = S_{\lambda+\lambda',\beta}$. So in that case
772: $S_{\lambda,\beta}$ does not have finite order.
773:
774: Such transformations occur as monodromy transformations
775: where not a sphere, but rather a quotient $S^3/G$ by a finite
776: group $G$ is vanishing, as we will explain now.
777: Consider the function defining the three-dimensional $A_1$-singularity:
778: \[
779: f: \C^4 \lra \C,\quad f(x,y,z,t)=x^2+y^2+z^2+t^2
780: \]
781: The fibre $F_s$ of $f$ over $s \in \C \setminus {0}$ is called the
782: Milnor fibre and can be identified with the cotangent bundel to the
783: sphere $\{(x,y,z,t) \in \R^4 \mid x^2+y^2+z^2+t^2=s\}$, which is
784: vanishing when $s \rightarrow 0$. We choose an orientation and let
785: $\delta$ be the homology class of this sphere. There is also a
786: covanishing cycle $\epsilon$ in the dual group
787: $\Hgrp^{\text{cl}}_3(F_s,\Z)$ (homology with closed support). One has
788: \[
789: \Hgrp_3(F_s,\Z) = \Z \delta,\quad
790: \Hgrp_3^{\text{cl}}(F_s,\Z) = \Z\epsilon,\quad
791: \langle \delta, \epsilon \rangle = 1
792: \]
793: Let $G \subset \mathrm{SU}(2) = S^3$ be a finite subgroup. $G$ then
794: acts linearly on $\R^4$ and by complexification on $\C^4$, leaving
795: invariant the function $f$ defining the $A_1$-singularity. Consider
796: the quotient map $\pi: \C^4 \lra X := \C^4/G$. The space $X$ will be
797: singular, but $f$ descends to a function $g: X \lra \C$, such that
798: $f=\pi \circ g$. So the fibre $G_s := g^{-1}(s)$ is the quotient of
799: $F_s$ by $G$. In the fibre $G_s$ there is a cycle $S^3/G$ vanishing
800: when $s \rightarrow 0$, with homology class $d \in \Hgrp_3(G_s,\Z)$.
801: As above there also exists a covanishing cycle $e \in
802: \Hgrp_3^{\text{cl}}(G_s,\Z)$ such that
803: \[
804: \Hgrp_3(G_s,\Z) = \Z d,\quad \Hgrp_3^{\text{cl}}(G_s,\Z) = \Z
805: e,\quad \langle d, e \rangle=1.
806: \]
807: The map $\pi$ induces maps $\pi^*$ and $\pi_*$ between the homology
808: groups of $F_s$ and $G_s$ and one easily sees that
809: \[
810: \pi_*(\delta) = |G|d,\quad \pi_*(\epsilon) = e,\quad
811: \pi^*(d) = \delta,\quad \pi^*(e)=|G|\epsilon
812: \]
813: The Picard-Lefschetz formula tells us that under the monodromy of $f$
814: the cycle $\delta$ remains fixed, whereas the cycle $\epsilon$ gets
815: mapped to $\epsilon -\delta$. From the fact that the monodromy
816: commutes with the group action we obtain, by taking $\pi_*$, that $d$
817: remains fixed, whereas $e$ gets mapped to $e-|G|d$. From this one
818: deduces in the usual way that the occurence of a singularity of type
819: $A_1/G$ will lead the monodromy transformation described by the
820: modified Picard-Lefschetz formula (see \cite{AGV})
821: \[
822: \gamma \mapsto \gamma -|G|\langle d, \gamma \rangle d
823: \]
824: The cycle $d$ should should give rise to a spherical object in the
825: derived category of the mirror, but only the $|G|$th power of the
826: Seidel-Thomas twist would arise from a monodromy transformation.
827:
828: \section{Computation of instanton numbers}
829: \label{app:inst}
830: According to \cite{GV1,GV2} (see also \cite{KKV}) we have the
831: following expansion for the partition function $F$ of the topological
832: string
833: \[
834: F =\sum_{g=0}^\infty \lambda^{2g-2} F_g = \sum_{g=0}^\infty \sum_d
835: \sum_{m=0}^\infty n^g_d \frac{1}{m} \biggl ( 2 \sin
836: \frac{m\lambda}{2} \biggr )^{2g-2} q^{dm}.
