math0309033/rr.tex
1: \documentclass[11pt]{article}
2: 
3: \setlength{\oddsidemargin}{1cm} \setlength{\evensidemargin}{1cm}\setlength{\textwidth}{14cm} % horizontal
4: \setlength{\topmargin}{0cm}\setlength{\textheight}{23cm} \setlength{\headheight}{.1in}\setlength{\headsep}{.3in}\setlength{\parskip}{.5mm} % vertical
5: 
6: \usepackage{amsmath,amssymb,amscd}
7: \usepackage[dvips]{graphics}
8: 
9: \newcommand{\rmk}{Contents}
10: \newcommand{\rmf}{Figures}
11: \renewcommand{\sectionmark}[1]{\markright{\thesection\ #1}}
12: \renewcommand{\subsectionmark}[1]{\markright{\thesubsection\ #1}}
13: 
14: \def\mapdownright#1{\Big\downarrow\rlap{$\vcenter{\hbox{$\scriptstyle#1$}}$}}
15: \def\mapdownleft#1{\rlap{$\vcenter{\hbox{$\scriptstyle#1$}}$}\,\,\,\Big\downarrow}
16: \def\mapse#1{\rlap{$\vcenter{\hbox{$\scriptstyle#1$}}$}\,\,\searrow}
17: \def\mapright#1{\smash{\mathop{\longrightarrow}\limits^{#1}}}
18: \def\mapleft#1{\smash{\mathop{\longleftarrow}\limits^{#1}}}
19: 
20: \renewcommand{\thetable}{\thetheorem}
21: \newtheorem{theorem}{Theorem}[section]
22: \newtheorem{nothing}[theorem]{$\!\!$}
23: \newtheorem{proposition}[theorem]{Proposition}
24: \newtheorem{conjecture}[theorem]{Conjecture}
25: \newtheorem{lemma}[theorem]{Lemma}
26: \newtheorem{example}[theorem]{Example}
27: \newtheorem{examples}[theorem]{Examples}
28: \newtheorem{construction}[theorem]{Construction}
29: \newtheorem{remark}[theorem]{Remark}
30: \newtheorem{algorithm}[theorem]{Algorithm}
31: \newtheorem{remarks}[theorem]{Remarks}
32: \newtheorem{corollary}[theorem]{Corollary}
33: \newtheorem{question}[theorem]{Question}
34: \newtheorem{tabl}[theorem]{Table}
35: \newtheorem{defin}[theorem]{Definition}
36: 
37: \newcommand{\prf}{\vspace{.05in}
38:                     \noindent {\sc Proof} \hspace{.05in}}
39: \newcommand{\prfc}{\vspace{.05in}
40:                     \noindent {\sc Proof of the Corollary}}
41: 
42: \newcommand{\prfth}{\vspace{.05in}
43:                     \noindent {\sc Proof of Theorem}}
44: 
45: \newcommand{\ethrm}{\hspace*{\fill}
46:                       $\Box$
47:                       \vspace{.1in}}
48: \newcommand{\noprf}{\vspace{-.2in}
49:                      \hspace*{\fill}
50:                       $\Box$
51:                       \vspace{.1in}}
52: 
53: \newcommand{\bp}{{\mathbb P}}
54: \newcommand{\bc}{{\mathbb C}}
55: \newcommand{\bF}{{\mathbb F}}
56: \newcommand{\bcstar}{\bc^*}
57: \newcommand{\bz}{{\mathbb Z}}
58: \newcommand{\bq}{{\mathbb Q}}
59: \newcommand{\br}{{\mathbb R}}
60: \newcommand{\ba}{{\mathbb A}}
61: \newcommand{\bbf}{{\mathcal F}}
62: \renewcommand{\L}{{\mathcal L}}
63: \renewcommand{\P}{{\mathcal P}}
64: \newcommand{\Q}{{\mathcal Q}}
65: \newcommand{\D}{{\mathcal D}}
66: \newcommand{\bk}{{\mathcal K}}
67: \newcommand{\bl}{{\mathcal L}} 
68: \newcommand{\bt}{{\mathcal T}}
69: \newcommand{\bn}{{\mathcal N}}
70: \newcommand{\bv}{{\mathcal V}}
71: \newcommand{\bigo}{{\mathcal O}}
72: \newcommand{\ox}{{\bigo}_X}
73: \newcommand{\ul}{\underline}
74: \newcommand{\ol}{\overline}
75: \newcommand{\ep}{\varepsilon} 
76: 
77: \newcommand{\aut}{{\mathop{\rm Aut}\nolimits}\,}
78: \newcommand{\cy}{Calabi--Yau }
79: 
80: \newcommand{\kod}[1]{H^1(#1, T_{#1})}
81: \newcommand{\restr}[1]{\!\!\mid_{\,#1}}
82: \newcommand{\levell}[1]{#1^{(l)}}
83: \newcommand{\rhostr}[1]{{\bF #1}}
84: \newcommand{\lin}[1]{\,\mid #1 \mid}
85: 
86: \newcommand{\cstar}{\bc\,^*}
87: 
88: \renewcommand{\mod}{\mathop{\rm mod}\nolimits}
89: \newcommand{\lcm}{\mathop{\rm lcm}\nolimits}
90: \newcommand{\hcf}{\mathop{\rm hcf}\nolimits}
91: \newcommand{\spec}{\mathop{\rm Spec\,}\nolimits}
92: \newcommand{\proj}{\mathop{\rm Proj\,}\nolimits}
93: \newcommand{\pic}{\mathop{\rm Pic\,}\nolimits}
94: \newcommand{\cl}{\mathop{\rm Cl\,}\nolimits}
95: \newcommand{\weil}{\mathop{\rm Weil\,}\nolimits}
96: \newcommand{\rk}{\mathop{\rm rk\,}\nolimits}
97: \newcommand{\bir}{\mathop{\rm Birat}\nolimits}
98: \newcommand{\df}{\mathop{\rm Def\,}\nolimits}
99: \newcommand{\codim}{\mathop{\rm codim}\nolimits}
100: \newcommand{\im}{\mathop{\rm Im}\nolimits}
101: \newcommand{\cha}{\mathop{\rm char}\nolimits}
102: \newcommand{\Hom}{\mathop{\rm Hom}\nolimits}
103: \newcommand{\tors}{\mathop{\rm Tors}\nolimits}
104: \newcommand{\rank}{\mathop{\rm rank}\nolimits}
105: \newcommand{\sing}{\mathop{\rm Sing}\nolimits}
106: \newcommand{\cone}{\mathop{\rm Cone}\nolimits}
107: \newcommand{\gr}{\mathop{\rm Gr}\nolimits}
108: \newcommand{\wgr}{\mathop{\rm wGr}\nolimits}
109: \newcommand{\agr}{\mathop{\rm aGr}\nolimits}
110: \newcommand{\ogr}{\mathop{\rm OGr}\nolimits}
111: \newcommand{\wogr}{\mathop{\rm wOGr}\nolimits}
112: \newcommand{\pf}{\mathop{\rm Pf}\nolimits}
113: 
114: \renewcommand{\Box}{\square}
115: 
116: \newcommand{\color}[6]{}
117: 
118: \begin{document}
119: 
120: \begin{center}
121: {\LARGE Orbifold Riemann--Roch for threefolds} \\
122: \vspace{0.15in}
123: {\LARGE with an application to Calabi--Yau geometry} \\
124: \vspace{0.2in}
125: {\large Anita Buckley and Bal\'{a}zs Szendr\H{o}i}\\
126: \vspace{0.15in}
127: {\large June 2004}
128: \vspace{0.1in}
129: \end{center}
130: 
131: {\small
132: \begin{center} {\sc abstract} \end{center}
133: {\leftskip=30pt \rightskip=30pt
134: We prove an orbifold Riemann--Roch formula for a polarized 3--fold $(X,D)$. 
135: As an application, we construct new families of projective Calabi--Yau 
136: threefolds.
137: \par}}
138: 
139: \section*{Introduction}
140: \label{introduction}
141: The aim of this paper is to state and prove a Riemann--Roch (RR) formula for a 
142: pair $(X,D)$ of a normal projective threefold $X$ with quotient singularities and
143: a $\bq$-Cartier Weil divisor $D$. Under certain conditions, 
144: we prove the existence of a formula
145: $$\chi(X,\bigo_X(D))=\mbox{ RR--type expression in }D+\sum_P c_P(D)+\sum_C s_C(D),$$
146: where $c_P(D)$ is the contribution from a singular point $P$ of $(X,D)$ 
147: and $s_C(D)$ is the contribution from a curve $C$ of singularities. 
148: These contributions depend on the type of the quotient singularities and on the 
149: embedding of $\sing X$ in $X$.
150: 
151: Explicit formulae of this type for surfaces and threefolds with isolated 
152: canonical singularities were first studied by Fletcher~\cite{fletcher1} and 
153: Reid~\cite{ypg}. Earlier Kawasaki~\cite{kawasaki} had proved 
154: a general Riemann--Roch formula for orbifolds. His formula involves a 
155: sum over loci in $X$ with constant inertia group, and specifies to the above 
156: form in our case. Our proof relies on equivariant Riemann--Roch and a computation of 
157: intersection numbers on a resolution, and is independent 
158: of Kawasaki's; it gives the contributions in an explicit form (which could also 
159: be deduced from Kawasaki's result with extra work). Kawasaki's analytic formula was 
160: extended by Toen~\cite{toen} to the context of Deligne--Mumford 
161: stacks using algebraic methods.
162: 
163: Under further conditions on $(X,D)$, vanishing implies that 
164: $\chi(X,\bigo_X(nD))$ is simply the dimension of $H^0(X,\bigo_X(nD))$ for $n>0$.  
165: The finite dimensional vector spaces 
166: $H^0(X,\bigo_X(nD))$ fit together into a graded ring 
167: $$R(X,D)=\bigoplus_{n \geq 0} H^0(X,\bigo_X(nD)).$$ 
168: If $D$ is assumed ample as well as being $\bq$-Cartier, this ring is finitely 
169: generated. A surjection 
170: \[k[x_0, \ldots, x_n] \twoheadrightarrow R(X,D) \]
171: from a graded ring $k[x_0, \ldots, x_n]$ generated by variables $x_i$ of 
172: weights $a_i$ corresponds to an 
173: embedding $$i\colon X\cong\proj R(X,D)\hookrightarrow \bp(a_0,\ldots, a_n)$$ of~$X$ 
174: into a weighted projective space, with $\bigo_X(D)$ isomorphic 
175: to $\bigo_X(1)=i^{\ast}\bigo_{\bp}(1)$. 
176: 
177: Our aim is to construct quasi-smooth and well formed threefolds in weighted 
178: projective space. This implies in particular that the only singularities of $X$ are 
179: quotient singularities induced by the weights of the weighted projective space. 