837: \]
838: The partion function $F$ can be defined physically or mathematically
839: using Gromov-Witten invariants. The above formula can then be
840: considered to define the Gopakumar-Vafa invariants $n^g_d$. In
841: contrast to e.g., the Gromov-Witten invariants, the Gopakumar-Vafa
842: invariants are conjectured to be always integral.
843:
844: We will restrict to the genus zero and genus one invariants.
845: Furthermore, we will only consider the 1-parameter case. In that case
846: we have the following formulas (see \cite{KKV})
847: \[
848: \partial_t^3 F_0 = n^0_0 + \sum_{\ell=1}^\infty \frac{n^0_\ell \ell^3
849: q^\ell}{1-q^\ell}
850: \]
851: with $n^0_0=H^3$ and
852: \[
853: \partial_t F_1 = \frac{c_2 \cdot H}{24} + \sum_{d=1}^\infty
854: \sum_{k=1}^\infty \bigl ( \tfrac{1}{12} n_d^0 + n_d^1 \bigr ) d
855: q^{kd}.
856: \]
857: Here we define the coordinate $t$ by $q=e^{-t}$. So if we can compute the
858: left hand sides of these equations the invariants $n^0_d$ and $n^1_d$
859: can easily be determined.
860:
861: To do so, we first introduce a special basis of solutions for the
862: equation \pref{eq:L} around a point of maximal unipotent monodromy,
863: i.e., a singular point where $\lambda=0$ is the only solution to the
864: indicial equation. In physical terms such a point (also called
865: MUM-point) corresponds to a large radius limit point.
866:
867: Suppose $z=0$ is a MUM-point. Then we can use the Frobenius method.
868: The idea is to consider a solution with values in the ring
869: $\C[\rho]/(\rho^n)$. We make the following ansatz for such a solution
870: \[
871: \tilde{y}(z)=\sum_{n=0}^\infty A(n,\rho) z^{n+\rho}
872: = y_0(z) + y_1(z) \rho + \dots + y_{n-1}(z) \rho^{n-1},
873: \]
874: where we define
875: \[
876: z^\rho = e^{\log z \cdot \rho} = 1 + \log z \cdot \rho +
877: \frac{\log^2 z}{\rho} \cdot \rho^2 + \dots + \frac{\log^{n-1}
878: z}{(n-1)!} \cdot \rho^{n-1}.
879: \]
880: Using $\theta z^{n+\rho} = (n+\rho) z^{n+\rho}$, where $\theta=z
881: \frac{d}{dz}$, we can translate the equation $L \tilde{y}=0$ into a
882: recursion relation for the $A(n,\rho)$. As initial condition for the
883: recursion we use $A(0,\rho)=1$. The $y_i$ we find in this way are
884: called the \emph{Frobenius basis}.
885:
886: Define power series $f_i$ by the following expression
887: \[
888: \sum_{n=0}^\infty A(n,\rho) z^n = f_0(z) + f_1(z) \rho + \dots +
889: f_{n-1}(z) \rho^{n-1}.
890: \]
891: Because $z^\rho \sum_{i=0}^{n-1} f_i(z) \rho^i = \sum_{i=0}^{n-1}
892: y_i(z) \rho^i$, we find
893: \[
894: y_i(z)=\sum_{j=0}^i \frac{\log^i z}{i!} f_{j-i}(z).
895: \]
896: Using the Frobenius base we can define a new coordinate
897: \begin{equation}
898: \label{eq:tcoord}
899: t=y_1(z)/y_0(z)=\log z + \frac{f_1(z)}{f_0(z)}.