180: Moreover, since we are interested in constructing threefolds with at worst 
181: canonical singularities, we restrict our study to curves of singularities which 
182: are generically of compound Du Val (cDV) type~\cite[Definition 2.1]{3-folds}. 
183: 
184: The main result of this paper is Theorem~\ref{thmdis}, which presents
185: an explicit RR formula of the above shape. As an application, we find new 
186: projective families of Calabi--Yau threefolds, via a study of their Hilbert 
187: series, which is brought into a compact form in Corollary~\ref{cyrhil}. 
188: The detailed and exhaustive analysis of families arising in this way, as well
189: as applications to other families of varieties such as Fanos and regular 
190: varieties of general type, will be presented elsewhere. 
191: 
192: \section{Definitions and notation}
193: \label{defnot}
194: 
195: We work over an algebraically closed field $k$ of characteristic zero. 
196: A $\bq$--\textit{divisor} on a normal variety $X$ is a formal linear combination 
197: of prime divisors with rational coefficients. 
198: A $\bq$--divisor $D$ is $\bq$--\textit{Cartier} if $mD$ is 
199: Cartier for some positive integer $m\neq 0$. In this case, 
200: if $C\subset X$ is a complete curve,
201: the \textit{degree} of $D$ on $C$ is defined by 
202: $$\deg D|_C=\frac{1}{m}\cdot \deg_C \bigo_X(mD).$$
203: If $X$ is projective, we also define the intersection $D\cdot c_2(X)$ as
204: $\frac{1}{m}f^{\ast}(mD)\cdot c_2(Y)$ computed on a resolution 
205: $f\colon Y\rightarrow X$, minimal over the cDV locus. 
206: A \textit{\cy threefold} is a normal projective threefold $X$ with canonical 
207: Gorenstein singularities, satisfying $K_X\sim \bigo_X$ (linear equivalence) 
208: and $H^1(X,\bigo_X)=0.$
209: 
210: A \textit{cyclic quotient singularity of type } $\frac{1}{r}(b_1,\ldots,b_n)$ is 
211: the quotient $\pi\colon \ba^n\rightarrow \ba^n/\mu_r$, where 
212: $\mu_r$ acts on $\ba^n$ by 
213: $$\mu_r\ni\varepsilon \colon\ (x_1,\ldots,x_n)\mapsto (\varepsilon^{b_1}x_1,\ldots,\varepsilon^{b_n}x_n)$$
214: We always assume that no factor of $r$ divides all the $b_i$, 
215: which is equivalent to the $\mu_r$-action being effective. 
216: The sheaf $\pi_{\ast}\bigo_{\ba^n}$ decomposes into eigensheaves
217: $$\bl_i=\{f\ | \ \varepsilon(f)=\varepsilon^i\cdot f\mbox{ for all } \varepsilon\in\mu_r\},$$
218: for $i=0,\ldots,r-1$. A singularity $Q\in X$ polarised by a Weil divisor $D$ is a
219: \textit{cyclic quotient singularity of type} 
220: $ _i\bigl(\frac{1}{r}(b_1,\ldots,b_n)\bigr),$
221: if $Q\in X$ is locally isomorphic to a point of type $\frac{1}{r}(b_1,\ldots,b_n)$ 
222: and $\bigo_X(D)\cong\bl_i$ near the singular point.
223: 
224: Let $(X,D)$ be a threefold $X$ containing a curve $C$ of singularities, 
225: equipped with a $\bq$-Cartier divisor $D$ which is Cartier away from $C$. 
226: Take a generic surface $S$ which intersects $C$ transversely 
227: in a finite number of points. Assume that every point in the 
228: intersection is a singular point of type $\ _k\bigl(\frac{1}{r}(1,-1)\bigr)$ 
229: on the polarised surface $(S, D|_S)$. Then $C\in X$ will be called 
230: a \textit{curve of singularities of transverse type} 
231: $\ _k\bigl(\frac{1}{r}(1,-1)\bigr)$, shorter a 
232: $\ _k\bigl(\frac{1}{r}(1,-1)\bigr)$ \textit{curve}, 
233: or often simply an $A_{r-1}$ \textit{curve}.
234: 
235: Note that every $A_{r-1}$ curve can contain a finite number 
236: of points of different type, 
237: which will be called \textit{dissident points.} A curve with dissident points is 
238: a \textit{dissident curve.} If an $A_{r-1}$ curve $C$ contains the 
239: dissident points $\{P_{\lambda}\ :\ \lambda\in\Lambda\}$ of types 
240: $\frac{1}{r \tau_{\lambda} }(a_{1\lambda},a_{2\lambda},a_{3\lambda})$, then 
241: we define the \textit{index of} $C$ to be 
242: \[\tau_C=\lcm_{\lambda\in\Lambda} \{\tau_{\lambda}\}.\] 
243: For an $A_{r-1}$ curve $C$ with no dissident points, $\tau_C=1.$
244: 
245: \begin{example}\rm The degree 13 hypersurface $X_{13}\subset \bp(1,1,2,3,6)$ in 
246: the weighted projective space with variables $x_1,x_2,y,z,t$ of the given degrees 
247: is a \cy threefold with two curves of singularities intersecting in a dissident 
248: point. The first $C_1=\{x_1=x_2=y=0\}$ is of type $\frac{1}{3}(1,2)$ and index~2 
249: because of the dissident point $(0,0,0,0,1)$ of type $\frac{1}{6}(1,2,3)$; 
250: likewise, $C_2=\{x_1=x_2=z=0\}$ is of type $\frac{1}{2}(1,1)$ and index~3.
251: \end{example}
252: 
253: For details on subvarieties in weighted projective space, consult~\cite{fletcher}. 
254: In particular, recall that a variety $X\subset\bp^{n}(a_0,\ldots,a_n)=\bp$ is 
255: \textit{quasi-smooth} if the affine cone $C_X\subset\ba^{n+1}$ over $X$ 
256: is smooth outside its vertex. In this case $X$ only has quotient singularities 
257: induced by the singularities of $\bp$. The weighted projective space 
258: $\bp(a_0,\ldots,a_n)$ is \textit{well formed} if 
259: $$(a_0,\ldots,\widehat{a_i},\ldots,a_n)=1 \ \mbox{ for each }i.$$ 
260: Moreover, $X$ of codimension $c$ in $\bp$ is \textit{well formed}, if it does 
261: not contain any $c+1$-codimensional singular stratum of $\bp$. 
262: Finally a pair $(X,D)$ of a variety $X$ and a $\bq$-divisor $D$ is quasi-smooth, 
263: respectively well formed, if there is an ample Cartier divisor $H$ on $X$, 
264: so that under the embedding $X\hookrightarrow\bp$ into a weighted projective 
265: space induced by the ample $\bq$-Cartier divisor $D+H$, the pair $X\subset\bp$ 
266: is quasi-smooth, respectively well formed.
267: 
268: \section{The Riemann--Roch formula}
269: \label{statethm} 
270: 
271: \subsection{The statement} 
272: 
273: \noindent Here is the main theorem of this paper:
274: 
275: \begin{theorem}
276: \label{thmdis}
277: Let~$(X,D)$ be a pair consisting of a normal projective threefold and a $\bq$-Cartier divisor, 
278: which is quasi-smooth and well formed. Assume further that the singularities 
279: of~$(X, D)$ consist of the following loci: 
280: \begin{itemize}
281: \item points $P\in X$ of type $_n\bigl(\frac{1}{s}(a_1,a_2,a_3)\bigr)$ (dissident and 
282: isolated), and 
283: \item curves $C\subset X$ of generic type $_k\bigl(\frac{1}{r}(1,-1)\bigr)$ with index 
284: $\tau_C$. 
285: \end{itemize}
286: Then for all positive integers $m$,
287: $$\begin{array}{rcl}
288: \chi(X,\bigo_X(mD)) &= & \chi(\bigo_X)+\frac{1}{12}mD(mD-K_X)(2mD-K_X)+m\displaystyle\frac{D\cdot c_2(X)}{12} \\
289:  & & \\
290:   & & \mbox{}+\displaystyle\sum_{P}c_P(mD)+\sum_{C}s_C(mD),
291: \end{array}$$ 
292: where
293: $$c_P(mD)=
294: \displaystyle\frac{1}{s}\sum_{\substack{\ep\in\mu_s \\ \ep^{a_i}\ne1\,\forall i = 1,2,3}}
295: \frac{\varepsilon^{-nm}-1}{(1-\varepsilon^{-a_1})(1-\varepsilon^{-a_2})(1-\varepsilon^{-a_3})}$$
296: and
297: $$\begin{array}{rcl}
298: s_C(mD)& = & -m\displaystyle\frac{\ol{mk}(r-\ol{mk})}{2r}\deg D|_C + \frac{\ol{mk}(r-\ol{mk})}{4r}\deg K_X|_C \\
299:  & & \\
300:       & & +\displaystyle\frac{\ol{mk}(r-\ol{mk})(r-2\cdot\ol{mk})}{12r^2\tau_C}N_C,   
301: \end{array}$$
302: where the integer $N_C$ is an invariant of $X$ in a neighbourhood of~$C$, and 
303: $\overline{\phantom{\Sigma}}$ denotes the smallest residue mod~$r$. 
304: \end{theorem}
305: 
306: \begin{remark}\rm
307: Note that a point $P\in X$ of type $_n\bigl(\frac{1}{s}(a_1,a_2,a_3)\bigr)$ is either
308: \begin{enumerate}
309: \item an isolated point singularity if $\hcf(a_i,s)=1$ for all $i=1,2,3,$\ \ or 
310: \item a dissident point on some curve if $\hcf(a_i,s)=\alpha_i\neq 1$ for some $i$. 
311: \end{enumerate}
312: In the first case, $c_P(mD)$ equals to the so-called basket contribution 
313: to Riemann--Roch~\cite{ypg}. In the second case, as the singularities along curves are 
314: of $A_{\alpha_i-1}$ transverse type, we must have $a_j+a_k=0$ mod $\alpha_i$ for different 
315: indexes $i,j,k$. Since $X$ is well formed, $a_i,a_j,s$ have no common divisor.
316: \end{remark}
317: 
318: \subsection{The outline of the proof}
319: \label{outline}
320: 
321: Let $(X,D)$ be a pair satisfying the conditions of Theorem~\ref{thmdis}. 
322: Choose a projective resolution $f\colon Y\rightarrow X$ which is crepant at the generic point of 
323: each one-dimensional component of the singular locus of $(X,D)$. The sheaf $\bigo_X(D)$ is 
324: a rank-1 reflexive sheaf on $X$; define $\L= f^*(\bigo_X(D))/({\rm torsion})$, 
325: a rank-one torsion-free sheaf on $Y$. Let also $H$ be an ample Cartier divisor on $X$. 
326: 
327: For some integers $n,N$, there exists a surjection 
328: $\bigo_X^N(-nH)\twoheadrightarrow\bigo_X(D)$ on $X$ 
329: which pulls back to a surjection 
330: $\bigo_Y^N\otimes f^*\bigo_X(-nH)\twoheadrightarrow\L$ on $Y$ 
331: and hence gives an exact sequence
332: $$0\to \bk \to \bigo_Y^N\otimes f^*\bigo_X(-nH) \to\L \to 0.$$
333: Under $f_\ast$, this becomes the long exact sequence
334: $$\begin{array}{cccccccc}
335: 0\to & f_{\ast}\bk    & \to & \bigo_X^N(-nH) & {\to} & f_{\ast}\L  & \to \\[6pt]
336:                  & R^1f_{\ast}\bk & \to & 0 &
337: \to                   & R^1f_{\ast}\L & \to \\[6pt]
338:                  & R^2f_{\ast}\bk &\to  & 0 &
339: \to                   & R^2f_{\ast}\L & \to  & 0,
340: \end{array}$$
341: where we used $f_{\ast}\bigo_Y\cong\bigo_X$ which holds since $X$ is normal, and 
342: the projection formula together with $R^if_{\ast}\bigo_Y=0$ for $i>0$ which holds 
343: as quotient singularities in characteristic zero are rational. It is easy to
344: see that $f_\ast\L\cong\bigo_X(D)$; from the exact sequence we also have 
345: $R^2f_{\ast}\L=0$ and $R^1f_{\ast}\L\cong R^2f_{\ast}\bk$, the latter 
346: necessarily supported on the isolated and dissident singularities of $(X,D)$.
347: 
348: The reflexivization of the rank-one torsion-free sheaf $\L$ on the smooth variety
349: $Y$ is a line bundle $\bigo_Y(D_Y)$. As $\bigo_X(D)$ is saturated, we also have
350: $f_\ast\bigo_Y(D_Y)\cong\bigo_X(D)$. Thus the exact sequence
351: \[ 0 \to \L \to \bigo_Y(D_Y) \to \Q \to 0\]
352: give rise to an injection
353: \[ f_\ast \Q \hookrightarrow R^1f_\ast \L \cong R^2 f_\ast \bk, \] 
354: proving that $f_\ast Q$ is supported on the isolated and dissident singular points 
355: of $(X,D)$. As $\Q$ is supported in codimension two on $Y$, this implies that all its 
356: higher pushforwards are also supported on these singular points.
357: 
358: Finally, by the Leray spectral sequence
359: \[
360: \chi(Y, \L) = \sum_i (-1)^i \chi(X, R^if_{\ast}\L), 
361: \]
362: which can be rewritten, using the above exact sequences and isomorphisms, as
363: \begin{eqnarray}\label{firststep}
364: \chi(X,\bigo_X(D)) & = & \chi(Y, \bigo_Y(D_Y)) + \chi(X, R^1f_{\ast}\bl) - \sum_i (-1)^i \chi(X, R^if_\ast \Q)\nonumber\\
365: & = & \chi(Y, \bigo_Y(D_Y)) + \P_1,
366: \end{eqnarray}
367: where, by our earlier remarks, $\P_1$ is a contribution from sheaves supported entirely 
368: on the isolated and dissident singular points of $(X,D)$. Our arguments in fact imply 
369: that this contribution is local in a stronger sense: it only depends on the analytic type 
370: of the isolated and dissident quotient singularities of $(X,D)$. 
371: This holds since the constructions of $\L$ and $\Q$ are universal, and an analytic 
372: isomorphism preserving the type of the quotient singularities 
373: necessarily gives an analytic isomorphism between these sheaves, and 
374: thus an equality of Euler characteristics. 
375: 
376: In the next step we will express $D_Y$ and $K_Y$ in terms of $D, K_X$ and 
377: the exceptional divisors of the resolution. For a singular point $P\in X$, let $\{{}^P\! G_j\}$
378: be the exceptional surfaces mapping to $P$ under $f$; similarly, for a curve $C\subset X$ 
379: of singularities, let $\{{}^C\! E_i\}$ and be the exceptional surfaces
380: mapping surjectively to $C$. Note that every dissident or isolated singular point $P$ 
381: is locally analytically isomorphic to $\ba^3/\mu_s$, and the configuration 
382: of $\{{}^P\! G_j\}$ depends only on the analytic singularity type of $P$. 
383: 
384: As $f$ is crepant at the generic point of each curve $C$, we have 
385: $$K_Y=f^{\ast}K_X+N,\mbox{ where } N=\sum_P\sum_j\,^P\!\gamma_{j}\, ^P\!G_j;$$
386: also
387: $$f^{\ast}D=D_Y+\sum_{C}R_C+M, \mbox{ where } R_C=\sum_i\,^C\!\alpha_{i}\,^C\!E_i 
388: \mbox{ and } M=\sum_P\sum_j\,^P\! \beta_{j}\,^P\!G_j,$$
389: with ${}^C\!\gamma_j,{}^C\!\alpha_i, {}^P\!\beta_j\in\bq$. Here $f^{\ast}D$ is by 
390: definition $\frac{1}{m}f^{\ast}(mD)$ for an integer $m$ which makes $mD$ Cartier.  
391: 
392: \begin{lemma}
393: \label{lemconfig} If $\overline{\phantom{\Sigma}}$ denotes the smallest residue mod~$r$, then
394: $$R_C=\displaystyle\frac{k}{r}\,^C\!E_1+\frac{\overline{2k}}{r}\, ^C\!E_2+\ldots+\frac{\overline{(r-1)k}}{r}\, ^C\!E_{r-1}.$$
395: \end{lemma}
396: 
397: \begin{prf} Choose a general transverse hypersurface $S$ and a point $Q\in S\cap C$. 
398: By assumption, on the polarised $(S,D|_S)$ the type of 
399: $Q$ is $_k\bigl(\frac{1}{r}(1,-1)\bigr)$.
400: Such a point has a unique crepant resolution 
401:  $f|_{\tilde{S}}\colon \tilde{S}\rightarrow S$, and we necessarily have 
402: $$(f|_{\tilde{S}})^\ast D|_S=D_Y|_{\tilde{S}}+
403: \frac{k}{r}F_1+\frac{\overline{2k}}{r}F_2+\ldots+
404: \frac{\overline{(r-1)k}}{r}F_{r-1},$$
405: where the $F_i$ are exceptional lines in $\tilde S$ forming an~$A_{r-1}$ 
406: configuration. 
407: 
408: Each $F_i$ is a fibre in one of the surfaces $^C\!E_j$. Since $X$ is assumed quasi-smooth 
409: and well formed, there is no ramification in codimension 1; thus $k$ and $r$ are coprime.
410: Therefore the integers $k, \overline{2k},\ldots,\overline{(r-1)k}$ 
411: are all different, which implies that $F_1,\ldots,F_{r-1}$ must be fibres in different 
412: surfaces $^C\!E_1,\ldots,^C\!E_{r-1}$. Indeed, if some $F_i$ would be fibres in the same
413: surface, the relevant coefficients should be equal, arising from an irreducible
414: divisor on the threefold. The statement follows. 
415: \end{prf}
416: \ethrm
417: 
418: In particular, there are $(r-1)$ irreducible exceptionals
419: ${^C\!E_1},\ldots,{^C\!E_{r-1}}$ over every $C$, each isomorphic to a blowup in a few points 
420: (over the dissident points) of a smooth surface geometrocally ruled over $C$.   
421: Moreover,  
422: \label{conref}
423: $^C\!E_i$ and $^C\!E_j$ intersect along a curve isomorphic to $C$ if $j=i+1$ and are 
424: disjoint otherwise. The situation is shown on Figure~\ref{pict5}. For two different curves $C$ 
425: and $\bar{C}$, the surfaces $^C\!E_i$ and $ ^{\bar{C}}\!E_j$  intersect only over the 
426: dissident points.
427: 
428: \begin{figure}[ht] 
429: \begin{center}
430: \input{conf.pstex_t}
431: \caption{Resolution of an $A_{r-1}$ curve $C$ with a dissident point $P$}
432: \label{pict5}
433: \end{center}
434: \end{figure}
435: 
436: We proceed to show that many intersection numbers between these divisors actually vanish. 
437: Choose an ample Cartier divisor $H$ on $X$ such that $D+H$ is also ample. 
438: Take $t$ large and divisible, so that the linear systems $|\,t\,(D+H)|$ and $|tH|$
439: contain nonsingular divisors $S_1$ and $S_2$ respectively.
440: Then $f^{\ast}S_1\cdot{^C\!E_i}$ is a finite union of fibres in $^C\!E_i$, since
441: $f^{\ast}S_1|_{^C\!E_i}$ is a pullback of $t(D+H)|_{C}$; the same holds for
442: $f^{\ast}S_2\cdot{^C\!E_i}$. Then 
443: \begin{eqnarray*} f^{\ast}(tD)\cdot f^{\ast}(tD)\cdot {^C\!E_i}&=&(f^{\ast}S_1-f^{\ast}S_2)\cdot (f^{\ast}S_1-f^{\ast}S_2) \cdot {^C\!E_i}\\ 
444: & = & f^{\ast}S_1\cdot (\mbox{fibres}) - f^{\ast}S_2\cdot (\mbox{fibres})\\ 
445: & = & S_1\cdot f_{\ast}(\mbox{fibres}) - S_2\cdot f_{\ast}(\mbox{fibres})\\
446: &=&0.
447: \end{eqnarray*}
448: using also the projection formula. Thus
449: $f^{\ast}D\cdot f^{\ast}D\cdot R_C=0$
450: and similarly $f^{\ast}D\cdot f^{\ast}K_X\cdot R_C=0.$
451: The projection formula also gives
452: $$(f^{\ast}D)^3=D^3,\ (f^{\ast}D)^2\cdot f^{\ast}K_X=D^2K_X,\ f^{\ast}D\cdot(f^{\ast}K_X)^2=DK_X^2.$$
453: The intersection numbers involving $M$ or $N$
454: depend only on the isolated or dissident singular points of $X$.
455: Moreover, for different curves of singularities, the intersections 
456: $$f^{\ast}D\cdot R_C R_{\bar{C}},\ f^{\ast}K_X\cdot R_C R_{\bar{C}}\ \mbox{ and }\ (R_C)^2R_{\bar{C}}$$
457: can only be nonzero because of intersections over dissident points.
458: 
459: Finally using~(\ref{firststep}), Riemann--Roch for the smooth threefold $Y$, and the birational
460: invariance of $\chi(\bigo_X)$, we obtain
461: \begin{eqnarray}
462: \label{eqena}
463: %\chi(\bigo_Y) +\frac{1}{12}D_Y(D_Y-K_Y)(2D_Y-K_Y)+\frac{1}{12}D_Yc_2(Y)
464: \chi(X, \bigo_X(D))& = & \chi(\bigo_X)+\frac{1}{12}D(D-K_X)(2D-K_X)+\frac{1}{12}f^{\ast}D\cdot c_2(Y) + \nonumber\\
465: & &  \frac{1}{2}f^{\ast}D \cdot \sum_CR_C^2-\frac{1}{4}f^{\ast}K_X \cdot \sum_CR_C^2 -\nonumber\\
466: & &  \frac{1}{6}\sum_CR_C^3-  \frac{1}{12}\sum_CR_C\cdot c_2(Y)+ \P_2. 
467: \end{eqnarray}
468: Here $\P_2$ denotes a quantity that depends, in addition to the earlier quantity $\P_1$, on 
469: intersection numbers of divisors over the isolated and dissident singularities of $(X,D)$. 
470: Since the latter are purely analytical, $\P_2$ also depends only on the analitic type
471: of these singularities. 
472: 
473: \subsection{The contribution from a curve of singularities} 
474: 
475: In this section we focus on the part of the contribution that depends only on a small 
476: analytic neighbourhood of the curves of singularities. During the argument we will
477: often meet divisors and intersection numbers which depend on the configuration of 
478: exceptional divisors over the dissident points, and thus depend only on the analytic 
479: type of these points. We will denote all such divisors by $\D$ and numerical 
480: contributions by $\P$, and will not worry about the exact expressions. 
481: 
482: The next two lemmas will analyze the intersection numbers in~(\ref{eqena}).
483: 
484: \begin{lemma}
485: \label{lemena}
486: For a $\ _k\bigl(\frac{1}{r}(1,-1)\bigr)$ curve $C$ of singularities of $(X,D)$,  
487: $$\frac{1}{2}f^{\ast}D\cdot R_{C}^2=-\deg D|_C\cdot \frac{\ol{k}(r-\ol{k})}{2r}$$
488: and  
489: $$\frac{1}{4}f^{\ast}K_X\cdot R_{C}^2=-\deg K_X|_C\cdot \frac{\overline{k}(r-\overline{k})}{4r}. $$
490: \end{lemma}
491: 
492: \begin{prf}
493: If $S\subset X$ is surface which intersects $C$ transversely in all points, then
494: the points in $C\cap S$ are all of type $\ _k\bigl(\frac{1}{r}(1,-1)\bigr)$ on $(S,D|_S)$. 
495: Riemann--Roch for a resolution $f|_{\tilde{S}}\colon \tilde{S}\rightarrow S$ of these 
496: points gives
497: $$  \chi(S, \bigo_S(D))=\chi(\bigo_S)+\frac{1}{2}D|_S(D|_S-K_S)+\frac{1}{2}B_{\ol{k}}^2,$$
498: as in the proof of Theorem 9.1 in \cite{ypg}. Here $B_{\ol{k}}=R_C|_{\tilde{S}}$
499: is the $\bq$-divisor on the surface $\tilde S$ contributing the so-called basket contribution 
500: $$\frac{1}{2}B_{\ol{k}}^2 = \#(C\cap S)\cdot\left(-\frac{\overline{k}(r-\overline{k})}{2r}\right).$$
501: As before, for some integer $t$ we can write $tD$ as a difference of two 
502: smooth surfaces $S_1, S_2$ both intersecting $C$ transversally; thus
503: $$
504: \frac{1}{2}f^{\ast}D\cdot R_{C}^2=\frac{1}{2t}(f^{\ast}S_2-f^\ast S_1)\cdot R_{C}^2=  
505: -\deg D|_C\cdot \frac{\overline{k}(r-\overline{k})}{2r}.
506: $$
507: A same argument also shows
508: $$\frac{1}{4}f^{\ast}K_X\cdot R_{C}^2=-\deg K_X|_C\cdot \frac{\overline{k}(r-\overline{k})}{4r}$$
509: which ends the proof of Lemma \ref{lemena}.
510: \end{prf}
511: \ethrm
512: 
513: \begin{lemma}
514: \label{lemdva}
515: Let $C$ be a $\, _k\bigl(\frac{1}{r}(1,-1)\bigr)$ curve in $(X,D)$. Then
516: $$\begin{array}{l}
517: R_{C}^3 + \frac{1}{2}R_{C}\cdot c_2(Y)=  \\
518: \\
519: -\frac{1}{r^2}\ol{k}(r-\ol{k})(r-2\ol{k})
520:  \left[(r-2)\left(1-g+\frac{1}{2}K_Y\,^C\!E_1\,^C\!E_2\right)+\,^C\!E_{r-2}^2\,^C\!E_{r-1}\right]+\P,
521: \end{array}$$
522: where $g$ is the genus of $C$.
523: \end{lemma}
524: 
525: \begin{prf}
526: >From the definition of $R_C$,
527: $$
528: R_{C}^3 =  \sum_{i=1}^{r-1} \left(\displaystyle\frac{\ol{ik}}{r}\right)^3E_i^3+3\sum_{i=1}^{r-2} \left(\displaystyle\frac{\ol{ik}}{r}\right)^2\frac{\ol{(i+1)k}}{r}E_i^2E_{i+1}+
529: \frac{\ol{ik}}{r}\left(\frac{\ol{(i+1)k}}{r}\right)^2E_iE_{i+1}^2
530: $$
531: and
532: $$R_{C}\cdot c_2(Y)  =\left(\frac{\ol{k}}{r}E_1+\frac{\ol{2k}}{r}E_2+\ldots+\frac{\ol{(r-1)k}}{r}E_{r-1}
533: \right)\cdot c_2(Y).$$
534: We will simplify these expressions using the properties of ruled surfaces.
535: Recall that each $E_i$ is isomorphic to a smooth ruled surface $\hat{E}_i$ blown up in a number 
536: of points. The blowups $E_i\stackrel{\pi_i}{\to}\hat{E}_i$ happen only over the 
537: dissident points. In other words, the resolution
538: $f\colon Y\rightarrow X$ contracts Exc$(\pi_i)\subset E_i$ to the dissident points on~$C$.
539: 
540: Denote $\gamma_i=E_i\cap E_{i+1}$ which is isomorphic to $C$ of genus $g.$
541: Note that $\pi_i(\gamma_i)=\pi_{i+1}(\gamma_i)$ is a section in both ruled surfaces $\hat{E}_i$ and $\hat{E}_{i+1}$. 
542: Then
543: $$K_{E_i}=\pi_i^{\ast}K_{\hat{E}_i}+ \D \sim -2\gamma_i+(2g-2+\gamma_i^2)f_i+\D,$$ 
544: where $f_i$ is a generic fibre on $E_i$. In particular 
545: $K_{E_i}^2=8(1-g)+\P.$ Similarly
546: $$K_{E_{i+1}}\sim -2\gamma_i+(2g-2+\gamma_i^2)f_{i+1}+\D.$$ 
547: By the adjunction formula we can compute
548: \begin{eqnarray*}
549: (\gamma_i^2)_{E_{i+1}}& = & (K_{E_i}-K_Y|_{E_i})E_{i+1}|_{E_i}\\
550: & = & -(\gamma_i^2)_{E_i}+2(g-1)-K_YE_iE_{i+1}+\P, 
551: \end{eqnarray*}
552: where $(\gamma^2)_E$ is the self-intersection of the curve $\gamma$ computed in the surface
553: $E\supset\gamma$. 
554: Similarly 
555: $$(\gamma_i^2)_{E_i}=-(\gamma_i^2)_{E_{i+1}}+2(g-1)-K_YE_iE_{i+1}+\P.$$
556: Together these give \label{rulled}
557: $$E_i^2E_{i+1}+E_iE_{i+1}^2=2(g-1)-K_YE_iE_{i+1}+\P\ \mbox{ for }i=1,\ldots,r-2.$$
558: 
559: Next, 
560: $$E_i^2E_{i+1}+E_{i+1}E_{i+2}^2=(\gamma_i^2)_{E_{i+1}}+(\gamma_{i+1}^2)_{E_{i+1}}+\P
561: \ \hbox{ for }\ i=1,\ldots,r-3$$
562: since $\pi_{i+1}(\gamma_i)$ and $\pi_{i+1}(\gamma_{i+1})$ are disjoint 
563: sections in~$\hat{E}_{i+1}.$
564: 
565: By the projection formula also the following holds:
566: $$K_YE_{i-1}E_{i}-K_YE_iE_{i+1}=\P\mbox{ and }K_YE_i^2+2K_YE_iE_{i+1}=\P.$$
567: 
568: Finally, putting together all of the above yields
569: $$\begin{array}{rcl}
570: E_i^2E_{i+1} & =&-(r-2-i)2(g-1)+(r-2-i)K_YE_1E_2+E_{r-2}^2E_{r-1}+\P,  \\[6pt]
571: E_iE_{i+1}^2 & =&(r-1-i)2(g-1)-(r-1-i)K_YE_1E_2-E_{r-2}^2E_{r-1}+\P, 
572: \end{array}$$
573: for all $i=1,\ldots,r-2$. 
574: 
575: Expressions involving the second Chern class $c_2(Y)$ can also be computed without 
576: difficulty. Let $\bn_i$ be the normal bundle of $E_i$ in 
577: the smooth threefold $Y$, and $\bt_{E_i},\ \bt_Y$ the tangent sheaves. 
578: There is an exact sequence
579: $$0\to \bt_{E_i}\to j_i^{\ast}\bt_Y \to \bn_i \to 0,$$
580: where $j_i\colon E_i\rightarrow Y$ is the inclusion.
581: Comparison of Chern polynomials and adjunction gives
582: $$
583: c_2(Y)\cdot E_i= c_2(E_i)-K_{E_i}E_i|_{E_i} = c_2(E_i)-(K_Y+E_i)E_i^2. 
584: $$ 
585: 
586: Also $c_2(E_i)$ can be expressed with other invariants
587: $$\frac{1}{12}\left(K_{E_i}^2+c_2(E_i)\right)=1+p_a(E_i)=1-g.$$
588: This in particular implies $c_2(E_i)=4(1-g)+\P.$
589: From
590: $$8(1-g)+\P=K_{E_i}^2=(K_Y+E_i)^2E_i=E_i^3+2K_YE_i^2+\P$$
591: we get
592: $$c_2(Y)\cdot E_i=-\frac{1}{2}E_i^3+\P=4(g-1)-2K_YE_1E_2+\P.$$
593: 
594: This reduces the expression for $R_C^3$ to
595: $$\begin{array}{l}
596: \displaystyle\frac{8(1-g)+4K_YE_1E_2}{r^3}\sum_{i=1}^{r-1}\left(\ol{ki}\right)^3+ \\
597: \displaystyle\frac{3}{r^3}\sum_{i=1}^{r-2} (\ol{ik})^2\ol{(i+1)k}\Big(-2(r-2-i)(g-1)+ (r-2-i)K_YE_1E_2+ E_{r-2}^2E_{r-1}\Big)+\\
598: \displaystyle\frac{3}{r^3}\sum_{i=1}^{r-2} \ol{ik}(\ol{(i+1)k})^2\Big(2(r-1-i)(g-1)- (r-1-i)K_YE_1E_2- E_{r-2}^2E_{r-1}\Big)+ \P 
599: \end{array}$$
600: and $R_{C}\cdot c_2(Y)$ reduces to
601: $$-\frac{4(1-g)+2K_YE_1E_2}{r}\sum_{i=1}^{r-1} \ol{ki}+\P.$$
602: 
603: Arguing as in \cite{morrison}, we can simplify 
604: $R_{C}^3 + \frac{1}{2}R_{C}\cdot c_2(Y)$ to the form
605: $$-\frac{1}{r^2}\ol{k}(r-\ol{k})(r-2\ol{k})
606: \left[(r-2)\left(1-g+\frac{1}{2}K_YE_1E_2\right)+E_{r-2}^2E_{r-1}\right]+\P.$$
607: This ends the proof of Lemma~\ref{lemdva}.
608: \end{prf}
609: \ethrm
610: 
611: \begin{corollary}
612: \label{rrformm}
613: Let $(X,D)$ satisfy the conditions of Theorem~\ref{thmdis}. Then for all positive integers $m$, 
614: \begin{eqnarray*}
615: \chi(X,\bigo_X(mD))&= & \chi(\bigo_X)+\frac{1}{12}mD(mD-K_X)(2mD-K_X) + m\frac{D\cdot c_2(X)}{12} +\\
616:  & &\\
617:  & &\sum_{C}s_C(mD)+ \P_3
618: \end{eqnarray*} 
619: where
620: $$\begin{array}{c}
621: s_C(mD) =-m\displaystyle\frac{\ol{mk}(r-\ol{mk})}{2r}\deg D|_C + \frac{\ol{mk}(r-\ol{mk})}{4r}\deg K_X|_C + \\
622:  \\
623:   \displaystyle\frac{\ol{mk}(r-\ol{mk})(r-2\cdot \ol{mk})}{6r^2}\left( (r-2)(1-g)+
624:  \frac{r-2}{2}K_Y\,^C\!E_1\,^C\!E_2+\,^C\!E_{r-2}^2\,^C\!E_{r-1}\right),   
625: \end{array}$$
626: and $\P_3$ is a contribution from the dissident and isolated singular points of $(X,D)$. 
627: \end{corollary}
628: 
629: \begin{prf}
630: For $m=1$, we only need to put the results of Lemma~\ref{lemena} and
631: Lemma~\ref{lemdva} into Formula~(\ref{eqena}) on page~\pageref{eqena}.
632: For $m>1$, consider $\bigo_X(mD)$, which is of transverse type 
633: $_{\ol{mk}}\bigl(\frac{1}{r}(1,-1)\bigr)$ on $C$, and repeat the proof using
634: $$f^{\ast}(mD)=D^{(m)}+\sum_{C}R_C^{(m)}+\D,$$
635: where $D^{(m)}$ is a Cartier divisor on $Y$, and 
636: $$R_C^{(m)}=\frac{\overline{mk}}{r}\,^C\!E_1+\frac{\overline{2mk}}{r}\, ^C\!E_2+\ldots+
637: \frac{\overline{(r-1)mk}}{r}\,^C\!E_{r-1}.$$
638: \end{prf}
639: \ethrm
640: 
641: \subsection{The contribution from a dissident or isolated singular point}
642: \label{three}
643: 
644: The analytically invariant contributions to RR from dissident 
645: or isolated singular points, denoted by $\P_3$ in Corollary~\ref{rrformm},
646: can be computed on any model that contains such singularities. 
647: We begin by showing the existence of such projective varieties.
648: 
649: \begin{proposition}
650: \label{exag}
651: Fix positive integers $s$ and $a_1,a_2,a_3$, and assume that $a_i,a_j,s$ have no common factor 
652: for all different $i,j\in\{1,2,3\}$. 
653: There exists a smooth projective 
654: 3--fold $Z$ together with an action of $\mu_s$ with the following properties: 
655: the action fixes a number of points on which a generator $\varepsilon\in\mu_s$ acts by 
656: $$\varepsilon\colon\ z_1,z_2,z_3\ \mapsto\ \varepsilon^{a_1}z_1,\varepsilon^{a_2}z_2,\varepsilon^{a_3}z_3.$$
657: If $\hcf(a_i,s)=\alpha_i\neq 1$, these points lie on curves
658: which are fixed by $\varepsilon^{\frac{s}{\alpha_i}}.$
659: Finally $\varepsilon^{\frac{s}{\alpha_i}}\in\mu_{\alpha_i}$ acts in the normal direction 
660: of each curve by
661: $$\varepsilon^{\frac{s}{\alpha_i}}\colon\ z_j,z_k\ \mapsto\ \varepsilon^{\frac{s}{\alpha_i}a_j} z_j,
662: \varepsilon^{\frac{s}{\alpha_i}a_k}z_k,\ \mbox{ for } j,k\in\{1,2,3\}-\{i\}$$
663: and freely away from the curve. 
664: \end{proposition}
665: 
666: \begin{prf}
667: We imitate the proof of (8.4) in~\cite{ypg}. Choose an integer $l\geq 3$ and consider action of
668: $\mu_s$ on $\bp^{l+3}(1,1,\ldots,1)$ given by 
669: $$x_1,x_{2},x_{3},x_{4},\ldots,x_{l+4}\ \mapsto\ 
670: \varepsilon^{a_1} x_{1},\varepsilon^{a_2} x_{2},\varepsilon^{a_3}x_{3},x_{4},\ldots,x_{l+4}.$$
671: This action fixes $\bp^{l}=\{x_1=x_2=x_3=0\}$ and acts in the normal direction by
672: $$\ x_1,x_2,x_3\ \mapsto\ \varepsilon^{a_1}x_1,\varepsilon^{a_2}x_2,\varepsilon^{a_3}x_3.$$
673: If $\hcf(a_i,s)=\alpha_i\neq 1$, the action is not free on $\bp^{l+1}=\{x_j=x_k=0\}$ for $j,k\in\{1,2,3\}-\{i\}$. 
674: This is fixed by 
675: $\varepsilon^{\frac{s}{\alpha_i}}$ which acts in the normal direction by
676: $$x_j,x_k\ \mapsto\ \varepsilon^{\frac{s}{\alpha_i}a_j} x_j,
677: \varepsilon^{\frac{s}{\alpha_i}a_k}x_k.$$
678: Another locus on which the action 
679: might not be free, is $\{x_{4},\ldots,x_{l+4}=0\}$. We will avoid this locus by defining
680: $$X\subset\bp^{l+3}/\mu_s,$$
681: as a complete intersection of $l$ general very ample divisors. Let $Z$ be the inverse image of 
682: $X$ under the quotient $\bp^{l+3}\rightarrow \bp^{l+3}/\mu_s $. 
683: Such $Z$ clearly satisfies the conditions in the proposition. 
684: \end{prf}
685: \ethrm
686: 
687: Let $X$ be a projective threefold with a singularity of type $\frac{1}{s}(a_1,a_2,a_3)$ 
688: as described in Proposition~\ref{exag}. 
689: We restricted our study to Du Val curves, which means that if
690: $\hcf(a_i,s)=\alpha_i\neq 1,$ then
691: $$a_j+a_k=0\ \mbox{ mod }\ \alpha_i.$$ 
692: Thus $\sing X$ equals to a number of dissident 
693: $$A_{\alpha_i-1}=\frac{1}{\alpha_i}(a_j,a_k)$$ curves, which intersect in dissident 
694: points of type
695: $\frac{1}{s}(a_1,a_2,a_3).$
696: 
697: Write $\pi\colon Z\rightarrow X$ for the quotient map 
698: and let $\bl_{n}$ be the $n$th eigensheaf of the action of 
699: $\varepsilon\in\mu_s$ on $\pi_{\ast}\bigo_Z.$ Then
700: $$\pi_{\ast}\bigo_Z=\bigoplus_{n=0}^{s-1}\bl_{n}$$
701: implies
702: $$H^p(Z,\bigo_Z)=\bigoplus_{n=0}^{s-1}H^p(X,\bl_{n}).$$
703: The group action on any $f\in\bl_{n}$ is given by $\varepsilon (f)=
704: \varepsilon^n \cdot f$, thus
705: $$\mbox{Tr}(\varepsilon : H^p(Z,\bigo_Z))=
706: \sum_{n=0}^{s-1}h^p(X,\bl_{n})\cdot \varepsilon^{n}. $$
707: Moreover,
708: $$\sum_p(-1)^p\mbox{Tr}(\varepsilon : H^p(Z,\bigo_Z))=
709: \sum_{n=0}^{s-1}\chi(X,\bl_{n})\cdot \varepsilon^{n}. $$
710: In order to simplify the notation, denote $\sum_p(-1)^p\mbox{Tr}(\varepsilon : H^p(Z,\bigo_Z))$ by $A_{\varepsilon}.$
711: Then
712: $$\sum_{n=0}^{s-1}\chi(X,\bl_{n})\cdot \varepsilon^{n}=A_{\varepsilon} \ \ \mbox{ and } \ \ \sum_{n=0}^{s-1}\chi(X,\bl_{n})=\chi(\bigo_Z).$$
713: 
714: The last two formulas can be considered as a linear system of $s$ equations in $\chi(X,\bl_{n})$ and 
715: variable $\varepsilon$, 
716: $$\left(\begin{array}{ccccc}
717: 1 &  1            &     1         & \ldots & 1 \\
718: 1 &  \varepsilon  & \varepsilon^2 & \ldots & \varepsilon^{s-1}  \\
719: 1 & \varepsilon^2 & \varepsilon^4 & \ldots &  \varepsilon^{2(s-1)}  \\
720: \ldots  & \ldots  &  \ldots       & \ldots &  \\
721: 1 & \varepsilon^{s-1} & \ldots    &  \ldots  & \varepsilon 
722: \end{array}\right)
723: \left(\begin{array}{c}
724: \chi(X,\bl_{0})\\
725: \chi(X,\bl_{1})\\
726: \chi(X,\bl_{2})\\
727: \vdots \\
728: \chi(X,\bl_{s-1})
729: \end{array}\right)=
730: \left(\begin{array}{c}
731: \chi(\bigo_Z)\\
732: A_{\varepsilon}\\
733: A_{\varepsilon^2}\\
734: \vdots\\
735: A_{\varepsilon^{s-1}}\\
736: \end{array}\right).$$
737: Write $\chi(X,\bl_{0})=\chi(\bigo_X)$ and eliminate $\chi(\bigo_Z)$ from the 
738: solution. We end up with 
739: \begin{eqnarray}
740: \label{eqypg}
741: \chi(X,\bl_{n}) &=&\chi(\bigo_X)+\frac{1}{s}\sum_{j=1}^{s-1}(\varepsilon^{-jn}-1)A_{\varepsilon^j}  \\
742: &=&\chi(\bigo_X)+\frac{1}{s}\sum_{\varepsilon\in\mu_s}(\varepsilon^{-n}-1)A_{\varepsilon}. \nonumber
743: \end{eqnarray}
744: 
745: We can compute $A_{\varepsilon}$ for all $\varepsilon\in\mu_s$ using the 
746: Atiyah--Singer--Segal equivariant RR formula~\cite{as},~\cite[p.565]{atiyah}.
747: If $\varepsilon^{a_i}\neq 1$ for all $i=1,2,3$, then the fixed locus of 
748: $\varepsilon$ is a number of points. In this case, for each point
749: $$
750: A_{\varepsilon}=\sum_p(-1)^p\mbox{Trace}(\varepsilon|H^p(Z,\bigo_Z)) = 
751: \frac{1}{(1-\varepsilon^{-a_1})(1-\varepsilon^{-a_2})(1-\varepsilon^{-a_3})}.
752: $$
753: If, on the other hand $\varepsilon^{a_i}= 1$ for one of the $i=1,2,3,$
754: then the fixed locus of $\varepsilon$ is a curve $C_i$. In this case, 
755: $\hcf(a_i,s)=\alpha_i\neq 1$ and $\varepsilon$ is an element of $\mu_{\alpha_i}$. 
756: The action in the normal direction equals 
757: $$\varepsilon^{a_j},\varepsilon^{a_k}=\varepsilon^{a_j},\varepsilon^{-a_j}.$$
758: Thus, we can use the equivariant Riemann--Roch formula for the cyclic group 
759: $\mu_{\alpha_i}$. Let $C_i$ have genus $g_i$, and let $x_1,x_2$ denote the first 
760: Chern classes of $\bn_{C_i/Z}=\bn_1\oplus\bn_2$. We get
761: $$\begin{array}{rcl}
762: A_{\varepsilon} & = & \sum_p(-1)^p\mbox{Trace}(\varepsilon|H^p(Z,\bigo_Z)) \\
763:  & & \\
764:  & = & \deg\left\{\displaystyle\frac{\mbox{td}(\bt_C)}{(1-\varepsilon^{-a} e^{-x_1})(1-\varepsilon^{-(r-a)} e^{-x_2})}\right\}_1\\
765:  & & \\
766:  & = & \frac{1-g_i}{(1-\varepsilon^{-a_j})(1-\varepsilon^{a_j})}
767: -\frac{\varepsilon^{-a_j}}{(1-\varepsilon^{-a_j})^2(1-\varepsilon^{a_j})}\deg_{C_i} x_1 
768: -\frac{\varepsilon^{a_j}}{(1-\varepsilon^{-a_j})(1-\varepsilon^{a_j})^2}\deg_{C_i} x_2. 
769: \end{array}$$
770: 
771: Since we assumed that $a_i,a_j,s$ have no common divisor for all $i,j=1,2,3$, 
772: we can split the sum over $\{\varepsilon\in\mu_s\}$ into four 
773: subsumes over (possibly empty) disjoint sets 
774: $$\begin{array}{l}
775: \{\varepsilon\in\mu_s \mbox{ s.t. }\varepsilon^{a_i}\neq 1\ \forall i=1,2,3\}, \\
776: \\
777: \{\varepsilon\in\mu_s \mbox{ s.t. }\varepsilon^{a_1}=1\}=\{\varepsilon\in\mu_{\alpha_1}\},\\
778: \\
779: \{\varepsilon\in\mu_s \mbox{ s.t. }\varepsilon^{a_2}=1\}=\{\varepsilon\in\mu_{\alpha_2}\} \mbox{ and}\\
780: \\
781: \{\varepsilon\in\mu_s \mbox{ s.t. }\varepsilon^{a_3}=1\}=\{\varepsilon\in\mu_{\alpha_3}\}.
782: \end{array}$$
783: This rewrites Formula~(\ref{eqypg}) into
784: $$\begin{array}{l}
785: \chi(X,\bl_{n})= \chi(\bigo_X) +\displaystyle\frac{1}{s}\sum_{\substack{\ep\in\mu_s \\ \ep^{a_i}\ne1\,\forall i = 1,2,3}} 
786: \frac{\varepsilon^{-n}-1}{(1-\varepsilon^{-a_1})(1-\varepsilon^{-a_2})(1-\varepsilon^{-a_3})} + \\
787: \\
788: \displaystyle\frac{1}{s}\sum_{(i,j)}\,\sum_{\scriptscriptstyle\varepsilon\in\mu_{\alpha_i}}(\varepsilon^{-n}-1)
789: \left( \frac{1-g_i}{(1-\varepsilon^{-a_j})(1-\varepsilon^{a_j})}
790: -\frac{\varepsilon^{-a_j}}{(1-\varepsilon^{-a_j})^2(1-\varepsilon^{a_j})}\deg_{C_i} x_1-\right. \\
791: \hfill \left. \displaystyle\frac{\varepsilon^{a_j}}{(1-\varepsilon^{-a_j})(1-\varepsilon^{a_j})^2}\deg_{C_i} x_2 \right),
792: \end{array}$$
793: where $(i,j)\in\{(1,2),\ (2,3), \ (3,1)\}$.
794: 
795: Since $c_1(\bn_{C_i/Z})=K_{C_i}-K_Z|_{C_i},$
796: we get $\deg_{C_i} x_1+\deg_{C_i} x_2=2(g_i-1)-\deg K_Z|_{C_i}.$
797: Then
798: $$\begin{array}{rcl}
799: \chi(X,\bl_{n}) & = &\chi(\bigo_X) + {\displaystyle\frac{1}{s}\sum_{\substack{\ep\in\mu_s \\ \ep^{a_i}\ne1\,\forall i = 1,2,3}}
800: \frac{\varepsilon^{-n}-1}{(1-\varepsilon^{-a_1})(1-\varepsilon^{-a_2})(1-\varepsilon^{-a_3})}}+\\
801: \\
802: &&\displaystyle \sum_{i=1}^3 \frac{\ol{nk_i}(\alpha_i-\ol{nk_i})}{4s}\deg K_Z|_{C_i} + \\
803: \\
804: &&\displaystyle\frac{1}{12s}\ol{nk_i}(\alpha_i-\ol{nk_i})(\alpha_i-2\cdot\ol{nk_i})(\deg_{C_i} x_1-\deg_{C_i} x_2),
805: \end{array}$$
806: where $\overline{\phantom{\Sigma}}$ denotes the smallest residue mod $\alpha_i$
807: and $k_i$ are defined by $k_ia_j=1$ mod $\alpha_i$. for $(i,j)$ as above.
808: 
809: We can think of $\chi(\bigo_X)$ as the RR-type expression for the pair $(X,L_{n})$, where 
810: $L_{n}=0\in \mbox{ Div}X\otimes\bq$ is the $\bq-$divisor corresponding to $\bl_{n}$. 
811: The rest of the formula is a sum of contributions coming from 
812: \begin{itemize}
813: \item dissident points of type $\frac{1}{s}(a_1,a_2,a_3),$ and
814: \item dissident curves of type $\frac{1}{\alpha_i}(a_j,-a_j),$
815: \end{itemize}
816: which are the singularities of~$(X,L_{n})$.
817: 
818: \begin{remark}\rm The attentive reader will notice that the argument in the last 
819: two sections is completely different: for $A_r$ curves, we use resolution of 
820: singularities and computations on ruled surfaces, whereas for (isolated or 
821: dissident) singular points, we use an auxiliary cover and equivariant 
822: Riemann--Roch. At least for curves without dissident points, the auxiliary cover 
823: construction also works, though with a twist. For an $A_r$ curve, a cyclic cover 
824: may not exist, but a $\mu_r\oplus\mu_r$ cover necessarily does and the argument 
825: goes through. The invariant $N_C$ of $X$ (the only term not explicitly defined in 
826: Theorem~\ref{thmdis}) depends only on the first order neighbourhood of the curve 
827: $C$ and can be expressed explicitly in terms of the genus C, and the singularity 
828: type and splitting of the normal bundle of the curve $\tilde C$ over $C$ 
829: in the auxiliary cover. For details, consult~\cite{anita}. 
830: \end{remark} 
831: 
832: \subsection{Conclusion of the proof} 
833: 
834: \begin{prfth}~\ref{thmdis}\hspace{5mm}
835: By Corollary~\ref{rrformm}, we only need to add the 
836: contributions from the dissident and isolated  
837: singular points to the RR formula. These were computed above. 
838: 
839: Let $C$ be a $_{\ol{mk}}\bigl(\frac{1}{r}(1,-1)\bigr)$ curve for $\bigo_X(mD)$ 
840: with dissident points
841: $$\{P_{\lambda}\ |\ \lambda\in\Lambda\} \mbox{ of types }\, _{\ol{mn_{\lambda}}}
842: \Bigl(\frac{1}{s_{\lambda} }(a_{1\lambda},a_{2\lambda},a_{3\lambda})\Bigr).$$ 
843: Note that for every $\lambda$ there exists $i$ such that 
844: $\hcf(a_{i\lambda},s_{\lambda})=r$, since $P_{\lambda}\in C$. Moreover,
845: $mn_{\lambda}a_{j\lambda}^{-1}=mk$ mod $r$. 
846: 
847: The contribution to RR from this locus is a sum of 
848: $$ \frac{\ol{mk}(r-\ol{mk})(r-2\cdot \ol{mk})}{6r^2}\left( (r-2)(1-g)+
849: \textstyle\frac{r-2}{2}K_Y\,^C\!E_1\,^C\!E_2+\,^C\!E_{r-2}^2\,^C\!E_{r-1}\right)$$
850: by Corollary~\ref{rrformm}, and of  
851: $$\begin{array}{l}
852: \frac{1}{12s_{\lambda}}\ol{mk}(r-\ol{mk})(r-2\cdot\ol{mk})(\deg_{C_{i\lambda}} x_1-\deg_{C_{i\lambda}} x_2)\\
853: \\
854: \mbox{}+\displaystyle\frac{1}{s_{\lambda}}\sum_{\substack{\ep\in\mu_{s\lambda} \\ \ep^{a_{i\lambda}}\ne1\,\forall i = 1,2,3}} \frac{\varepsilon^{-mn_{\lambda}}-1}{(1-\varepsilon^{-a_{1\lambda}})(1-\varepsilon^{-a_{2\lambda}})(1-\varepsilon^{-a_{3\lambda}})}.
855: \end{array}$$
856: for every point $P_{\lambda}$.
857: The first two rows of the sum can be written together as  
858: $$ \frac{\ol{mk}(r-\ol{mk})(r-2\cdot \ol{mk})}{12 r^2\tau_C}N_C,$$
859: where $\tau_C=\frac{1}{r}\lcm\{s_{\lambda}\}$ is the index of $C$ 
860: and the integer $N_C$ is an invariant of $X$ in the neighbourhood of $C$. 
861: \end{prfth}
862: \ethrm
863: 
864: \section{\cy threefolds}
865: \label{caya}
866: 
867: \subsection{The Hilbert series} 
868: 
869: When $X$ is a \cy threefold, the RR formula assumes a much more compact form.
870: 
871: \begin{corollary}
872: \label{thmcordis}
873: Let $X$ be a \cy threefold and let $D$ be a $\bq$-Cartier divisor.
874: Assume that $(X,D)$ is quasi-smooth and well formed with the 
875: following singularities: 
876: \begin{itemize}
877: \item points $Q$ of type $_n\bigl(\frac{1}{s}(a_1,a_2,a_3)\bigr),$
878: \item curves $C$ of generic type $_k\bigl(\frac{1}{r}(1,-1)\bigr)$ with index $\tau_C$. 
879: \end{itemize}
880: Then for all positive integers $m$,
881: $$\begin{array}{rcl}
882: h^0(X,\bigo_X(mD))&= & \displaystyle\frac{1}{6}m^3D^3+m\frac{D\cdot c_2(X)}{12}  \\
883:  & &\\
884:  & &\mbox{}+\sum_{Q}c_Q(mD)+\sum_{C}s_C(mD),
885: \end{array}$$ 
886: where
887: $$c_Q(mD)=
888: \frac{1}{s}\sum_{\substack{\ep\in\mu_s \\ \ep^{a_i}\ne1\,\forall i = 1,2,3}}
889: \frac{\varepsilon^{-nm}-1}{(1-\varepsilon^{-a_1})(1-\varepsilon^{-a_2})(1-\varepsilon^{-a_3})}$$
890: and
891: $$ s_C(mD)= -m\frac{\ol{mk}(r-\ol{mk})}{2r}\deg D|_C + \frac{\ol{mk}(r-\ol{mk})(r-2\cdot\ol{mk})}{12r^2\tau_C}N_C,$$
892: where $\overline{\phantom{\Sigma}}$ denotes the smallest residue mod $r$ and the 
893: integer $N_C$ is an invariant of $X$ in the neighbourhood of $C.$
894: \end{corollary}
895: 
896: \begin{prf}
897: This is a direct corollary of Theorem~\ref{thmdis}. By Kodaira vanishing, the
898: higher cohomologies of $\bigo_X(mD)$ vanish for $m>0$, and $\chi(\bigo_X)=0$ 
899: by Serre duality. 
900: \end{prf}
901: \ethrm
902: 
903: \begin{remark}\rm
904: Observe that for a $\frac{1}{2}(1,1)$ curve of singularities, 
905: the term involving $N_C$ vanishes. The Riemann--Roch formula for this special case 
906: was already proved in~\cite{balazs2}.
907: \end{remark}
908: 
909: \begin{corollary}
910: \label{cyrhil}
911: If $(X,D)$ satisfies the conditions of Corollary~\ref{thmcordis}, then the 
912: Hilbert series 
913: $$P_X(t)=1+\sum_{m=1}^{\infty}h^0(X,\bigo_X(mD))t^m.$$
914: can be written in the compact form
915: $$\begin{array}{rcl}
916: P_X(t)&=&
917: 1+\displaystyle\frac{D^3}{6}\cdot \frac{t^3+4t^2+t}{(1-t)^4}+\frac{c_2(X)\cdot D}{12}\cdot \frac{t}{(1-t)^2}\\
918:  && \\
919:  & &\mbox{}+\sum_Q \tilde{P}_Q(t)+\sum_C \tilde{P}_C(t), 
920: \end{array}$$
921: where every curve $C$ contributes 
922: $$\begin{array}{rcl}
923: \tilde{P}_C(t)&=&
924:   -\deg D|_C\left( \displaystyle\frac{1}{1-t^r}\sum_{i=1}^{r-1}i\frac{\ol{ik}(r-\ol{ik})}{2r}t^i
925: + \displaystyle\frac{rt^r}{(1-t^r)^2}\sum_{i=1}^{r-1}\frac{\ol{ik}(r-\ol{ik})}{2r}t^i \right)\\
926:  & &\\
927:  &&\mbox{}+ \displaystyle\frac{N_C}{12 r^2 \tau_c}\cdot \frac{1}{1-t^r}\sum_{i=1}^{r-1}\ol{ik}(r-\ol{ik})(r-2\cdot\ol{ik})t^i
928: \end{array}$$
929: and every singular point $Q$ contributes 
930: $$\tilde{P}_Q(t)=\frac{1}{1-t^s}\sum_{i=1}^{s-1}c_Q(iD)t^i.$$
931: Here 
932: $$c_Q(iD)=\frac{1}{s}\sum_{\substack{\ep\in\mu_s \\ \ep^{a_i}\ne1\,\forall i = 1,2,3}}
933: \frac{\varepsilon^{-ni}-1}
934: {(1-\varepsilon^{-a_1})(1-\varepsilon^{-a_2})(1-\varepsilon^{-a_3})}.$$ 
935: \end{corollary} 
936: 
937: \begin{prf} This follows directly from Corollary~\ref{thmcordis}, 
938: using elementary summation formulae of power series. 
939: \end{prf}
940: \ethrm
941: 
942: \subsection{Examples} 
943: \label{examples}
944: 
945: We apply Corollary~\ref{cyrhil} to the construction of new families of 
946: projective Calabi--Yau threefolds as follows. In the first step, we compute 
947: the Hilbert series $P(t)$ from Corollary~\ref{cyrhil} using the following 
948: input data:
949: \begin{itemize}
950: \item integers $h^0(X,D)$ and $h^0(X,2D);$
951: \item points $\left\{{}_n\bigl(\frac{1}{s}(a_1,a_2,a_3)\bigr)\right\};$
952: \item curves $\left\{{}_k\bigl(\frac{1}{r}(1,-1)\bigr)\mbox{ of degree } \deg D|_C, \mbox{ index } \tau_C \mbox{ and invariant } N_C \right\}.$
953: \end{itemize}
954: In the next step, we look for a set of weights $w_0,\ldots,w_n$ such that
955: $$Q(t)=P(t)\prod_{k=0}^{n}(1-t^{w_k})$$
956: is a polynomial. Then a plausible guess is that $(X,D)$ can be embedded in the 
957: ambient space $\bp(w_0,\ldots,w_n)$, and the shape of $Q(t)$ will suggest 
958: a set of generators and relations for the defining ideal. 
959: Compare~\cite{altinok, abr} for the philosophy and some explicit examples 
960: of this type of argument. As discussed in~\cite{abr} in great detail, 
961: this procedure is best done by a computer; we thank Gavin Brown for providing 
962: a computer program written in the programming language of the Magma computer 
963: algebra system~\cite{magma}. 
964: 
965: Lists of complete intersection \cy threefolds in weight\-ed projective spaces 
966: can certainly be generated in this way; such Calabi--Yau manifolds have been
967: listed by direct methods and extensively studied in the literature, for example 
968: in~\cite{cand1, cand2, cikreuzer}. Our computer search also generates more 
969: interesting examples in higher codimensions. 
970: 
971: \begin{example}
972: \label{cod3ex4wg} \rm Take the input data 
973: \begin{itemize}
974: \item $h^0(X,D)=3$ and $h^0(X,2D)=6$; 
975: \item points $\left\{{}_2\bigl(\frac{1}{3}(1,1,1)\bigr),{}_8\bigl(\frac{1}{9}(1,3,5)\bigr)\right\}$;
976: \item curve $\frac{1}{3}(1,2) \mbox{ with } \deg D|_C=\frac{1}{9},\ \tau_C=3,\ N_C=22$.
977: \end{itemize}
978: The Hilbert series is
979: \[ P(t)=\frac{Q(t)}{(1-t)^3(1-t^3)^2(1-t^5)(1-t^9)},\]
980: where
981: $$Q(t)=-t^{23}+t^{17}+t^{15}+2t^{13}+t^{11}-t^{12}-2t^{10}-t^8-t^6+1.$$
982: This suggests that $(X,D)$ could be realized as a codimension 3 threefold
983: in the weighted projective space $\bp^6(1^3,3^2,5,9)$. 
984: Indeed, define~$X\subset\bp^6(1^3,3^2,5,9)$ by the vanishing of 
985: the submaximal Pfaffians 
986: \begin{eqnarray*}
987: \label{pfeqp}
988:  & &\pf_1=b_7y_1+b_5 z+x_2 v,   \\
989:  & &\pf_2=a_7y_1+a_5 z+x_1 v,   \\
990:  & &\pf_3=y_2 y_1-a_5x_2+b_5x_1,  \\
991:  & &\pf_4=-y_2 z-a_7x_2+b_7x_1,  \\
992:  & &\pf_5=y_2 v-a_7b_5+a_5b_7, 
993: \end{eqnarray*}
994: of the $5\times 5$ skew-symmetric matrix 
995: \[ M = \left(\begin{matrix}  0 & y_2 & a_7 & a_5 & x_1 \\ 
996:                              & 0 & b_7 & b_5 & x_2 \\ 
997:                               & & 0 & v & -z \\
998:                                & && 0 & y_1 \\ 
999:                                 & & & & 0
1000:              \end{matrix}\right).\]
1001: Here $x_1,x_2,x_3,y_1,y_2,z, v$ are weighted variables of $\bp^6$ and $a_i, b_i$ 
1002: are general polynomials of degree~$i$. It is easy to check that $X$ is 
1003: quasi-smooth and well formed, and that its singular locus indeed consists of an 
1004: isolated singular point of type $\frac{1}{3}(1,1,1)$ 
1005: and an $A_2$ curve with a dissident point of type $\frac{1}{9}(1,3,5)$.
1006: 
1007: >From this representation, one can also prove that $X$ is a general quasilinear 
1008: section 
1009: $$X=(5)^2 \cap (7)^2\cap \mathcal{C}\!\wgr(2,5)\subset \bp(1^3,3^2,5^3,7^2,9).$$
1010: Here $\mathcal{C}\!\wgr(2,5)$ denotes the projective cone over $\wgr(2,5)$, 
1011: the weighted Grassmannian as defined in \cite{wg}.
1012: \end{example}
1013: 
1014: \begin{example}
1015: \label{tocod4exa}\rm Take the input data 
1016: \begin{itemize}
1017: \item $h^0(X,D)=2$ and $h^0(X,2D)=4$;
1018: \item point $_4\bigl(\frac{1}{5}(1,1,3)\bigr)$;
1019: \item curve $\frac{1}{3}(1,2) \mbox{ with } \deg D|_C=1,\ \tau_C=1,\ N_C=12$.
1020: \end{itemize}
1021: The computer output for $P(t)$ is
1022: \[ P(t)=\frac{t^{21}-3t^{15}-3t^{14}-3t^{13}+2t^{12}+6t^{11}+6t^{10}+2t^9-3t^8-3t^7-3t^6+1}{(1-t)^2(1-t^2)(1-t^3)^4(1-t^5)}.\]
1023: This suggests that $(X,D)$ could be realized as a codimension 4 threefold in the weighted 
1024: projective space $\bp^7(1^2,2,3^4,5)$, defined by nine relations of degrees $6,6,6$, $7,7,7$, 
1025: $8,8,8$ respectively. The existence of $X$ can be proved by the Type I 
1026: unprojection \cite{altinok, stavmile} starting from a codimension 3 variety 
1027: $Y \subset \bp^6(1,1,2,3,3,3,3)$ defined by a set of Pfaffians as in Example~\ref{cod3ex4wg}.
1028: We omit the details.
1029: \end{example}
1030: 
1031: \begin{example}
1032: \label{ex3wg} \rm Take the input data 
1033: \begin{itemize}
1034: \item $h^0(X,D)=2$ and $h^0(X,2D)=7$;
1035: \item point ${}_3\bigl(\frac{1}{4}(2,3,3)\bigr)$;
1036: \item curve $\frac{1}{2}(1,1) \mbox{ with } \deg D|_C=\frac{7}{4},\ \tau_C=2.$
1037: \end{itemize}
1038: The closed form of the Hilbert series is
1039: \[ \frac{1-3t^4-4t^5-t^6+6t^7+6t^8+2t^9-2t^{11}-6t^{12}-6t^{13}+t^{14}+4t^{15}+3t^{16}-t^{20}}{(1-t)^2(1-t^2)^4(1-t^3)^2(1-t^4)}.\]
1040: This suggests that $(X,D)$ could be realized as a codimension 5 threefold in the weighted 
1041: projective space $\bp^8(1^2,2^4,3^2,4)$. This turns out to be true indeed, although 
1042: the details are somewhat tedious. More conceptually, following~\cite{wg}, one can show that 
1043: $(X,D)$ can be realized as a general quasilinear section 
1044: $$X=\wogr(5,10) \cap(2)^2\cap (3)^4 \cap (4)\subset \bp(1^2,2^6,3^6,4^2), $$
1045: where $\wogr(5,10)$ denotes a weighted orthogonal Grassmannian~\cite{wg}.
1046: \end{example}
1047: 
1048: \begin{remark} \rm Two questions arise at this point: to what extent are these 
1049: Calabi--Yau threefold families ``new'', and how many families can one construct 
1050: this way. The second question will be discussed elsewhere; to give away the 
1051: (negative) punchline, however hard we try, so far we have been unable 
1052: to construct infinitely many families. In answer to the first question, 
1053: it is certainly possible that our examples are birational to one of the 
1054: gigantic number of toric complete intersection 
1055: Calabi--Yau threefolds~\cite{baty, bb, torickreuzer}. However, 
1056: our projective descriptions, in relatively simple ambient varieties 
1057: such as weighted Grassmannians~\cite{wg} and the 
1058: ``universal Type I unprojection''~\cite{stavmile}, are new. Such descriptions 
1059: are quite pretty in themselves, and can also be used for various purposes 
1060: such as computing their Hodge numbers and studying mirror symmetry 
1061: phenomena (details to follow). For these reasons alone, we believe 
1062: they deserve some attention. 
1063: \end{remark} 
1064: \vspace{0.1in}
1065: 
1066: \paragraph{Acknowledgements} We thank Miles Reid for his continued attention 
1067: and helpful advice during the course of this project, Gavin Brown for providing 
1068: us with the Magma computer program~\cite{magma} used to find the examples 
1069: presented in Section~\ref{examples}, and the referee for a very careful reading
1070: of our manuscript. We also thank the Isaac Newton Institute, Cambridge 
1071: for hospitality while part of this research was conducted, and the 
1072: Mathematics Institute of the University of Warwick for supporting the 
1073: research of A. B. with a Special Research Studentship. 
1074: 
1075: \begin{thebibliography}{99} 
1076: \addcontentsline{toc}{section}{References}
1077: {\small
1078: \bibitem{altinok} S. Alt\i nok, {\it Graded rings corresponding to polarised K3 surfaces and $\bq$-Fano 3-folds}, PhD. Thesis, University of Warwick, 1998.
1079: \bibitem{abr} S. Alt\i nok, G. Brown and M. Reid, {\it Fano 3-folds, $K3$ surfaces and graded rings}, in: Topology and geometry: commemorating SISTAG (ed. A. J. Berrick et al), 25--53, Contemp. Math., 314, Amer. Math. Soc., Providence, RI, 2002.
1080: \bibitem{as} M. F. Atiyah and G. B. Segal, {\it The index of elliptic operators. II}, Ann. of Math. {\bf 87} (1968) 531--545.
1081: \bibitem{atiyah} {M. F.Atiyah and I. M. Singer}, {\it The index of elliptic operators. III}, Ann. of Math. {\bf 87} (1968) 564--604.
1082: \bibitem{baty} V. V. Batyrev, {\it Dual polyhedra and mirror symmetry for Calabi-Yau hypersurfaces in toric varieties}  J. Algebraic Geom. {\bf 3} (1994) 493--535. 
1083: \bibitem{bb} L. A. Borisov, {\it Towards the mirror symmetry for Calabi-Yau complete intersections in Gorenstein toric Fano varieties}, alg-geom/9310001.
1084: \bibitem{magma} W. Bosma, J. Cannon and C.~Playoust, {\it The Magma algebra system I: The user language}, J. Symb. Comp. {\bf 24} (1997) 235--265. \newline See also http://www.maths.\linebreak[2]usyd.edu.au:8000/u/magma
1085: \bibitem{anita} A. Buckley, {\it Orbifold Riemann--Roch for threefolds and applications to Calabi--Yaus}, PhD. Thesis, University of Warwick, 2003.
1086: \bibitem{cand1} P. Candelas, A. M. Dale, C. A. L\"utken and  R. C. Schimmrigk, {\it Complete intersection Calabi-Yau manifolds}, Frontiers of high energy physics (London, 1986, ed. I. G. Halliday), 88--134, Hilger, Bristol, 1987. 
1087: \bibitem{cand2} P. Candelas, M. Lynker and R. Schimmrigk, {\it Calabi-Yau manifolds in weighted ${\mathbb P}^4$}, Nuclear Phys. B {\bf 341} (1990) 383--402.
1088: \bibitem{wg}  {A. Corti} and {M. Reid}, {\it Weighted Grassmannians}, Algebraic geometry. A volume in memory of Paolo Francia (ed. M. C. Beltrametti et al), 141--163, de Gruyter, Berlin, {2002}. 
1089: \bibitem{fletcher1} A. R. Fletcher, {\it Contributions to Riemann-Roch on projective $3$-folds with only canonical singularities and applications} Algebraic geometry, Bowdoin, 1985 (Brunswick, Maine, 1985, ed. S. J. Bloch), 221--231, Proc. Sympos. Pure Math., 46, Part 1, Amer. Math. Soc., Providence, RI, 1987.
1090: \bibitem{fletcher} {A. R. Fletcher}, {\it Working with weighted complete intersections}, Explicit birational geometry of 3--folds (ed. A. Corti and M. Reid), 101--173, London Math. Soc. Lecture Note Ser., 281, Cambridge Univ. Press, Cambridge, {2000}. 
1091: \bibitem{morrison} {K. Intriligator, D. R. Morrison} and {N. Seiberg}, {\it Five--dimensional supersymmetric gauge theories and degenerations of Calabi--Yau spaces}, Nuclear Phys. B {\bf 497} (1997) 56--100.
1092: \bibitem{kawasaki} Kawasaki, T. {\it The Riemann-Roch theorem for complex $V$-manifolds}, Osaka J. Math.  {\bf 16}  (1979) 151--159.
1093: \bibitem{torickreuzer} M. Kreuzer and H. Skarke, {\it Complete classification of reflexive polyhedra in four dimensions}, Adv. Theor. Math. Phys. {\bf 4} (2002) 1209-1230.
1094: \bibitem{cikreuzer} {M. Kreuzer, E. Riegler} and {D. A. Sahakyan}, {\it Toric complete intersections and weighted projective space}, J. Geom. Phys. {\bf 46} (2003) 159--173.
1095: \bibitem{stavmile} {S. Papadakis} and {M. Reid}, {\it Kustin--Miller unprojection without complexes}, J. Algebraic Geom. {\bf 13} (2004) 563--577.
1096: \bibitem{3-folds}  {M. Reid}, {\it Canonical 3--folds}, Journ\'{e}es de g\'{e}om\'{e}trie alg\'{e}brique d'Angers (ed. A. Beauville), 1979, 273--310.
1097: %\bibitem{pagoda}  {M. Reid}, {\it Minimal models of canonical 3--folds}, Algebraic Varieties and Analytic Varieties (ed. S. Iitaka), Advanced Studies in Pure Mathematics {\bf 1} (1983) 131--180. 
1098: \bibitem{ypg}  {M. Reid}, {\it Young person's guide to canonical singularities},  Algebraic geometry (Bowdoin 1985, ed. S. Bloch), 345--414, Proc. Sympos. Pure Math. 46, Part 1, AMS, Providence, RI, 1987.
1099: \bibitem{balazs2} {B. Szendr\H oi}, {\it Calabi--Yau threefolds with a curve of singularities and counter\-examples to the Torelli problem II}, Math. Proc. Cambridge Philos. Soc. {\bf 129} (2000) 193--204.
1100: \bibitem{toen} {B. Toen}, {\it Th\'eor\'emes de Riemann-Roch pour les champs de Deligne-Mumford}, $K$-Theory {\bf 18} (1999) 33--76.
1101: }
1102: \end{thebibliography}
1103: 
1104: \noindent {\small \sc Department of Mathematics, University of Ljubljana
1105: 
1106: \noindent Jadranska 19, Ljubljana 1000, Slovenia 
1107: 
1108: \noindent E-mail address: {\tt anita.buckley@fmf.uni-lj.si}
1109: 
1110: \vspace{0.1in}
1111: 
1112: \noindent {\rm and}
1113: 
1114: \vspace{0.1in}
1115: 
1116: \noindent Department of Mathematics, Utrecht University
1117: 
1118: \noindent PO. Box 80010, NL-3508 TA Utrecht, The Netherlands
1119: 
1120: \noindent E-mail address: \tt szendroi@math.uu.nl}
1121: 
1122: \end{document}