900: \end{equation}
901:
902: There are basically two ways to compute $\partial_t^3 F_0$. The
903: starting point of the first one is the \emph{Yukawa coupling} in the
904: $z$ coordinate
905: \[
906: K_{zzz}=\exp\bigl ({\textstyle -\frac{1}{2} \int a_3(z) \dop z} \bigr ),
907: \]
908: where $a_3(z)$ is one of the coefficients from \pref{eq:L}. The claim
909: is that $\partial_t^3 F$ is the following transformation of this
910: function to the $t$-coordinate defined in \pref{eq:tcoord}
911: \begin{equation}
912: \label{eq:Kttt}
913: \partial_t^3 F(t) = \frac{K_{zzz}(z(t))}{y_0^2(z(t))
914: \bigl ( \frac{dt}{dz} \bigr )^3}
915: \end{equation}
916:
917: Now recall that a Calabi-Yau equation has to satisfy a list of
918: conditions (see \cite{AZ,AESZ}). One of these can be written as
919: \begin{equation}
920: \label{eq:AZ}
921: a_1=\frac{1}{2}a_2a_3-\frac{1}{8}a_3^3 + a_2' - \frac{3}{4}a_3a_3'-
922: \frac{1}{2}a_3''
923: \end{equation}
924: According to Proposition~1 from \cite{AZ} this condition is equivalent
925: to the two conditions
926: \begin{align}
927: \label{eq:AZ1}
928: \frac{d^2}{dt^2} \frac{y_2}{y_0} &= \frac{\exp\bigl (-\frac{1}{2}\int
929: a_3(z) \dop z \bigr )}{y_0^2 \bigl ( \frac{dt}{dz}\bigr )^3}, \\
930: \label{eq:AZ2}
931: \frac{d^2}{dt^2} \frac{y_3}{y_0} &= t \frac{d^2}{dt^2} \frac{y_2}{y_0}.
932: \end{align}
933: The second condition is equivalent to the existence of a function $G$
934: and a constant $c$ such that
935: \[
936: \Pi(t):=\frac{1}{y_0} \begin{pmatrix} y_0 \\ y_1 \\ y_2 \\ y_3
937: \end{pmatrix} =
938: \begin{pmatrix}
939: 1 \\
940: t \\
941: \partial_t G - c\\
942: t\partial_t G - 2 G
943: \end{pmatrix}.
944: \]
945: The vector $\Pi(t)$ is called the \emph{normalized period vector}.
946: Using the second condition \pref{eq:Kttt} translates to
947: \[
948: \partial_t^3 F_0 = \partial_t^3 G(t) = \partial_t^2 \frac{y_2}{y_0}.
949: \]
950: This yields a different way of computing $\partial_t^3 F_0$ and
951: therefore also the instanton numbers $n^0_d$.
952:
953: To compute $\partial_t F_1$ we use the recipe from \cite{BCOV1,BCOV2}
954: based on an analysis of the so called holomorphic anomaly. We will use
955: the following formula from \cite{BCOV1} (using our notation and
956: adapted slightly for the case we are studying):
957: \begin{equation}
958: \label{eq:F1}
959: \partial_t F_1 = \partial_t \log \left (
960: \frac{z^{1+\frac{c_2 \cdot H}{12}}f(z)}{y_0^{4-\frac{c_3}{12}}
961: \frac{\partial t}{\partial z}} \right ).
962: \end{equation}
963: In this formula one needs the geometrical data $c_2 \cdot H$ and $c_3$
964: which can usually be determined from the monodromy calculation and/or
965: conifold period. However, the main problem with this formula is the
966: function $f$ which is a holomorphic function of $z$ that still has to
967: be determined. We will use an ansatz for $f$ to reduce this problem to
968: the determination of a finite number of parameters. To describe this
969: ansatz note that because of the special form of a Calabi-Yau equation
970: we can write
971: \[
972: a_4(z)= z^4 \Delta(z)= z^4 \prod_i (\Delta_i(z))^{k_i},
973: \]
974: for some polynomial $\Delta(z)$, which we call the
975: \emph{discriminant}. The $\Delta_i(z)$ are the irreducible factors
976: (over $\R$) of $\Delta(z)$. Our ansatz is then the following
977: \[
978: f(z) = \prod_i (\Delta_i(z))^{s_i},
979: \]
980: where the exponents $s_i \in \Q$ still have to be determined. The
981: (apparent) singular points of the operator are the zeros of the
982: discriminant $\Delta(z)$ (and $0$ and $\infty$). So each of the
983: factors $\Delta_i(z)$ corresponds via its zeros to one or more
984: (apparent) singular points. To determine the exponents we look at the
985: monodromies around the corresponding singular points. When the
986: singular point is a conifold, i.e., the monodromy is of the form
987: $S_{1,v}$, then the exponent is generally assumed to be
988: $-\frac{1}{6}$. We generalize this to $-\frac{\lambda}{6}$ for
989: monodromies of the form $S_{\lambda,v}$ for arbitrary $\lambda$. When
990: the monodromy is the identity, we put the exponent to zero. These
991: rules already allow us to deal with many equations. However,
992: monodromies of other types for which we do not know of a sensible
993: guess do occur.
994:
995: % To define the genus one instanton numbers we need a generalization of
996: % the multiple cover formula \pref{eq:mc0}. We will use the modern
997: % approach using Gopakumar-Vafa invariants (see \cite{} for a comparison
998: % of the Gopakumar-Vafa invariants and the older definitions of genus
999: % one instanton numbers).
1000: % \[
1001: % \partial_t F_1 = \frac{c_2 \cdot H}{24} + \sum_{d=1}^\infty
1002: % \sum_{k=1}^\infty \bigl ( \tfrac{1}{12} n_d^0 + n_d^1 \bigr ) d
1003: % q^{kd}
1004: % \]
1005: % The first one (formula (13) in \cite{BCOV1}) expresses the derivative
1006: % of the genus 1 topological potential in terms of the genus 0 and genus
1007: % 1 instanton numbers and the Chern class $c_{n-1}$
1008: % \[
1009: % \partial_{t^a} F_1^{\text{top}} = \frac{(-1)^{n-1}}{12} \int k_a
1010: % \wedge c_{n-1} - 2 \sum_s n^1_s \sum_{n=1}^\infty \frac{ns_a q^{ns}}{1-q^{ns}}
1011: % -\frac{1}{6} \sum_s n^0_s s_a \frac{q^s}{1-q^s}.
1012: % \]
1013: % Here $t^a$ are coordinates on the moduli spaces corresponding to a
1014: % basis $\{k_a\}_{a=1}^{h^{1,1}}$ of $H^{1,1}(X)$, $s$ is a multi-index
1015: % and $q^s=\exp\bigl (-\sum_{a=1}^{h^{1,1}} t^a k_a \bigr )$. In case we
1016: % are studying here $n=3$ and $h^{1,1}=1$. So we can simplify the
1017: % formula to
1018: % \[
1019: % \partial_t F_1^{\text{top}} = \frac{1}{12} \int k
1020: % \wedge c_2 - 2 \sum_s n^1_s \sum_{n=1}^\infty \frac{ns q^{ns}}{1-q^{ns}}
1021: % -\frac{1}{6} \sum_s n^0_s s \frac{q^s}{1-q^s}.
1022: % \]
1023: % Kind of multiple cover formula
1024: % Alternative: use GV-invariants (slightly different, but probably better)
1025:
1026: % \section{Hadamard Product}
1027:
1028: % {\bf Theorem:} The set of solutions in $k[[x]]$ of geometric differential
1029: % equations is a $k$-vector subspace stable under Cauchy and Hadamard Product.
1030:
1031: % The \textsc{Hadamard} product of two power series
1032: % \[f(x)=\sum_n a_n x^n\]
1033: % and
1034: % \[g(x)=\sum_n b_n x^n\]
1035: % is the series
1036: % \[f*g(x)=\sum_n a b_n x^n\]
1037: % Another way of writing this is
1038: % \[f*g(x) dx= (1/2\pi i) \int_{\gamma} f(s) g(t) ds dt / (x-st) \]
1039:
1040: % If $f(s)$ and $g(t)$ arise as period integrals of the form
1041: % \[f(s)=\int_\Gamma (\frac{dy}{F(y,s)}\]
1042: % \[g(t)=\int_\Delta (\frac{dz}{G(z,t)}\]
1043: % then
1044: % \[f*g(x)=int_{\Gamma \times \Delta \times \gamma} \frac{dy\wedge dz\wedge ds dt}{F(y,s)G(z,t)(x-st)} \]
1045:
1046: % In terms of local systems one has the following construction.
1047: % Given two local systems $L_1$ and $L_2$ on $\C^*$, one
1048: % can form the external tensor product
1049: % \[p_1^*(L_1) \otimes p_2^*(L_2)\]
1050: % oon $\C^* \times \C^*$. The multiplication map
1051: % \[ \mu: \C^* \times \C^* \lra \C^*, (y,z) \mapsto yz \]
1052: % gives a way to produce local systems
1053: % \[ R\mu_*(p_1^*(L_1) \otimes p_2^*(L_2))\]
1054: % on $\C^*$.
1055:
1056: % Problem: Express the Monodromy of a Hadamard-product in terms
1057: % of the mondormies of the factors.
1058:
1059: \bibliographystyle{plain}
1060: \bibliography{jointproj}
1061: \end{document}
1062:
1063: %%% Local Variables:
1064: %%% mode: latex
1065: %%% TeX-master: t
1066: %%% End: