0806.3230/A.tex
1: %
2: % Linear precision for toric surface patches
3: %
4: % Kristian Ranestad
5: % Frank Sottile
6: % Hans-Christian Graf von Bothmer
7: %
8: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
9: \documentclass[12pt,a4paper]{amsart}
10: %\documentclass[12pt]{amsart}
11: \usepackage{amssymb,amsmath}
12: \usepackage{graphicx}
13: \usepackage{colordvi}
14: \usepackage[all]{xy}
15: 
16: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
17: % US letter
18: %\headheight=8pt     \topmargin=20pt
19: %\textheight=620pt   \textwidth=465pt
20: %\oddsidemargin=1pt \evensidemargin=1pt
21: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
22: %    A4 paper
23: \headheight=8pt     \topmargin=10pt
24: \textheight=660pt     \textwidth=440pt
25: \oddsidemargin=-1pt   \evensidemargin=-1pt
26: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
27: 
28: 
29: \newtheorem{IntroTheorem}{Theorem}
30: \newtheorem{IntroCor}[IntroTheorem]{Corollary}
31: 
32: \numberwithin{equation}{section}
33: \newtheorem{theorem}[equation]{Theorem}
34: \newtheorem{lemma}[equation]{Lemma}
35: \newtheorem{proposition}[equation]{Proposition}
36: \newtheorem{cor}[equation]{Corollary}
37: \newtheorem{question}[equation]{Question}
38: 
39: \theoremstyle{definition}
40: \newtheorem{definition}[equation]{Definition}
41: 
42: %\theoremstyle{example}
43: \newtheorem{example}[equation]{Example}
44: 
45: %\theoremstyle{remark}
46: \newtheorem{remark}[equation]{Remark}
47: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
48: %\def\publname{\Red{{\scriptsize 19 June 2009}} \Blue{{\scriptsize\tt ArXiv.org:0806.3230}}}
49: %\def\currentvolume{}
50: %\def\currentissue{}
51: %\PII{}
52: %\copyrightinfo{}{}
53: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
54: \newcounter{FNC}[page]
55: \def\fauxfootnote#1{{\addtocounter{FNC}{2}$^\fnsymbol{FNC}$%
56:      \let\thefootnote\relax\footnotetext{$^\fnsymbol{FNC}$#1}}}
57: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
58: \hyphenation{pa-ram-e-trized}
59: 
60: \newcommand{\rat}{\ {\relbar\rightarrow}\ }
61: \newcommand{\longrat}{{\;\relbar\relbar\rightarrow\;}}
62: 
63: \newcommand{\ba}{{\bf a}}
64: \newcommand{\bb}{{\bf b}}
65: \newcommand{\bv}{{\bf v}}
66: 
67: \newcommand{\calA}{{\mathcal A}}
68: 
69: \newcommand{\C}{{\mathbb C}}
70: \renewcommand{\P}{{\mathbb P}}
71: \newcommand{\AZ}{{\mathbb A}}
72: \newcommand{\Q}{{\mathbb Q}}
73: \newcommand{\R}{{\mathbb R}}
74: \newcommand{\Z}{{\mathbb Z}}
75: 
76: \newcommand{\sumdelta}{\sum \delta}
77: 
78: \newcommand{\QED}{\hfill\raisebox{-1pt}{\includegraphics{figures/QED.eps}}\vspace{8pt}}
79: \newcommand{\eqed}{\hfill\raisebox{-1pt}{\includegraphics{figures/QED.eps}}}
80: \renewcommand{\qed}{\hfill\QED}
81: 
82: %\renewcommand{\Blue}[1]{\Cyan{#1}}
83: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
84: \begin{document}
85: 
86: \title{Linear precision for toric surface patches}
87: 
88: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
89: \author[H-C. Graf v. Bothmer]{Hans-Christian Graf von Bothmer}
90: \address{Mathematisches Institut, Georg-August-Universiti\"at G\"ttingen, 
91:    Bunsenstr. 3-5, D-37073 G\"ottingen Germany}
92: \email{bothmer@uni-math.gwdg.de}
93: \urladdr{http://www.crcg.de/wiki/index.php5?title=User:Bothmer}
94: 
95: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
96: \author[K.~Ranestad]{Kristian Ranestad}
97: \address{Matematisk institutt\\
98:          Universitetet i Oslo\\
99:          PO Box 1053, Blindern\\
100:          NO-0316 Oslo\\
101:          Norway}
102: \email{ranestad@math.uio.no}
103: \urladdr{http://www.math.uio.no/\~{}ranestad}
104: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
105: \author[F.~Sottile]{Frank Sottile}
106: \address{Department of Mathematics\\
107:          Texas A\&M University\\
108:          College Station\\
109:          TX \ 77843-3368\\
110:          USA}
111: \email{sottile@math.tamu.edu}
112: \urladdr{http://www.math.tamu.edu/\~{}sottile}
113: \thanks{Work of Sottile supported by NSF grants CAREER DMS-0538734 and DMS-0701050,  
114:         the Institute for Mathematics and its Applications, and Texas Advanced Research
115:    Program under Grant No. 010366-0054-2007} 
116: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
117: \subjclass[2000]{14M25, 65D17}
118: \keywords{B\'ezier patches, geometric modeling, linear precision, Cremona transformation, toric
119: patch}
120: 
121: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
122: \begin{abstract}
123:  We classify the homogeneous polynomials in three variables whose toric polar linear
124:  system defines a Cremona transformation.
125:  This classification includes, as a proper subset, the classification of toric
126:  surface patches from 
127:  geometric modeling which have linear precision.
128:  Besides the well-known tensor product patches and B\'ezier triangles, we identify a
129:  family of toric patches with trapezoidal shape, each of which has linear precision.
130:  Furthermore, B\'ezier triangles and tensor product patches are special cases of
131:  trapezoidal patches.
132:  
133: \end{abstract}
134: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
135: \maketitle
136: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
137: 
138: \begin{center}
139: Communicated by Wolfgang Dahmen and Herbert Edelsbrunner
140: \end{center}
141: 
142: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
143: %
144: \section*{Introduction} 
145: %
146: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
147: 
148: While the basic units in the geometric modeling of surfaces are B\'ezier triangles and
149: rectangular tensor product patches, some applications call for multi-sided $C^\infty$
150: patches (see~\cite{KK00} for a discussion).
151: Krasauskas's toric B\'ezier patches~\cite{KR02} are a flexible and mathematically
152: appealing system of such patches.
153: These are based on real toric varieties from algebraic geometry, may have shape any
154: polytope $\Delta$ with integer vertices, and they include the classical B\'ezier patches as special
155: cases. 
156: For descriptions of multisided patches and toric patches, see~\cite{Gold}.
157: 
158: More precisely, given a set of lattice points in $\Z^n$ with convex hull $\Delta$,
159: Krasauskas defined toric B\'ezier functions, which are polynomial
160: blending functions associated to each lattice point. 
161: This collection of lattice points and toric B\'ezier functions, together with a positive
162: weight associated to each lattice point is a \Blue{{\sl toric patch}}.
163: Choosing also a control point in $\R^d$ for each lattice point leads to a map
164: $\Phi \colon \Delta \to \R^d$ whose image may be used in modeling.
165: If we choose the lattice points themselves as control points we obtain the 
166: \Blue{{\sl tautological map}} $\tau \colon \Delta \to \Delta$, which  
167: is a bijection. 
168: If the tautological map has a rational inverse, 
169: then the toric patch has \Blue{{\sl linear precision}}. 
170: 
171: 
172: The lattice points and weights of a toric patch are encoded in
173: a homogeneous multivariate polynomial $F(x_0,\dotsc,x_n)$ with positive coefficients,
174: with every such polynomial corresponding to a toric patch.
175: In~\cite{GS} it was shown that the toric patch given by $F$ has linear precision
176: if and only if the associated toric polar linear system,
177: %
178:  \[
179:     T(F)\ =\  \Bigl\langle\; x_0\frac{\partial F}{\partial x_0},\,
180:            x_1\frac{\partial F}{\partial x_1},\,
181:            \dotsc,\,
182:            x_n\frac{\partial F}{\partial x_n}\;\Bigr\rangle\,,
183:  \]
184: %
185: defines a birational map $\Blue{\Phi_F}\colon\P^n \rat \P^n$.
186: This follows from the existence of a rational reparameterization transforming 
187: the tautological map into $\Phi_F$.
188: The polar linear system is toric because the derivations $x_i\frac{\partial}{\partial x_i}$
189: are vector fields on the torus $(\C^\times)^n\subset\P^n$.
190: 
191: When $T(F)$ defines a birational map, we say that $F$ defines a 
192: \Blue{{\sl toric polar Cremona transformation}}.
193: We seek to classify all such homogeneous polynomials $F$ without the
194: restriction that the coefficients are positive or even real.
195: This is a variant of the classification of homogeneous polynomials $F$ whose polar linear
196: system 
197: (which is generated by the partial derivatives $\frac{\partial F}{\partial x_i}$) 
198: defines a  birational map.
199: Dolgachev~\cite{Do00} classified all such square free polynomials in 3 variables
200: and those in 4 variables that are products of linear forms.\medskip
201: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
202: 
203: \noindent{\bf Definition.}
204:  Two polynomials $F$ and $G$ are called \Blue{{\sl equivalent}} if they can be transformed
205:  into each other by successive invertible monomial substitutions, multiplications with
206:  Laurent monomials, or scalings of the variables. \medskip
207: 
208: The property of defining a 
209: toric polar Cremona transformation is preserved under this equivalence.
210: Our main result is the classification (up to equivalence) of
211: homogeneous polynomials in three variables that 
212: define toric polar Cremona transformations.
213: 
214: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
215: \begin{IntroTheorem}\label{T:Classification}
216:  A homogeneous polynomial $F$ in three variables that defines a toric polar
217:  Cremona transformation is equivalent to one of the following
218:  \begin{enumerate}
219:   \item  $(x+z)^a(y+z)^b$ for $a,b\geq 1$,
220:   \item  $(x+z)^a\bigl( (x+z)^d + yz^{d-1}\bigr)^b$ for $a\geq 0$ and $b,d\geq 1$, or 
221:   \item  $\bigl(x^2+y^2+z^2-2(xy+xz+yz)\bigr)^d$, for $d\geq 1$.
222:  \end{enumerate}
223: %
224: \end{IntroTheorem}
225: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
226: 
227: When $a=0$ and $d=1$ in (2), we obtain the polynomial
228: $(x+y+z)^b$, which corresponds to a B\'ezier triangular patch of degree $b$ used in
229: geometric modeling.
230: Similarly, the polynomials $F$ in (1) correspond to tensor product patches, which are also  
231: common in geometric modeling.
232: These are also recovered from the polynomials in (2) when $d=0$, after multiplying by $z^b$.
233: Less-known in geometric modeling are \Blue{{\sl trapezoidal patches}}, which 
234: correspond to the polynomials of (2) for general parameters $a,b,d$.
235: Their blending functions and weights are given in Example~\ref{Ex:trapezoid}.
236: 
237: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
238: \begin{IntroCor}\label{T:Toric_patches}
239:   The only toric surface patches possessing linear precision are
240:   tensor product patches, B\'ezier triangles, and the trapezoidal patches of 
241:   Example~$\ref{Ex:trapezoid}$.
242: \end{IntroCor}
243: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
244: 
245: The polynomials of Theorem~\ref{T:Classification}(3) cannot arise in geometric
246: modeling, for they are not equivalent to a polynomial with positive coefficients.
247: 
248: We remark that the notion of linear precision used here and
249: in~\cite{GS} is more restrictive than typically used in geometric
250: modeling.
251: There, linear precision often means that there are control points in $\Delta$
252: so that the resulting map $\Delta\to\Delta$ is the identity. 
253: We include these control points in our definition of a patch to give a 
254: precise definition that enables the mathematical study of linear precision.
255: Nevertheless, this restrictive classification will form the basis for a more
256: thourough study of the general notion of linear precision for toric
257: patches.
258: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
259: 
260: In Section~\ref{S:LPPP}, we review definitions and results from~\cite{GS}
261: about linear precision for toric patches, including Proposition~\ref{P:TPLS} which
262: asserts that a toric patch has linear precision if and only if a
263: polynomial associated to the patch defines a toric polar Cremona
264: transformation, showing that Corollary~\ref{T:Toric_patches} 
265: follows from Theorem~\ref{T:Classification}.
266: We also show directly that polynomials associated to B\'ezier
267: triangles, tensor product patches, and trapezoidal patches define
268: toric polar Cremona transformations.
269: In particular, this implies that trapezoidal patches have linear precision.
270: In Section~\ref{S:Basics}, we prove that the above equivalence preserves the
271: property of defining a toric polar Cremona transformation. 
272: %We also establish an important restriction lemma.
273: Then we give our proof of Theorem~\ref{T:Classification}.
274: Three important ingredients of this proof are established in the
275: remaining sections.
276: In Section~\ref{S:contracted}, we show that if all factors of $F$ are contracted, 
277: then $F$ has two such contracted factors and we identify them.
278: In Section~\ref{S:not_contracted}, we classify the non contracted factors of $F$,
279: and we conclude in  Section~\ref{S:technical} with an analysis of possible singularities
280: of the curve defined by $F$.
281: 
282: Most of our proofs use elementary notions from algebraic
283: geometry as developed in~\cite{CLO}.
284: The only exceptions are in Section~\ref{Sec:genus}, where we blow up
285: a binomial curve to compute its arithmetic genus, and
286: Section~~\ref{S:technical}, which uses the resolution of base points of
287: a linear series.\medskip
288: 
289: \noindent {\bf Notation.}  
290: We shall use the term \Blue{{\sl linear system}} on $\P^2$ both for a vector space 
291: of forms and for the projective space of curves that they define.  
292: More generally a linear system on a surface defines a rational map to a
293: projective space.  
294: A common factor in the linear system can be removed without changing this map, so we shall
295: say that two linear systems are equivalent if they define the same rational map.  
296: For example,
297: \[
298:   \langle F,G,H \rangle\ \equiv\ \langle xF, xG,xH \rangle\,.
299: \]
300: 
301: 
302: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
303: %
304: \section{Linear precision and toric surface patches}\label{S:LPPP}
305: %
306: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
307: 
308: We review some definitions and results of~\cite{GS}.
309: See~\cite{Cox03,KR02,So03} for more on toric varieties and their relation to geometric
310: modeling.
311: 
312: Let $\Delta\subset\R^n$ be a lattice polytope (the vertices of $\Delta$ lie
313: in the integer lattice $\Z^n$).
314: This may be defined by its facet inequalities
315: \[
316:   \Delta\ =\ \{s\in\R^n \,:\, h_i(s)\geq 0\,, i=1,\dotsc,N\}\,.
317: \]
318:  Here, $\Delta$ has $N$ facets and for each $i=1,\dotsc,N$, 
319:  $\Blue{h_i}(s):=\langle \bv_i,s\rangle + c_i$ is the linear function defining
320:  the $i$th facet, where $\bv_i\in\Z^n$ is the (inward oriented) primitive vector normal to 
321:  the facet and $c_i\in\Z$. 
322: 
323:  Let $\calA\subset\Delta\cap\Z^n$ be any subset of the integer points of
324:  $\Delta$ which includes its vertices.
325:  Let $w=\{w_\ba: \ba\in\calA\}\subset \R_>$ be a collection of
326:  positive weights.
327:  For each $\ba\in\calA$, Krasauskas defined the \Blue{{\sl toric B\'ezier function}}
328: %
329:  \begin{equation}\label{Eq:toric-Bezier}
330:    \Blue{\beta_\ba}(s)\ :=\
331:     w_\ba\cdot h_1(s)^{h_1(\ba)}h_2(s)^{h_2(\ba)}\dotsb h_N(s)^{h_N(\ba)}\,.
332:  \end{equation}
333: %
334: Then $(\beta_\ba(s) :  \ba\in\calA)$ are the
335: blending functions for the 
336: \Blue{{\sl toric B\'ezier patch of shape $(\calA,w)$}}.
337: 
338: 
339: Given \Blue{{\sl control points}}
340: $b=\{\bb_\ba :  \ba\in\calA\}\subset\R^m$ we may define the map 
341: %
342:  \begin{equation}\label{Eq:TBPwCP}
343:   \begin{array}{rcl}
344:     \Phi\ \colon\ \Delta &\longrightarrow& \R^m\,,\\
345:         s&\longmapsto&    
346:    {\displaystyle \frac{\sum_{\ba\in\calA} \bb_\ba \cdot \beta_\ba(s)}%
347:         {\sum_{\ba\in\calA} \beta_\ba(s)}}\ .
348:   \end{array}
349:  \end{equation}
350: %
351: Precomposing the function $\beta_\ba(s)$ with a homeomorphism $\psi\colon\Delta\to\Delta$
352: gives a new function $\beta_\ba(\psi(s))$.
353: Using these new functions in place of the original functions $\beta_\ba$
354: in~\eqref{Eq:TBPwCP} does not change the shape $\Phi(\Delta)$ but will alter the
355: parameterization of the patch. 
356: 
357: The toric B\'ezier patch of shape $(\calA,w)$ 
358: has \Blue{{\sl linear precision}} if the tautological map
359: %
360:  \begin{eqnarray}
361:     \tau\ \colon\ \Delta &\longrightarrow& \Delta\nonumber\\
362:         s&\longmapsto&   
363:    \frac{\sum_{\ba\in\calA} \ba \cdot \beta_\ba(s)}%
364:         {\sum_{\ba\in\calA} \beta_\ba(s)}\nonumber
365:  \end{eqnarray}
366: %
367: is the identity.
368: While this may not occur for the given blending functions, Theorem~1.11 in~\cite{GS}
369: asserts that there is a unique reparameterization by a homeomorphism
370: $\psi\colon\Delta\to\Delta$ so that $(\beta_\ba(\psi(s)) :  \ba\in\calA)$ has linear
371: precision. 
372: Unfortunately, these new functions $\beta_\ba(\psi(s))$ may not be easy
373: to compute.
374: The toric patch of shape $(\calA,w)$ has 
375: \Blue{{\sl rational linear precision}} if these new functions 
376: $\beta_\ba(\psi(s))$ are rational functions or polynomials.
377: This property has an appealing mathematical reformulation.
378: 
379: Given data $(\calA,w)$ as above, suppose that 
380: $d:=\max\{\Blue{|\ba|}:=a_1+\dotsb+a_n :  \ba\in\calA\}$ is
381: the maximum degree of a monomial $x^\ba$ for $\ba\in\calA$.
382: Define the homogeneous polynomial
383: \[
384:     \Blue{F_{\calA,w}}(x_0,x_1,\dotsc,x_n)\ :=\ 
385:     \sum_{\ba\in\calA} w_\ba\; x_0^{d-|\ba|} x^\ba\,.
386: \]
387: The \Blue{{\sl toric polar linear system}} of $F_{\calA,w}$
388: is the linear system generated by its toric derivatives,
389: %
390:  \begin{equation}\label{Eq:TPLS}
391:     \Blue{T(F_{\calA,w})}\ :=\ 
392:       \Bigl\langle\; x_0\frac{\partial F_{\calA,w}}{\partial x_0},\,
393:            x_1\frac{\partial F_{\calA,w}}{\partial x_1},\,
394:            \dotsc,\,
395:            x_n\frac{\partial F_{\calA,w}}{\partial x_n}\;\Bigr\rangle\,.
396:  \end{equation}
397: %
398: 
399: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
400: \begin{proposition}[Corollary 3.13 of~\cite{GS}]\label{P:TPLS}
401:   The toric patch of shape $(\calA,w)$ has rational linear precision if and only if
402:   its toric polar linear system~$\eqref{Eq:TPLS}$ defines a birational isomorphism
403:   $\P^n \rat \P^n$.
404: \end{proposition}
405: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
406: 
407: We illustrate Proposition~\ref{P:TPLS} through some examples of patches with
408: linear precision. 
409: 
410: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
411: \begin{example}[B\'ezier curves]\label{Ex:curves}
412:  Let $\calA:=\{0,1,\dotsc,d\}\subset\R$.
413:  If we choose weights $w_i:=\binom{d}{i}$, then the toric B\'ezier functions are
414: %
415:  \begin{equation}\label{Eq:Bez_Curve}
416:    \beta_i(s)\ :=\ \tbinom{d}{i} s^i (d-s)^{d-i}\,,
417:    \qquad i=0,1,\dotsc,d\,.
418:  \end{equation}
419: %
420:  The polynomial is
421: \[
422:    F_{\calA,w}\ =\ \sum_{i=0}^d \tbinom{d}{i} x^i y^{d-i}\ =\ (x+y)^d\,,
423: \]
424:  and its associated toric polar linear system is
425: \[
426:   T(F_{\calA,w})\ =\ \left\langle xd(x+y)^{d-1},\, yd(x+y)^{d-1}\right\rangle
427:   \ \equiv\ \langle x, y\rangle\,,
428: \]
429: which defines the identity map $\P^1\to\P^1$.
430: Thus the toric patch with shape $(\calA,w)$ has rational linear precision.
431: 
432: Substituting $s=d\cdot t$ and removing the common factor of $d^d$, the 
433: toric B\'ezier functions~\eqref{Eq:Bez_Curve} become the univariate Bernstein
434: polynomials, the blending functions for B\'ezier curves.
435: Up to a coordinate change, this is the only toric patch in dimension 1
436: which has rational linear precision~\cite[Example 3.15]{GS}.
437: More precisely, the toric polar linear system of a homogeneous polynomial $F(x,y)$ that is
438: prime to $xy$ defines a birational isomorphism $\P^1\to \P^1$ if and only if $F$ is the
439: pure power of a linear form that does not vanish at either coordinate point $[0:1]$ and
440: $[1:0]$.  
441: \eqed
442: \end{example}
443: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
444: We specialize to the case $n=2$ for the remainder of this paper.
445: Our homogeneous coordinates for $\P^2$ are $[x:y:z]$ and the toric polar linear system
446: defines the map
447: %
448:  \begin{equation}\label{Eq:TPCP}
449:    [\;x\ :\ y\ :\ z\;]\ \longmapsto\ 
450:    \Bigl[\; x\frac{\partial F_{\calA,w}}{\partial x}\ :\ 
451:             y\frac{\partial F_{\calA,w}}{\partial y}\ :\ 
452:             z\frac{\partial F_{\calA,w}}{\partial z}\;\Bigr]\,.
453:  \end{equation}
454: 
455: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
456: \begin{example}[Quadratic Cremona Transformation]\label{Ex:cremona}
457:  Before giving examples of toric surfaces patches with linear precision,
458:  we describe the classical quadratic Cremona transformation, a birational map
459:  on the projective plane.
460:  This is defined by
461: %
462:  \[
463:     \varphi\ \colon\   [\;x\ :\ y\ :\ z\;]\ \longmapsto\ 
464:    [\;yz\ :\ zx\ :\ xy\;]\,.
465:  \]
466: %
467:  At points where $xyz\neq0$, we have
468:  $\varphi([x:y:z])=[\frac{1}{x}:\frac{1}{y}:\frac{1}{z}]$,
469:  which shows that $\varphi$ is an involution.
470:  The map $\varphi$ is undefined at the three coordinate points 
471:  $[1:0:0]$, $[0:1:0]$, and $[0:0:1]$, which are its \Blue{{\sl basepoints}}.
472:  For $xy\neq 0$ we have $\varphi([x:y:0])=[0:0:xy]=[0:0:1]$, and so the map $\varphi$
473:  \Blue{{\sl contracts}} the coordinate line $z=0$ to the point $[0:0:1]$.
474:  The other coordinate lines are also contracted by $\varphi$.
475:  Furthermore, as $\varphi([1:ty:tz])=[t^2yz:tz:ty]=[tyz:z:y]$, we see that if $y,z$ are fixed but
476:  $t$ approaches zero, then $\varphi([1:ty:tz])$ approaches $[0:z:y]$.
477:  Thus the map $\varphi$ blows up the basepoint $[1:0:0]$ into the line $x=0$.
478: 
479:  We call this map, or any map obtained from it by linear substitution, a
480:  standard quadratic Cremona transformation.
481:  There is a second, non standard, quadratic Cremona given by
482:  $[x:y:z]\mapsto[x^2 :  yz : xz]$.
483:  We leave the computation of its contracted curves and the resolution of its basepoints as
484:  an exercise for the reader.
485: \end{example}
486: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
487: 
488: 
489: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
490: \begin{example}[Tensor product patches]\label{Ex:tensor}
491: Let $a,b$ be positive integers.
492: Let $\calA$ be the integer points in the $a\times b$ rectangle,
493: $\{(i,j) :  0\leq i\leq a,\ 0\leq j\leq b\}$, 
494: and select weights $w_{i,j}:=\binom{i}{a}\binom{j}{b}$.
495: Then the corresponding toric B\'ezier functions are
496: %
497:  \begin{equation}\label{Eq:TPP}
498:    \beta_{i,j}(s,t)\ :=\ \tbinom{i}{a}\tbinom{j}{b}
499:          s^i(a-s)^{a-i} t^j(b-t)^{b-j}\,,
500:  \end{equation}
501: %
502: and the homogeneous polynomial is
503: \[
504:    F_{\calA,w} = \sum_{i=0}^a\sum_{j=0}^b \tbinom{i}{a}\tbinom{j}{b}x^iy^jz^{a+b-i-j}
505:     \ =\ (x+z)^a(y+z)^b\,.
506: \]
507: Removing the common factor $(x+z)^{a-1}(y+z)^{b-1}$ from the partial
508: derivatives of $F_{\calA,w}$ shows that
509: %
510:  \begin{multline*}
511:   \qquad T(F_{\calA,w})\ \equiv\ \langle  ax(y+z),\ by(x+z),\ z(a(y+z) \,+\,b(x+z))\rangle\\
512:   =\  \langle  (x+z)(y+z),\ z(x+z),\  z(y+z)\rangle\,, \qquad
513:  \end{multline*}
514: %
515: %  Let A,B,C be the first three quadratics
516: %  set   D := A/a - B/b = z(x-y)
517: %  C + aD = (a+b)z(x+z)
518: %  C - bD = (a+b)z(y+z)
519: %  These can be added to A or B to get the first of the three simpler quadrics
520: %
521: which defines a quadratic Cremona transformation with base points
522: \[
523:    \bigl\{ [1:1:-1],\ [1:0:0],\  [0:1:0]\bigr\}\,.
524: \]
525: By Proposition~\ref{P:TPLS}, this patch~\eqref{Eq:TPP} has rational linear precision.
526: This is well-known, as after a change of coordinates, these blending functions define
527: the tensor product patch of bidegree $(a,b)$, which has rational linear precision.
528: \eqed
529: \end{example}
530: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
531: 
532: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
533: \begin{example}[B\'ezier triangles]
534: Let $\calA:=\{(i,j) :  0\leq i,j\mbox{ and } i+j\leq d\}$.
535: These are the integer points inside the triangle below.
536: %
537:  \[
538:    \begin{picture}(112,66)(-31,-7)
539:     \put(-1,-1){\includegraphics{figures/Triangle.eps}}
540:     \thicklines
541:     \put(-31,48){$(0,d)$}  
542:     \put( 54,-3){$(d,0)$}
543:     \put(-31,-3){$(0,0)$}   
544:    \end{picture}
545:  \]
546: %
547: If we select weights $\Blue{w_{i,j}}:=\frac{d!}{i!j!(d-i-j)!}$, the toric B\'ezier functions are 
548: %
549:  \begin{equation}\label{Eq:BT}
550:    \beta_{i,j}(s,t)\ :=\ \tfrac{d!}{i!j!(d-i-j)!}
551:               s^it^j(d-s-t)^{d-i-j}\,,
552:  \end{equation}
553: %
554: and the homogeneous polynomial is
555: \[
556:    F_{\calA,w} = \sum_{i+j+k=d} \tfrac{d!}{i!j!k!}x^iy^jz^k
557:     \ =\ (x+y+z)^d\,.
558: \]
559: Its toric polar linear system is
560: \[
561:    T(F_{\calA,w}) = \left\langle xd(x+y+z)^{d-1}, yd(x+y+z)^{d-1}, zd(x+y+z)^{d-1}\right\rangle
562:    \ \equiv\ \langle x, y, z\rangle\,,
563: \]
564: which defines the identity map $\P^2\to\P^2$.
565: Thus the patch with blending functions~\eqref{Eq:BT} has rational linear precision.
566: These blending functions are essentially the standard bivariate Bernstein polynomials,
567: which are used in B\'ezier triangles and have linear precision.
568: \eqed
569: \end{example}
570: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
571: 
572: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
573: \begin{example}[Trapezoids]\label{Ex:trapezoid}
574: Let $b,d\geq 1$ and $a\geq0$ be integers, and set 
575: \[
576:   \calA\ :=\ \{ (i,j)\,:\, 0\leq j\leq b\ 
577:     \mbox{ and }\  0\leq i\leq a+db-dj\}\,,
578: \]
579: which are the integer points inside the trapezoid below.
580: \[
581:   \begin{picture}(238,53)(-34,-8)
582:     \put(-3,-3){\includegraphics{figures/Trapezoid.eps}}
583:     \put( 80,30){$(a,b)$}
584:     \put(-30,-3){$(0,0)$}   
585:     \put(135,-3){$(a+db,0)$}
586:     \put(-30,30){$(0,b)$}  
587:   \end{picture}
588: \]
589: Choose weights  $\Blue{w_{i,j}}:=\binom{b}{j}\binom{a+db-dj}{i}$.
590: Then the toric B\'ezier functions are
591: %
592:  \begin{equation}\label{Eq:Trapezoid}
593:    \beta_{i,j}(s,t)\ :=\ \binom{b}{j}\binom{a+db-dj}{i}
594:        s^i(a+db-s-dt)^{a+db-dj-i} t^j(b-t)^{b-j}\,,
595:  \end{equation}
596: %
597: and the homogeneous polynomial  is
598: \[
599:    F_{\calA,w}\ =\ \sum_{j=0}^b\sum_{i=0}^{a+db-dj} \tbinom{b}{j}\tbinom{a+db-dj}{i}
600:       x^i y^j z^{a+db-i-j}\ =\ 
601:     (x+z)^a\bigl( (x+z)^d + y z^{d-1}\bigr)^b\,.
602: \]
603: The partial derivatives of $F_{\calA,w}$ have common factor
604: $(x+z)^{a-1}\bigl( (x+z)^d + y z^{d-1}\bigr)^{b-1}$.
605: %
606: %  We remove this common factor to obtain
607: %
608: % A   ax [ (x+z)^d + yz^{d-1} ] + bdx (x+z)^d ,
609: % B   byz^{d-1}(x+z) , 
610: % C   az [ (x+z)^d + yz^{d-1} ] + bdz (x+z)^d + b(d-1)(x+z)yz^{d-1} .
611: %
612: % C-(d-1)B = z ( a [ (x+z)^d + yz^{d-1} ] + bd (x+z)^d )
613: %        A = x ( a [ (x+z)^d + yz^{d-1} ] + bd (x+z)^d )
614: %
615: % Adding these, we get
616: %
617: %   (a +bd) (x+z)^{d+1}  + a (x+z) yz^{d-1}
618: %
619: % Subtracting a B/ b, gives
620: %
621: % (a +bd) (x+z)^{d+1}
622: %
623: Removing this and performing some linear algebra shows that 
624: %
625:  \[
626:     T(F_{\calA,w})\ \equiv\ \left\langle  (x+z)^{d+1},\ yz^{d-1}(x+z),\ 
627:     z( (a+bd)(x+z)^d + ayz^{d-1} )\right\rangle\,.
628:  \]
629: %
630: This has a base point at $[1:0:-1]$ of multiplicity 1 and 
631: one at $[0:1:0]$ of multiplicity $d$.
632: To see that it defines a birational map, work in the affine chart where $x+z\neq 0$,
633: and assume that $x=1-z$.
634: Then the corresponding rational map is
635: \[
636:   (y,z)\ \longmapsto\ (yz^{d-1}, (a+bd)z^d+yz^{d-1})\,.
637: \]
638: Changing coordinates, this is $(y,z)\mapsto(yz^{d-1},z)$,
639: which is a bijection when $z\neq 0$.
640: \eqed
641: \end{example}
642: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
643: 
644: 
645: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
646: \begin{remark}\label{R:trapezoid}
647:  The first three patches are widely used and implemented in CAD software.  
648:  The trapezoid patch reduces to the B\'ezier triangle when $a=0$ and $d=1$, and to the
649:  tensor product patch when $d=0$.
650:  While the trapezoid patch for general parameters has not been used explicitly in modeling, 
651:  special cases of it have appeared implicitly.
652:  For example, a rational ruled surface in $\R^3$ of degree $2a+d$ with directrix of
653:  minimal degree $a$ and general one of degree $a+d$~\cite[\S5.2]{PW01} is the image of such
654:  a patch (here, $b=1$). 
655:  B\'ezier quad patches on a sphere bounded by circular arcs of minimal type
656:  $(2,4)$~\cite{Kr06} are also trapezoidal.
657:  Some quad patches on rational canal surfaces~\cite{Kr07} can be represented by trapezoidal
658:  patches with $b=2$. 
659:  The full possibilities for the use of the trapezid patch in modeling have yet to be
660:  developed. 
661: \QED
662: \end{remark}
663: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
664: 
665: 
666: 
667: 
668: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
669: %
670: \section{Toric polar Cremona transformations}\label{S:Basics}
671: %
672: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
673: 
674: We classify toric surface  ($n=2$) patches with rational linear precision
675: through the algebraic relaxation of classifying the homogeneous polynomials 
676: (\Blue{{\sl forms}}) $F=F(x,y,z)\in \C[x,y,z]$ 
677: whose toric polar linear system defines a birational map $\P^2\rat\P^2$.
678: Write $F_x$ for $\frac{\partial}{\partial x}F$, and the same for the other variables
679: $y$ and $z$.
680: We will write $F=0$ or simply $F$ for the reduced  curve defined by $F$ in $\P^2$.
681: 
682: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
683: \begin{definition}
684:  Let $F$ be a form.
685:  The vector space $\Blue{T(F)}:=\langle xF_x, yF_y,zF_z \rangle$ defines the 
686:  \Blue{{\sl toric polar linear system}} of curves on $\P^2$ and the 
687:  \Blue{{\sl toric polar map}}
688: %
689:  \begin{eqnarray*}
690:     \Blue{\varphi_F}\ \colon\ \P^2 &\longrat& \P^2\\
691:         {} [\;x\ :\ y\ :\ z\;]&\longmapsto& [\;xF_x\ :\ yF_y\ :\ zF_z\;]\,,
692:  \end{eqnarray*}
693: % 
694:  which maps curves in $T(F)$ into lines in the target $\P^2$.
695: We say that $F$ defines a 
696: \Blue{{\sl toric polar Cremona transformation}} if
697:  this map is birational. 
698: \end{definition}
699: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
700: We establish some elementary properties of such forms $F$ and linear 
701: systems $T(F)$, and then give our classification of forms that define toric
702: polar Cremona transformations.
703: 
704: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
705: %
706: %
707: \subsection{Equivalence of forms}
708: There are some transformations which send one form defining a toric polar Cremona
709: transformation into another such form. 
710: Those which are invertible define an equivalence relation on forms, and our classification
711: is up to this equivalence. 
712: 
713: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
714: \begin{lemma}\label{L:power}
715:  A form $F$ defines a toric polar Cremona transformation if and only if every power $F^a$
716:  for $a>0$ defines a toric polar Cremona transformation.
717: \end{lemma}
718: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
719: 
720: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
721: \noindent{\it Proof.}
722:  The toric polar linear systems of $F$ and $F^a$ are equivalent,
723: \[
724:    \left\langle x \tfrac{\partial F^a}{\partial x},\;
725:                 y \tfrac{\partial F^a}{\partial y},\;
726:                 z \tfrac{\partial F^a}{\partial z}\right\rangle\ =\ 
727:    \langle axF^{a-1}F_x,\, ayF^{a-1}F_y,\,azF^{a-1}F_z\rangle \ \equiv\ 
728:    \langle xF_x,\, yF_y,\,zF_z\rangle\,.
729:     \eqno{\QED}
730: \]
731: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
732: 
733: 
734: The linearity of differentiation implies that $F(x,y,z)$ defines a toric polar Cremona  
735: transformation if and only if $F(ax,by,cz)$ defines a toric polar Cremona  
736: transformation, for all non zero $a,b,c\in\C$.
737: Call this \Blue{{\sl scaling the variables}}.
738: Multiplication by a monomial also preserves the property of defining a toric polar Cremona
739: transformation. 
740: 
741: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
742: \begin{lemma}\label{L:monomial}
743:   A form $F$ defines a toric polar Cremona transformation if and only if
744:   $x^ay^bz^cF$ defines a toric polar Cremona transformation, for any positive integers
745:   $a,b,c$. 
746: \end{lemma}
747: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
748: 
749: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
750: \begin{proof}
751:   It suffices to check that $T(xF)\equiv T(F)$.
752:   Note that $T(xF)$ is
753: \[
754:       \left\langle x \tfrac{\partial}{\partial x} xF,\;
755:                 y \tfrac{\partial}{\partial y} xF,\;
756:                 z \tfrac{\partial}{\partial z} xF\right\rangle\ =\ 
757:    \langle xF + x^2F_x,\, yxF_y,\,zxF_z\rangle \ \equiv\ 
758:    \langle xF_x,\, yF_y,\,zF_z\rangle\,,
759: \]
760:   which is $T(F)$.
761:   The last equivalence follows by removing the common factor $x$ and
762:   applying the Euler relation, which is $xF_x+yF_y+zF_z=\deg(F)F$.
763: \end{proof}
764: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
765: 
766: The calculations in the proof of Lemma~\ref{L:monomial}
767: hold when the exponents $a,b,c$
768: are \Blue{{\sl any}} integers.
769: Consequently, $F$ may be any homogeneous Laurent polynomial.
770: For example,
771: \[
772:    y^{-1}z^{-1} + x^{-2}yz^{-1} + x^{-2}y^{-1}z 
773:    - 2 ( x^{-1}z^{-1} + x^{-1}y^{-1} + x^{-2} )
774: \]
775: is a Laurent form defining a toric polar Cremona transformation.
776: (This is the form of Theorem~\ref{T:Classification}(3) with $d=1$ multiplied by the
777: monomial $x^{-2}y^{-1}z^{-1}$.) 
778: 
779: A third class of transformations are the invertible monomial transformations.
780: A vector $\alpha=(\alpha_1,\alpha_2,\alpha_3)\in\Z^3$ corresponds to a (Laurent) monomial
781: $\Blue{t^\alpha}:=x^{\alpha_1}y^{\alpha_2}z^{\alpha_3}$ of degree $\Blue{|\alpha|}:=\alpha_1+\alpha_2+\alpha_3$.
782: Let $\alpha,\beta,\gamma\in\Z^3$ be three exponent vectors and consider the map
783: $\C^3\to\C^3$ defined by
784: %
785:  \begin{equation}\label{Eq:monom_map}
786:     (x,y,z)\ \longmapsto\ (t^\alpha,t^\beta,t^\gamma)\,.
787:  \end{equation}
788: %
789: 
790: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
791: \begin{lemma}\label{L:monomial trans}
792:  The formula~$\eqref{Eq:monom_map}$ defines a rational map $\P^2\rat\P^2$ if and only if\/
793:  $|\alpha|=|\beta|=|\gamma|$. 
794:  This map is invertible if and only if $\alpha-\gamma$ and $\beta-\gamma$ form a basis for 
795:  $\{v\in\Z^3  :  |v|=0\}$.
796: \end{lemma}
797: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
798: 
799: We prove Lemma~\ref{L:monomial trans} later.
800: Suppose that $\calA:=\{\alpha,\beta,\gamma\}\subset\Z^3$ satisfies the hypotheses of
801: Lemma~\ref{L:monomial trans} so that the map $\Blue{\varphi_\calA}\colon\P^2\rat\P^2$
802: defined by~\eqref{Eq:monom_map} is a birational isomorphism.
803: It induces a map $\Blue{\varphi^*_\calA}$ on monomials $x^ay^bz^c$ by
804: %
805:  \begin{equation}\label{Eq:imt}
806:    \varphi^*_\calA(x^ay^bz^c)\ :=\ t^{\alpha a+\beta b+\gamma c}\ =:\ 
807:     \Blue{t^{\Blue{\calA {\bf a}}}}\,,
808:  \end{equation}
809: %
810:  where ${\bf a}:=(a,b,c)^T$ and $\calA {\bf a}$ is the multiplication of the 
811:  vector ${\bf a}$ by the matrix $\calA$ whose columns are $\alpha,\beta,\gamma$.
812:  When $\calA\in\mbox{Mat}_{3\times 3}\Q$ is invertible and satisfies the hypothesis of
813:  Lemma~\ref{L:monomial trans}, we call $\varphi^*_\calA$ an 
814:  \Blue{{\sl invertible monomial transformation}}. 
815:  Under the hypotheses of Lemma~\ref{L:monomial trans}, the condition that $\calA$ is
816:  invertible is equivalent to $|\alpha|\neq 0$.
817: 
818: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
819: \begin{lemma}\label{L:imt}
820:   A form $F$ defines a toric polar Cremona transformation if and only if
821:   $\varphi^*_\calA(F)$ does for any invertible monomial transformation $\varphi^*_\calA$. 
822: \end{lemma}
823: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
824: 
825: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
826: \begin{proof}
827:  By~\eqref{Eq:imt}, the toric derivative 
828:  $x\frac{\partial}{\partial x}\varphi^*_\calA(t^{\bf a})$
829:  is 
830: \[
831:    (A_1\cdot {\bf a}) t^{\calA {\bf a}}\ =\ 
832:   \varphi^*_\calA((A_1\cdot {\bf a})t^{\bf a})
833:   \ =\  A_1\cdot \varphi^*_\calA( x\tfrac{\partial}{\partial x}t^{{\bf a}},\,
834:                           y\tfrac{\partial}{\partial y}t^{{\bf a}},\,
835:                           z\tfrac{\partial}{\partial z}t^{{\bf a}})^T\,,
836: \]
837:  where 
838:  $A_1$ is the first row of the matrix $\calA$.
839:  Thus
840: %
841:  \begin{eqnarray*}
842:      \left(  x\tfrac{\partial}{\partial x}\varphi^*_\calA(t^{{\bf a}}),\,
843:              y\tfrac{\partial}{\partial y}\varphi^*_\calA(t^{{\bf a}}),\,
844:              z\tfrac{\partial}{\partial z}\varphi^*_\calA(t^{{\bf a}}) \right)^T
845:      &=&
846:      \calA\left(\varphi^*_\calA( x\tfrac{\partial}{\partial x}t^{{\bf a}},\,
847:                           y\tfrac{\partial}{\partial y}t^{{\bf a}},\,
848:                           z\tfrac{\partial}{\partial z}t^{{\bf a}})^T\right)\,,\\
849:   &=& \varphi^*_\calA( x\tfrac{\partial}{\partial x}t^{{\bf a}},\,
850:                           y\tfrac{\partial}{\partial y}t^{{\bf a}},\,
851:                           z\tfrac{\partial}{\partial z}t^{{\bf a}})^T\,.
852:  \end{eqnarray*}
853:  as $\calA$ is invertible.
854:  Thus we have the relation between the toric polar linear systems
855: \[
856:    T( \varphi^*_\calA(F))\ =\ \varphi^*_\calA(T(F))
857: \]
858:  for any homogeneous polynomial $F$.
859:  The lemma follows as $\varphi_\calA$ is birational.
860: \end{proof}
861: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
862: 
863: 
864: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
865: \begin{definition}
866:  Let $\Blue{\C[x^{\pm}]}:=\C[x,x^{-1}, y,y^{-1}, z,z^{-1}]$ be the ring of Laurent
867:  polynomials, the coordinate ring of the torus $(\C^*)^3$.
868:  This is $\Z$-graded by the total degree of a monomial.
869:  Forms $F,G\in\C[x^{\pm}]$ are \Blue{{\sl equivalent}} if 
870:  $G=\psi^*(F)$, where $\psi$ is a composition of
871: \begin{itemize}
872:   \item[(a)] scaling variables, $[x:y:z]\mapsto [ax:by:cz]$, or
873:   \item[(b)] multiplication by a monomial, or
874:   \item[(c)] an invertible monomial transformation.
875: \end{itemize}
876:  Our classification is up to this equivalence.
877:  By (b), it is no loss to assume that a Laurent form $F$ is an ordinary form
878:  (in $\C[x,y,z]$).
879: \end{definition}
880: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
881: 
882: 
883: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
884: \begin{proof}[Proof of Lemma~$\ref{L:monomial trans}$]
885:  The first statement is clear as a rational map $\C^3\to\C^3$ drops to a map
886:  $\P^2\rat\P^2$ if and only if it is defined by homogeneous rational forms of the same
887:  degree. 
888:  For the second, consider the map $(\C^*)^3\to\P^2$ defined by
889: \[
890:   \varphi\ \colon\ t:=(x,y,z)\ \longmapsto\ [t^\alpha\,:\,t^\beta\,:\,t^\gamma]\,,
891: \]
892:  and suppose that we have $s,t\in(\C^*)^3$ with $\varphi(s)=\varphi(t)$. After rescaling
893:  in the source, we may assume that $t=(1,1,1)$.
894:  In particular, $s$ is a solution of
895: \[
896:    s^\alpha\ =\ \lambda\,,\qquad
897:    s^\beta\ =\ \lambda\,,\qquad\mbox{and}\qquad
898:    s^\gamma\ =\ \lambda\,,
899: \]
900:  for some $\lambda\in\C^*$.
901:  But we also have
902: %
903:  \begin{equation}\label{Eq:u_eq}
904:    s^{\alpha-\gamma}\ =\ 1\ =\ s^{\beta-\gamma}\,.
905:  \end{equation}
906: %
907:  Since $|\alpha|=|\beta|=|\gamma|$, solutions to~\eqref{Eq:u_eq} include the diagonal
908:  torus $\Delta:=\{(a,a,a) :  a\in\C^*\}$, so we see again that $\varphi$ is defined on
909:  the dense torus $(\C^*)^3/\Delta$ of $\P^2$.
910:  This map on the dense torus is an isomorphism if and only if points of the diagonal torus
911:  are the only solutions to~\eqref{Eq:u_eq},
912:  which is equivalent to the condition that the exponents $\alpha-\gamma$ and
913:  $\beta-\gamma$ are a basis for the free abelian group
914:  $\{v\in\Z^3 :  |v|=0\}$.
915: \end{proof}
916: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
917: 
918: 
919: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
920: %
921: %
922: \subsection{Proof of Theorem~\ref{T:Classification}}
923: Let $F$ be a form defining a toric polar Cremona transformation.
924: We first classify the possible irreducible factors of $F$, and then 
925: determine which factors may occur together.
926: The classification of the factors of $F$ occupies
927: Sections~\ref{S:contracted}, \ref{S:not_contracted}, and~\ref{S:technical}.
928: 
929: 
930: Under the rational map $\varphi_F$ each component of each curve in the toric polar linear
931: system is either contracted (mapped to a point) or mapped to a dense subset of a curve.
932: As $F\in T(F)$, this in particular holds for the curve $F=0$, whose
933: components correspond to the irreducible factors of $F$.
934: (We always take the reduced structure on this curve, that is,  we consider the set of zeroes
935: of $F$, not the scheme defined by $F$.)
936: An irreducible factor of $F$ is \Blue{{\sl contracted}}, respectively 
937: \Blue{{\sl not contracted}}, if the corresponding component of the curve $F=0$ is
938: contracted by the linear system $T(F)$, respectively not contracted. 
939: There are three possibilities for the factors of $F$.
940: %
941:  \begin{equation}\label{Eq:Trilemma}
942:   \begin{minipage}[c]{12cm}
943:      \begin{enumerate}
944:       \item $F$ has no contracted factors, or 
945:       \item $F$ has only contracted factors, or 
946:       \item $F$ has both contracted and non contracted factors.
947:      \end{enumerate}
948:   \end{minipage}
949:  \end{equation}
950: %
951: 
952: 
953: We get information about the factors of $F$ from a  simple, but useful restriction lemma.
954: 
955: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
956: \begin{lemma}\label{L:restriction}
957:  Suppose that $G$ is an irreducible factor of a form $F$.
958:  Then the restrictions of the toric polar linear systems of $F$ and of $G$ to the curve
959:  $G=0$ are equivalent.
960: \end{lemma}
961: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
962: 
963: \noindent{\it Proof.}
964:  Write $F = G^nH$ with $G$ and $H$ coprime. 
965:  Then 
966: %
967:  \[ 
968:    T(F)\ =\ 
969:     \langle nxG^{n-1}G_xH + xG^nH_x, \ 
970:             nyG^{n-1}G_yH + yG^nH_y, \ 
971:             nzG^{n-1}G_zH + zG^nH_z \rangle\,.
972:  \]
973: %
974:  After factoring out $G^{n-1}$, restricting to $G=0$, and factoring out $nH$ we obtain
975: %
976:  \[
977:     T(F)|_G\ \equiv \ \langle xG_x, yG_y, zG_z \rangle|_{G}\ =\  T(G)|_G\,.\eqno{\QED}
978:  \]
979: %
980: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
981: 
982: A cornerstone of our classification is that there is a strong restriction on the 
983: singularities of the curve $F=0$.
984: A singular point $p$ of a curve is \Blue{{\it ordinary}} if locally near $p$, the curve
985: consists of $r>1$ smooth branches that meet transversally at $p$.
986: In Section~\ref{S:technical} we prove the following theorem.
987: \medskip
988: 
989: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
990: \begin{theorem}\label{Th:singularities}
991:   If a form $F$ coprime to $xyz$ defines a toric polar
992:   Cremona transformation, then the curve $F=0$ has at most one singular point
993:   outside the coordinate lines, and if there is such a point, then all factors of $F$ are
994:   contracted.  Furthermore, if this singular point is ordinary,
995:  then $F$ has at most two distinct factors through the singular point.
996: \end{theorem}
997: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
998: 
999: 
1000: Since the toric polar map $\varphi_F$ is birational,
1001: each component of a curve in the toric polar linear system $T(F)$ is either contracted 
1002: or mapped birationally onto a line, and at most one component of a 
1003: curve is not contracted.
1004: This in particular holds for $F$.
1005: 
1006: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1007: \begin{cor}\label{Co:non_contr}
1008:   At most one factor of $F$ is not contracted.
1009: \end{cor}
1010: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1011: 
1012: 
1013: In Section~\ref{S:not_contracted} we classify the possible non contracted factors of $F$.
1014: 
1015: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1016: \begin{theorem}\label{T:not_contracted}
1017:  If $F$ is an irreducible form defining a curve with no
1018:  singularities outside the coordinate lines whose toric polar linear
1019:  system maps this curve birationally onto a line,
1020:  then $F$ is equivalent to one of the following forms,
1021: %
1022:  \begin{enumerate}
1023: 
1024:   \item $x^2+y^2+z^2-2(xy+xz+yz)$, or
1025: 
1026:   \item $(x+z)^d + yz^{d-1}$, for some integer $d\geq 1$.
1027: 
1028:  \end{enumerate}
1029: %
1030: \end{theorem}
1031: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1032: 
1033: Example~\ref{Ex:trapezoid} with $a=0$ and $b=1$ shows that the 
1034: second class of forms define toric polar Cremona
1035: transformations, and the following example shows that the first class also does.
1036: 
1037: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1038: 
1039: \begin{example}\label{Ex:relaxation}
1040:  The form $F$ of Theorem~\ref{T:not_contracted}(1) has the toric polar linear system 
1041: %
1042:  \begin{eqnarray}
1043:    T(F)& =& \langle x^2-xy-xz,\ y^2-xy-yz,\ z^2-xz-yz\rangle \nonumber\\
1044:    &=&\langle \Blue{(x{-}y{-}z)}\Brown{(y{-}x{-}z)},\ 
1045:  \label{eq:strange_equality}
1046:              \Blue{(x{-}y{-}z)}\Purple{(z{-}x{-}y)},\ 
1047:              \Purple{(y{-}x{-}z)}\Brown{(z{-}x{-}y)}\rangle\ , 
1048:  \end{eqnarray}
1049: %
1050:  which defines a quadratic Cremona transformation with base points
1051: \[
1052:    \bigl\{ [1:0:1],\ [0:1:1],\  [1:1:0]\bigr\}\,.
1053: \]
1054:  To help see the equality~\eqref{eq:strange_equality}, note that
1055:  $-xF_x-yF_y+zF_z= \Blue{(x-y-z)}\Brown{(y-x-z)}$.
1056: 
1057:  This example shows that the algebraic relaxation (seeking polynomials $F$ with arbitrary
1058:  coefficients whose toric
1059:  polar linear system is birational) of the original problem from geometric modeling 
1060:  has solutions which do not come from geometric modeling, as the coefficients of $F$
1061:  cannot simultaneously be made positive. \eqed 
1062: \end{example}
1063: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1064: 
1065: In Section~\ref{S:contracted}, we study the possible contracted factors of $F$
1066: and prove the following lemma.
1067: 
1068: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1069: \begin{lemma}\label{L:contr}
1070:   Suppose that $G$ is a contracted factor of $F$.
1071:   Then, up to a permutation of variables, $G=x^a+\alpha y^bz^{a-b}$ with $a$, $b$
1072:   coprime and $\alpha\neq 0$.
1073: \end{lemma}
1074: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1075: 
1076: We now present a proof of Theorem~\ref{T:Classification},
1077: following the three cases of~\eqref{Eq:Trilemma}.
1078: 
1079: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1080: %
1081: \subsubsection{$F$ has no contracted factors}
1082: %
1083: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1084: In this case, Corollary~\ref{Co:non_contr} implies that $F$ has a single irreducible
1085: factor.
1086: Since this factor is not contracted  it is equivalent 
1087: to one of the forms described in Theorem~\ref{T:not_contracted}.
1088: Since these forms define toric polar Cremona transformations, Lemma~\ref{L:power} implies
1089: that any power of such a form defines a toric polar Cremona transformation.
1090: In particular, $F$ defines a toric polar Cremona transformation.
1091: This establishes part (2) of Theorem~\ref{T:Classification}(2) when
1092: $a=0$ and also part (3).
1093: 
1094: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1095: %
1096: \subsubsection{$F$ has only contracted factors}
1097: %
1098: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1099: We outline the proof in this case, which is carried out in
1100: Section~\ref{S:contracted}. 
1101: Suppose first that $F$ defines a toric polar Cremona transformation.
1102: By Lemma~\ref{L:contr}, every contracted factor is a binomial.
1103: We first show that any two contracted factors of $F$ are simultaneously equivalent to 
1104: \[
1105:    (z+x)\qquad\mbox{and}\qquad(z+y)\,,
1106: \]
1107: and in particular they meet outside the coordinate lines.
1108: 
1109: Suppose that all factors of $F$ are contracted, then we show that $F$ has at least two 
1110: irreducible factors.
1111: We next show that if $F$ has three or more factors, then we may assume that they intersect
1112: transversally at $[1:1:-1]$.   
1113: Hence $F=0$ has an ordinary singularity of multiplicity at least $3$, which contradicts
1114: the last part of Theorem \ref{Th:singularities}. 
1115: Therefore  $F$ is equivalent to
1116: %
1117:  \begin{equation}\label{F:two_factors}
1118:    (x+z)^a(y+z)^b\,,
1119:  \end{equation}
1120: %
1121: with $a,b>0$.
1122: By Example~\ref{Ex:tensor}, any such form defines a
1123: toric polar Cremona transformation,
1124: which completes the proof of Theorem~\ref{T:Classification}(1).
1125: 
1126: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1127: %
1128: \subsubsection{$F$ has both contracted and non contracted factors}
1129: %
1130: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1131: By Corollary~\ref{Co:non_contr}, $F$ has a unique
1132: non contracted factor.
1133: It also has a unique contracted factor.
1134: Indeed, any two contracted factors meet outside the coordinate
1135: lines, and so the curve $F=0$  is singular outside the coordinate lines.
1136: Then Theorem~\ref{Th:singularities} implies that all factors of $F$ are contracted, 
1137: a contradiction. 
1138: All that remains is to examine the different possibilities for
1139: the factors of $F$.
1140: We show that the non contracted factor cannot be equivalent to $x^2+y^2+z^2-2(xy+xz+yz)$.
1141: We then show that if the non contracted factor is equivalent to $(x+z)^d + yz^{d-1}$, then
1142: (after putting it into this form) the contracted factor is $x+z$.
1143: Example~\ref{Ex:trapezoid} shows that all possibilities
1144: \[
1145:    (x+z)^a \bigl((x+z)^d + yz^{d-1} \bigr)^b
1146: \]
1147: with $a\geq 0$ and $b>0$ define toric polar Cremona transformations, 
1148: which completes the proof of Theorem~\ref{T:Classification}.
1149: These claims about the non contracted factors are proven in the following two lemmas.
1150: 
1151: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1152: \begin{lemma}
1153:   If $F$ has a non contracted factor equivalent to 
1154:   $x^2+y^2+z^2-2(xy+xz+yz)$, then it has no other factors.
1155: \end{lemma}
1156: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1157: 
1158: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1159: \begin{proof}
1160:  Suppose that $F$ has two factors, $Q:=x^2+y^2+z^2-2(xy+xz+yz)$, and a contracted factor
1161:  $G$.
1162:  Since permuting the variables does not change $Q$, Lemma~\ref{L:contr} implies that 
1163:  $G=x^A+\alpha y^a z^{A-a}$, with $A,a\geq 0$ coprime and $\alpha$ non zero.
1164: 
1165:  By Theorem~\ref{Th:singularities}, $G$ and $Q$ can meet only on the coordinate lines.
1166:  If we substitute the parameterization $[x:y:z]=[(s+t)^2:s^2:t^2]$ of $Q$ into
1167:  $G$, we obtain
1168: %
1169:  \begin{equation}\label{Eq:subst}
1170:     (s+t)^{2A}\ +\ \alpha s^{2a} t^{2A-2a}\,.
1171:  \end{equation}
1172: %
1173: % Since $Q$ meets the coordinate lines only when $s=0$ or $t=0$ or $s+t=0$, 
1174:  The condition that $Q$ and $G$ meet only on the coordinate lines implies that the only
1175:  factors of~\eqref{Eq:subst} are $s$, $t$, or $s+t$.
1176:  But this implies that $A=a=0$, contradicting our assumption that $G$
1177:  was a non trivial
1178:  factor of $F$.
1179: \end{proof}
1180: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1181: 
1182: 
1183: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1184: \begin{lemma}
1185:   If $F$ has the  non contracted factor $(x+z)^d + yz^{d-1}$, then its other irreducible
1186:   factor must be $x+z$. 
1187: \end{lemma}
1188: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1189: 
1190: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1191: \begin{proof}
1192:  Suppose that $F$ has two factors, $Q:=(x+z)^d + yz^{d-1}$ with $d\geq 1$ and a contracted
1193:  factor $G$, which is necessarily a binomial.
1194:  Multiplying $G$ by a monomial and scaling, we may assume it has the form
1195:  $\alpha+(-1)^ax^Ay^az^{-A-a}$ with $A\geq 0$ and $A,a$ coprime.
1196:  By Theorem~\ref{Th:singularities}, $G$ and $Q$ can meet only on the coordinate lines.
1197:  We solve $Q=0$ for $y$ to obtain $y=-(x+z)^d/z^{d-1}$ and then substitute this into 
1198:  $G$ to obtain
1199: \[
1200:    \alpha+(-1)^ax^A\left(-\tfrac{(x+z)^d}{z^{d-1}}\right)^a z^{-A-a}
1201:    \ =\ \alpha + x^A(x+z)^{ad}z^{-A-ad}\,.
1202: \]
1203:  If we multiply this by $z^{A+ad}$ if $a\geq 0$ and by $z^A(x+z)^{-ad}$ if $a<0$ (and replace
1204:  $a$ by $-a$), this becomes either
1205: \[
1206:    \alpha z^{A+ad}\ +\ x^A (x+z)^{ad}\qquad\mbox{or}\qquad
1207:    \alpha z^A(x+z)^{ad}\ +\ x^A z^{ad}\,.
1208: \]
1209:  Since $G$ and $Q$ can meet only on the coordinate lines, the only possible factors of
1210:  these polynomials are $x$, $z$, and $x+z$.
1211:  Neither $x$ nor $z$ can be a factor as $A\neq ad$ and the coefficients are non zero,
1212:  so $x+z$ is the only factor.
1213:  But then we must have $a=0$, $\alpha=1$, and $A=1$,
1214:  so that $G=1+xz^{-1}$, or, clearing the denominator, $G=x+z$.
1215: \end{proof}
1216: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1217: 
1218: 
1219: 
1220: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1221: %
1222: \section{contracted factors}\label{S:contracted}
1223: %
1224: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1225: 
1226: We study the contracted factors of a form $F$ that defines a toric polar Cremona
1227: transformation.
1228: We first prove Lemma~\ref{L:contr}, that any contracted factor of $F$ is a binomial.
1229: 
1230: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1231: \begin{proof}[Proof of Lemma~$\ref{L:contr}$]
1232:   By Lemma~\ref{L:restriction}, the restrictions of the toric polar maps of $F$ and of $G$
1233:   to the curve $G=0$ coincide.
1234:   Let $T(G)$ be the toric polar linear system of $G$.  
1235:   Since it contracts $G$, $T(G)|_G=\langle xG_x, yG_y, zG_z \rangle|_G$ is
1236:   one-dimensional, and so $T(G)$ is only two-dimensional.
1237:   Thus there is a linear relationship among the toric derivatives of $G$,
1238: \[
1239:     q_1 xG_x\,+\, q_2 yG_y \,+\, q_3 zG_z \ =\ 0\,.
1240: \]
1241:   Writing $G=\sum_i m_i$ as a sum of terms $m_i= \alpha_i x^{a_i}y^{b_i}z^{c_i}$,
1242:   we see that $q_1a_i+q_2b_i=q_3c_i$ for all $i$.
1243:   Thus we may assume that $q_1,q_2,q_3\in\Z$ and they are coprime.
1244:   Permuting variables, we may assume that the $q_j$ are non negative.
1245:   Since $G$ is homogeneous, say of degree $d$, we have $a_i+b_i+c_i=d$, and so
1246: \[
1247:    (q_1+q_3)a_i + (q_2+q_3)b_i\ =\ q_3 d\,.
1248:  \]
1249: % 
1250:   Thus $G(x,y,1)$ is a weighted homogeneous polynomial of degree $q_{3}d$. 
1251:   Since $G$ is irreducible, the only possibilities are $G(x,y,1)=x$ or
1252:   $G(x,y,1)=y$ (neither can occur as $F$ is coprime to $xyz$), or
1253:   $G(x,y,1)=x^a + \alpha y^b$ with $a$ and $b$ coprime, $\alpha\neq 0$ and  
1254:   $(q_{1}+q_{3})a = (q_{2}+q_{3})b$.
1255:   Since $G$ is irreducible, $z$ does not divide $G$ and $G(x,y,1)$ must have degree $d$. 
1256:   Therefore after possibly interchanging $x$ and $y$ we see that
1257:   $G$ has the form claimed.
1258: \end{proof}
1259: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1260: 
1261: Any two contracted factors may be put into a standard form.
1262: 
1263: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1264: \begin{lemma}\label{L:two_factors}
1265:   If $G$ and $H$ are two contracted factors of $F$, then, up to equivalence
1266:   $GH=(x+z)(y+z)$.
1267: \end{lemma}
1268: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1269: 
1270: In particular, any two contracted factors of $F$ meet outside the coordinate lines.
1271: 
1272: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1273: \begin{proof}
1274:   Up to a permutation of the variables, each factor is a prime binomial of the form 
1275: \[
1276:     x^A + \alpha y^az^{A-a}\,,
1277: \]
1278:  where $A,a\geq 0$ are coprime and $\alpha$ is non zero.
1279:  
1280:  By Theorem~\ref{Th:singularities}, $F$ has at most one singularity outside the
1281:  coordinate lines.
1282:  Points common to two factors of $F$ are singular, so 
1283:  the factors $G$ and $H$ define curves that meet at most once outside
1284:  the coordinate axes.
1285:  To study such points, we dehomogenize and set $z=1$.
1286:  Multiplying $G$ and $H$ by monomials, we may suppose that they have the form
1287: %
1288:  \begin{equation}\label{Eq:GandH}
1289:   G\ =\ \alpha\ +\ x^Ay^a\qquad\mbox{and}\qquad
1290:   H\ =\ \beta\ + \ x^By^b\,.
1291:  \end{equation}
1292: %
1293:  Since these are irreducible binomials, $1=\gcd\{|A|,|a|\}=\gcd\{|B|,|b|\}$,
1294:  and since they are coprime $Ab-Ba\neq 0$.
1295: 
1296:  The points common to the two components are the solutions to $G=H=0$.
1297:  The number of solutions to such a zero-dimensional binomial system is
1298:  $|Ab-Ba|$~\cite[\S~3.2]{Stu}.
1299:  Since there can be at most one such point,
1300:  $|Ab-Ba|=1$.
1301:  Interchanging the roles of $(A,a)$ and $(B,b)$ if necessary, we may assume that 
1302:  $Ab-Ba=1$.
1303:  Under the invertible substitution $x=x^by^{-a}$ and $y=x^{-B}y^A$,~\eqref{Eq:GandH}
1304:  becomes $\alpha+x$ and $\beta+y$.
1305:  Scaling $x$ and $y$ and rehomogenizing, 
1306:  we may assume that the binomials are $x+z$ and $y+z$.
1307: \end{proof}
1308: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1309: 
1310: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1311: \begin{lemma}\label{L:irreducible}
1312:   If $F$ has only a single irreducible factor, then that factor is not contracted.
1313: \end{lemma}
1314: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1315: 
1316: In particular a form with all factors contracted must have at least two factors. 
1317: 
1318: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1319: \begin{proof}
1320:   Let $G$ be the irreducible factor of $F$, then $F=G^a$ for some $a>1$.
1321:   By Lemma~\ref{L:power} the toric polar map $\varphi_F$ of $F$ coincides with the toric
1322:   polar map $\varphi_G$ of $G$. 
1323:   If $G=0$ is contracted, then, as in the proof of  Lemma~\ref{L:contr}, 
1324:   $T(G)|_G$ is one-dimensional and thus $T(G)$ is only two-dimensional
1325:   so that $xG_x$, $yG_y$ and $zG_z$ are dependent. 
1326:   But then $\varphi_G$, and hence $\varphi_F$, cannot be birational.
1327: \end{proof}
1328: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1329: 
1330: We classify forms $F$ defining a toric polar Cremona
1331: transformation with all factors contracted.
1332: 
1333: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1334: \begin{theorem}\label{T:all_contr}
1335:   If all factors of $F$ are contracted, then $F$ is equivalent to
1336:   $(x+z)^a(y+z)^b$, for some $a,b>0$.
1337: \end{theorem}
1338: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1339: 
1340: 
1341: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1342: \begin{proof}
1343:  Suppose that all factors of $F$ are contracted.
1344:  By Lemma~\ref{L:two_factors}, we may assume that two of the irreducible factors of $F$
1345:  are $x+z$ and $y+z$.
1346:  We only need to show that there are no further contracted factors of $F$.
1347:  Suppose there is another contracted factor.
1348:  After multiplying by a monomial, this will have the form
1349: \[
1350:    \gamma\ +\ x^Cy^cz^{-C-c}\,,
1351: \]
1352:  with $\gamma\in\C^*$ and $C>1$.
1353:  By Theorem~\ref{Th:singularities}, $F$ has at most one singularity outside of the
1354:  coordinate lines, and so this factor must meet the other factors only in the 
1355:  point $[1:1:-1]$ where they meet.
1356:  It follows that $\gamma=\pm 1$ and $C=|c|=1$.
1357:  
1358:  We see that the only possible irreducible factors of $F$ are
1359: \[ 
1360:   x+z\,,\quad y+z\,,\quad z^2-xy\,,\quad\mbox{ and }\quad
1361:   y-x\,.
1362: \]
1363: Since these four factors have distinct tangents at $[1:1:-1]$, the singularity of $F$ at
1364: this point is ordinary. 
1365: By the last part of Theorem \ref{Th:singularities}, $F$ can have at most two distinct
1366: factors through $[1:1:-1]$ 
1367: so the theorem follows.
1368: \end{proof}
1369: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1370: 
1371: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1372: %
1373: \section{Irreducible polynomials}\label{S:not_contracted}
1374: %
1375: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1376: 
1377: We classify irreducible factors of $F$ which are not contracted by the toric polar Cremona
1378: transformation.
1379: Specifically, we prove the following theorem.
1380: \medskip
1381: 
1382: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1383: \noindent{\bf Theorem~\ref{T:not_contracted}.}
1384: {\it 
1385:  If $F$ is an irreducible form defining a curve with no
1386:  singularities outside the coordinate lines whose toric polar linear
1387:  system maps this curve birationally onto a line,
1388:  then $F$ is equivalent to one of the following forms,
1389: %
1390:  \begin{enumerate}
1391: 
1392:   \item $x^2+y^2+z^2-2(xy+xz+yz)$, or
1393: 
1394:   \item $(x+z)^d + yz^{d-1}$, for some integer $d\geq 1$.
1395: 
1396:  \end{enumerate}
1397: %
1398: }\medskip
1399: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1400: 
1401: Since the curve $F=0$ is rational, it has a parameterization
1402: $\Blue{\gamma}\colon \P^1\to\P^2$ which determines $F$ up to a constant.
1403: The composition of $\gamma$ with the toric polar Cremona transformation of $F$ is a map 
1404: $\P^1\to \P^2$ of degree 1.
1405: We will deduce from this and the location of the singularities of $F$ that there are exactly
1406: three distinct irreducible factors appearing in $\gamma$. 
1407: 
1408: Applying quadratic Cremona transformations 
1409: puts $\gamma$ into a standard form from which the 
1410: hypothesis on the singularities of $F=0$ restricts $F$ to be equivalent to one of the forms
1411: of Theorem~\ref{T:not_contracted}.
1412: An important technical part of this argument is the local contribution to the arithmetic
1413: genus of a singular point of a binomial curve, which we compute in Section~\ref{Sec:genus}.
1414: 
1415: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1416: %
1417: \subsection{Linear factors in $\gamma$}
1418: %
1419: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1420: Suppose that $F$ is a form of degree $d$ that satisfies the hypotheses of
1421: Theorem~\ref{T:not_contracted} and let $\gamma:=[f:g:h]\,\colon\P^1\to\P^2$ parameterize
1422: the curve $F=0$.
1423: Then $f$, $g$, and $h$ are are coprime forms of degree $d$ on $\P^1$.
1424: Because the toric polar map $\varphi_F$ sends the image of $\gamma$ (the curve $F=0$)
1425: isomorphically onto a line, the map $\P^1\to\P^2$ with components
1426: %
1427:  \begin{equation}\label{E:components}
1428:    f F_x(\gamma),\  gF_y(\gamma),\ \mbox{ and }\ \  hF_z(\gamma)
1429:  \end{equation}
1430: % 
1431: has degree 1, and thus these forms become linear after removing common factors.
1432: We study their syzygy module to show that there are only three distinct irreducible
1433: factors in $fgh$.\smallskip 
1434: 
1435: Choose homogeneous coordinates $[s,t]$ on $\P^1$ with $st$ coprime to
1436: $fgh$. 
1437: As $\gamma$ parameterizes $F$, we have $F(\gamma)\equiv 0$ on $\P^1$.
1438: Differentiating with respect to $s$ and $t$ gives
1439: %
1440:  \begin{equation}\label{Eq:first_relation}
1441:    \left[\begin{matrix}f_s&g_s&h_s\\f_t&g_t&h_t\end{matrix}\right]
1442:    \left[\begin{matrix}F_x(\gamma)\\F_y(\gamma)\\F_z(\gamma)\end{matrix}\right]
1443:    \ =\ \left[\begin{matrix}0\\0\end{matrix}\right]\ .
1444:  \end{equation}
1445: %
1446: Using the Euler relations $sf_s+tf_t=df$ (and the same for $g$ and $h$) gives the
1447: syzygy
1448: %
1449:  \[
1450:     f F_x(\gamma)\ +\ gF_y(\gamma)\ +\ hF_z(\gamma)\ =\ 0\,.
1451:  \]
1452: %
1453: Multiplying the first row of~\eqref{Eq:first_relation} by $fgh$ gives a second syzygy, so
1454: we have 
1455: %
1456:  \begin{equation}\label{Eq:syzygies}
1457:    \left[\begin{matrix}f_sgh&fg_sh&fgh_s\\1&1&1\end{matrix}\right]
1458:    \left[\begin{matrix}fF_x(\gamma)\\gF_y(\gamma)\\hF_z(\gamma)\end{matrix}\right]
1459:    \ =\ \left[\begin{matrix}0\\0\end{matrix}\right]\ .
1460:  \end{equation}
1461: %
1462: %The rows of this $2\times 3$ matrix are linearly independent over $\C(s,t)$ and so they
1463: %generate the syzygy module over $\C[s,t]$, up to saturation.
1464: An equivalent set of syzygies is given by the rows of the matrix
1465: %
1466:  \begin{equation}\label{Eq:other_syzygies}
1467:    \left[\begin{matrix}f_sgh-fgh_s&fg_sh-fgh_s&0\\1&1&1\end{matrix}\right]\ .
1468:  \end{equation}
1469: %
1470: Since the three components~\eqref{E:components} share a common factor whose removal yields
1471: linear forms $(p,q,r)$ with the same syzygy matrix, $r=-p-q$ and 
1472: the removal of common factors from the first row of~\eqref{Eq:other_syzygies}
1473: gives the syzygy $(-q,p,0)$.
1474: 
1475: There are three sources for common factors of the first row of~\eqref{Eq:other_syzygies}.
1476: %
1477:  \begin{enumerate}
1478:   \item Common factors of $f$ and $f_s$, of $g$ and $g_s$, or of $h$ and $h_s$,
1479: 
1480:   \item common factors of some pair of $f$, $g$, or $h$, and
1481: 
1482:   \item common factors of $f_sh-fh_s$ and $g_sh-gh_s$.
1483:  \end{enumerate}
1484: %
1485: 
1486:    A common factor of the third type that is not of type (1) or (2) vanishes at a point $p\in\P^1$ where 
1487: %
1488:  \[
1489:   \mbox{\rm rank} \left[\begin{matrix}f_s&g_s&h_s\\f&g&h\end{matrix}\right]
1490:    \ \leq\ 1\,.
1491:  \]
1492: %
1493:  The Euler relation implies that we also have 
1494: %
1495:  \begin{equation}\label{E:t_differential}
1496:   \mbox{\rm rank} \left[\begin{matrix}f_s&g_s&h_s\\tf_t&tg_t&th_t\end{matrix}\right]
1497:    \ \leq\ 1\,,
1498:  \end{equation}
1499: %
1500:  and so $t$ is a common factor of the third type.
1501:  Suppose now that $t(p)\neq 0$.
1502:  Then~\eqref{E:t_differential} shows that 
1503:  the differential of $\gamma$ does not have full rank at $p$,
1504:  and so the the curve $F=0$ is singular at $\gamma(p)$.
1505:  But such a singular point must lie on a coordinate line of $\P^2$ and so 
1506:  one of $f$, $g$, or $h$ vanishes at $p$.
1507:  Without loss of generality, suppose that $f(p)=0$.
1508:  Then $f_s(p)h(p)=0$, as $f_sh-fh_s$ vanishes at $p$,
1509:  and so the common factor vanishing at $p$ divides either $f_s$ or $h$, and
1510:  is therefore a factor of type (1) or (2).
1511:  Thus $t$ is the only factor of type (3) that is not of type (1) or (2).
1512:  As $fgh$ is coprime to $t$, the common
1513:  factor $t$ has multiplicity 1.
1514: 
1515: Now suppose that $\ell$ is a linear factor of $fgh$ with $\ell^a$, $\ell^b$, and $\ell^c$
1516: exactly dividing $f$, $g$, and $h$, respectively.
1517: Then $\ell^{a+b+c}$ exactly divides $fgh$ and $\ell^{a+b+c-1}$ exactly divides the entries
1518: in the first row of~\eqref{Eq:other_syzygies}.
1519: It follows that if the prime factorization of $fgh$ is $\ell_1^{a_1}\dotsb \ell_k^{a_k}$,
1520: then the common factor of the first row of~\eqref{Eq:other_syzygies} is
1521: \[
1522:     t \ell_1^{a_1-1} \dotsb \ell_k^{a_k-1}\,.
1523: \]
1524: This has degree $3d+1-k$.
1525: Since the entries in the first row of~\eqref{Eq:other_syzygies} have degree $3d-1$, and 
1526: removing this common factor gives linear forms, we have that
1527: \[
1528:    3d-1\ =\ 1+3d+1-k\,,
1529: \]
1530: or $k=3$.
1531: Thus there are exactly three distinct linear factors dividing $fgh$.
1532: 
1533: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1534: %
1535: \subsection{Arithmetic genus of binomial germs}\label{Sec:genus}
1536: %
1537: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1538: 
1539: We compute $\delta_{(0,0)}$, the contribution at the origin to the arithmetic genus of a
1540: curve $C$ with germ 
1541: %
1542:  \begin{equation}\label{E:binomial_germ}
1543:    (x^a\: -\: y^b)\; u(x,y)\ =\ 0\,,
1544:  \end{equation}
1545: %
1546: where $u(0,0)\neq 0$.
1547: Write $p_a(C)$ for the arithmetic genus of a curve $C$.
1548: 
1549: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1550: \begin{lemma}\label{L:binomial_germ}
1551:   Let $C$ be a curve on a smooth surface $S$ with germ~$\eqref{E:binomial_germ}$
1552:   given in local coordinates $(x,y)$ of a point $p\in S$.
1553:   If $\widetilde{C}$ is obtained from $C$ by a sequence of blowups in $p$ and points
1554:   infinitely near $p$ and is smooth at all points infinitely near $p$, then
1555: %
1556:  \begin{equation}\label{E:arith_genus}
1557:    p_a(\widetilde{C})\ =\ p_a(C)\ -\ 
1558:    \tfrac{1}{2}\bigl( (a-1)(b-1) + \gcd(a,b) - 1  \bigr)\,.
1559:  \end{equation}
1560: %
1561: \end{lemma}
1562: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1563: 
1564: We call the difference $p_a(C)-p_a(\widetilde{C})$ the 
1565: \Blue{{\sl $\delta$-invariant of $C$ at $p$}}.
1566: 
1567: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1568: \begin{proof}
1569: Recall the formula~\cite[Cor V.3.7]{Hart} for the arithmetic genus of
1570: the strict transform $C'$ of a curve $C$ obtained by blowing up a point of multiplicity
1571: $a$ in $C$,
1572: %
1573:  \begin{equation}\label{E:genus_contribution}
1574:       p_a(C')\ =\ p_a(C)\ -\ \tfrac{1}{2}a(a-1)\,.
1575:  \end{equation}
1576: %
1577: 
1578: Let $C$ be defined near $p$ by~\eqref{E:binomial_germ}.
1579: Then $C$ has multiplicity  $\min\{a,b\}$ at $p$.
1580: If $\min\{a,b\}=1$, then $C=\widetilde{C}$ and~\eqref{E:arith_genus}
1581: becomes $p_a(\widetilde{C})=p_a(C)$.
1582: 
1583: If $a=b$, then $C$ consists of $a$ smooth branches meeting pairwise transversally at $p$.
1584: Blowing up $p$ separates these branches so that $C'=\widetilde{C}$.
1585: By~\eqref{E:genus_contribution}, we have
1586: \[
1587:    p_a(C')\ =\ p_a(C)\ -\ \tfrac{1}{2}a(a-1)\ =\ 
1588:     p_a(C)\ -\ \tfrac{1}{2}\bigl((a-1)(a-1) + a -1\bigr)\,,
1589: \]
1590: which establishes the lemma in this case.
1591: 
1592: We complete the proof by induction on the maximum of the
1593: exponents of $x$ and $y$.
1594: Suppose that $a<b$.
1595: Then $C$ is tangent to the curve $y=0$ at $p$ and so to compute the blowup $C'$, we 
1596: substitute $x=xy$ in~\eqref{E:binomial_germ} to obtain
1597: \[
1598:    y^a \; (x^a-y^{b-a})\,\cdot\,u(xy,y)\,.
1599: \]
1600: The exceptional divisor ($y=0$) has multiplicity $a$, and the curve $C'$ has local
1601: equation $(x^a-y^{b-a}) u'$, where $u'(x,y)=u(xy,y)$ and so $u'(0,0)\neq 0$.
1602: Since $a,b-a<b=\max\{a,b\}$, our induction hypothesis applies to $C'$ to give
1603: \[
1604:    p_a(\widetilde{C})\ =\ p_a(C')\ -\ 
1605:    \tfrac{1}{2}\bigl( (a-1)(b-a-1)+\gcd(a,b-a) -1  \bigr)\,.
1606: \]
1607: Using~\eqref{E:genus_contribution}, this becomes
1608: %
1609:  \begin{eqnarray*}
1610:    p_a(\widetilde{C})&=& p_a(C)\ -\ \tfrac{1}{2}a(a-1)\ -\ 
1611:    \tfrac{1}{2}\bigl( (a-1)(b-a-1)+\gcd(a,b-a) - 1  \bigr)\\
1612:     &=&  p_a(C)\ -\ \tfrac{1}{2}\bigl( (a-1)(b-1) + \gcd(a,b)-1  \bigr)\,,
1613:  \end{eqnarray*}
1614: %
1615:  which completes the proof.
1616: \end{proof}
1617: 
1618: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1619: %
1620: \subsection{Classification}
1621: %
1622: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1623: 
1624: Suppose that $\gamma\colon\P^1\to\P^2$ parameterizes the curve $F=0$.
1625: We may assume that the three linear forms dividing components of $\gamma$ are $s$, $t$,
1626: and $\Blue{\ell}:=-(s+t)$. 
1627: Since the components are relatively prime forms of degree $d$, there are seven
1628: possibilities for $\gamma$, up to permuting components and factors.\medskip 
1629: 
1630: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1631: \begin{tabular}{rlcrlcrl}
1632:  I   & $[s^at^{d-a} : t^b\ell^{d-b} : s^{d-c}\ell^c]$ & \ &
1633:  II  & $[s^at^b\ell^{d-a-b} :  s^{d-c}t^c :  \ell^d]$  & \ &
1634:  III & $[s^at^{d-a} : t^b\ell^{d-b} : \ell^d]$   \\
1635:  IV  & $[s^at^{d-a} : s^{d-b}t^b : \ell^d]$       & \ &
1636:  V   & $[s^at^b\ell^{d-a-b} :  t^d :  \ell^d]$ & \ &
1637:  VI  & $[s^at^{d-a} : t^d : \ell^d]$        \\
1638:  VII & $[s^d : t^d : \ell^d]$\smallskip
1639: \end{tabular}
1640: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1641: 
1642: We assume that all exponents appearing here are positive.
1643: 
1644: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1645: \begin{theorem}\label{T:gamma_classification}
1646:  Suppose that $\gamma$ is a curve with parameterization one of the types {\rm I}---{\rm VII}.
1647:  \begin{enumerate}
1648:   \item If $\gamma$ has type {\rm I} and is smooth outside the coordinate lines, then
1649:          $\gamma$ is equivalent to either
1650: \[
1651:     [s^2t^2\,:\, t^2\ell^2\,:\,  s^2\ell^2]
1652:     \qquad\mbox{or}\qquad
1653:     [s^{d-1}t\,:\, t^{d-1}\ell\,:\, s^{d-1}\ell]\,.
1654: \]
1655: %
1656:   \item If $\gamma$ does not have type {\rm I}, then it may be transformed into a curve of
1657:         type {\rm I} via quadratic Cremona transformations. 
1658: 
1659: \end{enumerate}
1660: \end{theorem}
1661: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1662: 
1663: We deduce Theorem~\ref{T:not_contracted} from
1664: Theorem~\ref{T:gamma_classification}.
1665: 
1666: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1667: \begin{proof}[Proof of Theorem~$\ref{T:not_contracted}$]
1668: Suppose that $\gamma$ has the first form in Theorem~\ref{T:gamma_classification}(1).
1669: Apply the standard Cremona transformation  $[x:y:z]\mapsto [yz:xz:xy]$ to $\gamma$  and
1670: remove the common factor $s^2t^2\ell^2$ to obtain
1671: \[
1672:    [s^2t^2\ell^4 \,:\, s^4t^2\ell^2 \,:\, s^2t^4\ell^2]\ =\ 
1673:    [\ell^2\,:\, s^2\,:\, t^2]\,,
1674: \]
1675: which satisfies $x^2+y^2+z^2-2(xy+xz+yz)=0$, the curve in 
1676: Theorem~\ref{T:not_contracted}(1).
1677: 
1678: Suppose that $\gamma$ has the second form in Theorem~\ref{T:gamma_classification}(1).
1679: Set $a:=d-1$ to obtain $[s^at:t^a\ell:s^a\ell]$.
1680: If $a=0$, this parameterizes the line $x+y+z=0$.
1681: If $a>0$, apply the standard Cremona transformation and 
1682: multiply the $y$-coordinate by $(-1)^a$ to obtain
1683: \[
1684:   [s^at^a\ell^2 : (-1)^as^{2a}t\ell : s^at^{a+1}\ell]\ =\ 
1685:   [t^{a-1}\ell : -(-s)^a : t^a]\,.
1686: \]
1687: Since $\ell=-(s+t)$, we have
1688: \[
1689:    (x+z)^a\ =\ (-st^{a-1})^a\ =\ (-s)^a (t^a)^{a-1}\ =\ -yz^{a-1}\,.
1690: \]
1691: This gives all curves of the form in Theorem~\ref{T:not_contracted}(2).
1692: \end{proof}
1693: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1694: 
1695: We prove Theorem~\ref{T:gamma_classification} in the
1696: following subsections. 
1697: 
1698: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1699: %
1700: %
1701: \subsection{Curves of type I}
1702: 
1703: 
1704: Suppose that $\gamma=[s^at^{d-a}:t^b\ell^{d-b}: s^{d-c}\ell^c]$ is a rational curve of type I.
1705: If $\gamma$ parameterizes a curve satisfying the hypotheses of
1706: Theorem~\ref{T:not_contracted}, then it can be singular only on the coordinate lines.
1707: Since all six exponents appearing in $\gamma$ are positive, these singularities 
1708: occur at the coordinate points.
1709: As $\gamma$ is rational, its arithmetic genus
1710: must equal the sum of its $\delta$-invariants at these singular points. 
1711: 
1712: In the neighborhood of the coordinate point $[0:0:1]$, the curve has germ 
1713: $(x^b-y^{d-a})u=0$, where $u(0,0)\neq 0$.
1714: By Lemma~\ref{L:binomial_germ} the $\delta$-invariant is
1715: \[
1716:    \tfrac{1}{2}\bigl( (b-1)(d-a-1) + \gcd(b,d-a) - 1  \bigr)\,.
1717: \]
1718: A similar formula holds at the other points $[1:0:0]$ and $[0:1:0]$.
1719: Summing these, equating with $p_a(C)=\binom{d-1}{2}$, and multiplying by $2$, we
1720: obtain  
1721: %
1722:  \begin{multline}\label{Eq:delta-invariants}
1723:  \qquad (d-1)(d-2)\ =\  d(a+b+c-3) -(ab+ac+bc) \\ + \gcd(a,d-c) 
1724:     + \gcd(b,d-a) + \gcd(c,d-b) \,. \qquad 
1725:  \end{multline}
1726: %
1727: 
1728: We may assume that the coordinates and forms $s,t,\ell$ have been chosen so that 
1729: $a$ is the maximum exponent and thus $d-a$ is the minimum.
1730: We have $a\geq d-c$ and $b\geq d-a$, and there are two cases to consider
1731: %
1732:  \begin{equation}\label{Eq:alternative}
1733:     c\geq d-b \qquad\mbox{or}\qquad d-b\geq c\,.
1734:  \end{equation}
1735: %
1736: We study each case separately, beginning with $c\geq d-b$.
1737: 
1738: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1739: \begin{proposition}\label{P:solutions}
1740:  The solutions to~$\eqref{Eq:delta-invariants}$ in the polytope \Blue{$P$}
1741:  defined by the inequalities
1742: %
1743:  \begin{eqnarray}
1744:    d-1 &\geq\ a\ \geq& d-c\nonumber\\
1745:    d-1 &\geq\ b\ \geq& d-a\label{Eq:ineqs}\\
1746:    d-1 &\geq\ c\ \geq& d-b\nonumber
1747:  \end{eqnarray}
1748: %
1749:  are $(d{-}1,d{-}1,1)$, $(d{-}1,1,d{-}1)$, and $(1,d{-}1,d{-}1)$, for any $d\geq 2$, and 
1750:  $(2,2,2)$ when $d=4$.
1751: \end{proposition}
1752: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1753: 
1754: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1755: \begin{proof}
1756:  Since $\gcd(\alpha,\beta)\geq\min(\alpha,\beta)$, for positive integers $\alpha$ and
1757:  $\beta$, we may simplify~\eqref{Eq:delta-invariants} to obtain the inequality
1758:  % (a-1)*(d-c-1)+(b-1)*(d-a-1)+(c-1)*(d-b-1) +(d-c+d-b+d-a) -3;
1759:  % ad - a  +  b d - b  +  c d - c     - a c - b a - c b
1760: %
1761: \[
1762:    (d-1)(d-2)\ \geq\ 
1763:     (d-1)(a+b+c) -(ab+ac+bc) \,.
1764: \]
1765: %
1766:  Let $\Blue{Q}=Q(a,b,c)$ be the symmetric quadratic form defined by the right hand side of
1767:  this inequality. 
1768:  We find its maximum values on the polytope $P$.
1769:  First, the Hessian of $Q$ is
1770: \[
1771:    \mbox{hess}(Q)\ =\ 
1772:     \left(\begin{matrix}0&-1&-1\\-1&0&-1\\-1&-1&0\end{matrix}\right)\ .
1773: \]
1774:  This has negative eigenvalue $-2$ with eigenvector $(1,1,1)$
1775:  and positive eigenvalue $1$ with two-dimensional eigenspace $a+b+c=0$.
1776:  In particular, $Q$ cannot take a maximum value in the interior of the polytope $P$ or in
1777:  any of its facets.
1778:  It can only take a maximum value in an edge that is parallel to the negative eigenspace
1779:  $(1,1,1)$.
1780: 
1781:  The polytope $P$ is a symmetric bipyramid over the triangle whose vertices are
1782: %
1783:  \begin{equation}\label{Eq:triangle}
1784:    (d{-}1,\,d{-}1,\,    1)\,,\ 
1785:    (d{-}1,\,    1,\,d{-}1)\,,\ 
1786:    (    1,\,d{-}1,\,d{-}1)\,.
1787:  \end{equation}
1788: %
1789:  and with apices $(d{-}1,d{-}1,d{-}1)$ and $(\frac{d}{2},\frac{d}{2},\frac{d}{2})$.
1790:  Since $P$ has no edge parallel to the negative eigenspace, $Q$ takes its maximum value 
1791:  at vertices of $P$.
1792: 
1793:  The form $Q$ takes value $0$ at $(d{-}1,d{-}1,d{-}1)$, 
1794:  $\frac{3}{4}d(d-2)$ at $(\frac{d}{2},\frac{d}{2},\frac{d}{2})$, and 
1795:  $(d-1)(d-2)$ at the vertices~\eqref{Eq:triangle} of the triangle, and so the 
1796:  vertices of the triangle give solutions.
1797:  When $d=2$, $P$ degenerates to a point $(1,1,1)$, which is a solution
1798:  to~\eqref{Eq:delta-invariants}.
1799:  The only remaining possibility is that the point $(\frac{d}{2},\frac{d}{2},\frac{d}{2})$
1800:  satisfies~\eqref{Eq:delta-invariants}.
1801:  But then 
1802: \[
1803:     (d-1)(d-2)\ =\ \tfrac{3}{4}d(d-2)\,,
1804: \]
1805: in which case $d=4$ and so $(a,b,c)=(2,2,2)$ is the only other solution.
1806: \end{proof}
1807: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1808: 
1809: If we take the alternative inequality in~\eqref{Eq:alternative}, 
1810: $d-b\geq c$, then $Q$ becomes 
1811: % (a-1)*(d-c-1)+(b-1)*(d-a-1)+(c-1)*(d-b-1) +(d-c+c+d-a) -3;
1812: \[
1813:    d(a+b+c-1) -a -ab-ac-bc\,.
1814: \]
1815: Replacing the third inequality in~\eqref{Eq:ineqs} by $d{-}b\geq c\geq 1$ defines a
1816: tetrahedron with vertices 
1817: \[
1818:    (d{-}1,\, d{-}1,\, 1),\  (d{-}1,\,1,\,d{-}1),\ 
1819:    (d{-}1,\, 1,\, 1),\ (\tfrac{d}{2},\tfrac{d}{2},\tfrac{d}{2})\,.
1820: \]
1821: Similar arguments as in the proof of Proposition~\ref{P:solutions} give the additional
1822: solution $(d{-}1,1,1)$ to~\eqref{Eq:delta-invariants}.
1823: By symmetry, we also obtain  solutions $(1,d{-}1,1)$ and $(1,1,d{-}1)$.\medskip
1824: 
1825: 
1826: We write the curves $\gamma$ corresponding to these solutions.
1827: The solution $(a,b,c)=(2,2,2)$ gives the expression  
1828: \[
1829:     [s^2t^2\,:\, t^2\ell^2\,:\, s^2\ell^2]\,,
1830: \]
1831: for $\gamma$ and the solutions $(a,b,c)=(d{-}1,d{-}1,1)$ give the expressions
1832: %
1833:  \begin{equation}\label{E:forms}
1834:     [s^{d-1}t\,:\, t^{d-1}\ell\,:\, s^{d-1}\ell]\,.
1835:  \end{equation}
1836: %
1837: The other two symmetric solutions give equivalent curves.
1838: Lastly, the solutions $(a,b,c)=(d{-}1,1,1)$ give the expressions
1839: \[
1840:     [s^{d-1}t\,:\, t\ell^{d-1}\,:\, s^{d-1}\ell]\,,
1841: \]
1842: which become the expressions~\eqref{E:forms} under 
1843: $x\leftrightarrow z$ and $t\leftrightarrow \ell$.
1844: 
1845: 
1846: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1847: %
1848: %
1849: \subsection{Reduction to curves of type I}
1850: 
1851: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1852: %
1853: %
1854: \subsubsection{Quadratic Cremona transformations} 
1855: We will show how curves of types II---VII are equivalent to curves of type I through
1856: quadratic Cremona transformations.
1857: We will sometimes use the non standard quadratic Cremona transformation
1858: \[
1859:    [x\colon y\colon z]\ \longmapsto  [z^2\colon xz\colon xy]\,.
1860: \]
1861: Permuting the variables gives five other non standard quadratic Cremona transformations.
1862: 
1863: 
1864: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1865: %
1866: %
1867: \subsubsection{Curves of type II}
1868: Suppose that $\gamma=[s^at^b\ell^{d-a-b}:s^{d-c}t^c:\ell^d]$ has type II.
1869: We show that this may be transformed into a curve of type I by induction on $d$.
1870: We will either transform $\gamma$ into a curve of type I or one of type II of lower
1871: degree. 
1872: Since $a+b<d=(d-c)+c$, interchanging $s$ and $t$ if necessary, we may assume that
1873: $b<c$.
1874: Applying the standard Cremona $[xy:xz:yz]$ transformation and removing the common factor of
1875: $t^b\ell^{d-a-b}$ gives
1876: \[
1877:    [s^{a+d-c}t^{b+c}\ell^{d-a-b}\,:\, s^at^b\ell^{2d-a-b}
1878:     \,:\, s^{d-c}t^c\ell^d]\ =\ 
1879:    [s^{a+d-c}t^c\,:\, s^a\ell^d\,:\, s^{d-c}t^{c-b}\ell^{a+b}]\,.
1880: \]
1881: There remains a common power of $s$ that we can remove.
1882: There are three cases to consider.
1883: 
1884: \begin{enumerate}
1885:  \item If $a>d-c$, then we remove the common factor of $s^{d-c}$ to obtain a curve of
1886:        type I.
1887:  \item If $a=d-c$, then $d=a+c$ and we remove the common factor of $s^a=s^{d-c}$ to
1888:        obtain 
1889: %
1890:  \[
1891:       [s^at^c\,:\, \ell^{a+c}\,:\, t^{c-b}\ell^{a+b}] \ =\ [x\,:\,y\,:\,z]\,.
1892:  \]
1893: %
1894:       We now apply the non standard Cremona transformation $[z^2:xz:xy]$ and remove the common
1895:       factor $t^{c-b}\ell^{a+b}$ to obtain
1896: %
1897:  \begin{equation}\label{Eq:one_case_of_many}
1898:       [t^{2c-2b}\ell^{2a+2b}\,:\, s^at^{2c-b}\ell^{a+b}\,:\, s^at^c\ell^{a+c}]
1899:        \ =\ 
1900:       [t^{c-b}\ell^{a+b}\,:\, s^at^c\,:\,s^at^b\ell^{c-b}]\,.
1901:  \end{equation}
1902: %
1903:     If $b<c-b$, then we remove the common factor $t^b$ to obtain a curve of type I.
1904:     If $b=c-b$, then we remove $t^b$ to get
1905: %
1906:  \[
1907:      [\ell^{a+b}\,:\, s^at^b\,:\, s^a\ell^b]\ =\ [x\,:\,y\,:\,z]\,.
1908:  \]
1909: %
1910:    Applying the non standard Cremona transformation $[z^2\,:\,xy\,:\,yz]$ to get
1911: %
1912:  \[
1913:     [s^{2a}\ell^{2b}\,:\, s^at^b\ell^{a+b}\,:\, s^{2a}t^b\ell^b]\ =\ 
1914:     [s^a\ell^b\,:\, t^b\ell^a\,:\, s^at^b]\,,
1915:  \]
1916: %
1917:   which has type I.
1918:   Finally, if $b>c-b$, then we remove the common factor of $t^{c-b}$
1919:   from~\eqref{Eq:one_case_of_many} to obtain
1920: %
1921:  \[
1922:      [\ell^{a+b}\,:\, s^at^b\,:\,s^at^{2b-c}\ell^{c-b}]\,,
1923:  \]
1924: %
1925:   which has type II and degree $a+b<a+c=d$.
1926: 
1927:  \item If $a<d-c$, then we instead apply the non standard Cremona transformation
1928:     $[x^2 :  yz : xz]$ to $\gamma$ and remove the common factor $s^at^b\ell^{d-a-b}$
1929:        to obtain 
1930: %
1931:   \[ 
1932:      [s^at^b\ell^{d-a-b}\,:\, s^{d-c-a}t^{c-b}\ell^{a+b}\,:\, \ell^d]\,.
1933:   \]
1934: %
1935:     Removing the final common factor $\ell^{\min\{d-a-b,a+b\}}$ gives another curve of
1936:     type II, but of lower degree.
1937: \end{enumerate}
1938: 
1939: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1940: %
1941: %
1942: \subsubsection{Curves of type III}
1943: Suppose that $\gamma=[s^at^{d-a}:t^b\ell^{d-b}:\ell^d]$ has type III.
1944: We apply the standard Cremona transformation $[xy:xz:yz]$ and remove the common factor $\ell^{d-b}$ to
1945: obtain 
1946: \[
1947:    [s^a t^{b+d-a}\,:\, s^at^{d-a}\ell^b\,:\,t^b\ell^d]\,.
1948: \]
1949: There remains a common power of $t$ that we can remove.
1950: There are three cases to consider.
1951: 
1952: \begin{enumerate}
1953:  \item If $b>d-a$, we factor out $t^{d-a}$ to get the type I curve, 
1954: \[
1955:    [s^at^b\,:\, s^a\ell^b\,:\, t^{a+b-d}\ell^d]\,.
1956: \]
1957: 
1958:  \item If $b=d-a$, we instead apply the non standard Cremona transformation 
1959:        $[y^2 : xy : xz]$ to $\gamma$ and factor out $t^b\ell^a$ to  get the type I
1960:        curve, 
1961: \[
1962:       [t^b\ell^a\,:\, s^at^b\,:\, s^a\ell^b]\,.
1963: \]
1964: 
1965:  \item If $b<d-a$, we factor out $t^b$ to obtain
1966: \[
1967:     [s^at^{d-a}\,:\, s^at^{d-a-b}\ell^b\,:\, \ell^d]\,,
1968: \]
1969:     which has type II, and we have already shown how to reduce a curve of type II to a
1970:     curve of type I.
1971: \end{enumerate}
1972: 
1973: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1974: %
1975: %
1976: \subsubsection{Curves of type IV}
1977: Suppose that $\gamma=[s^at^{d-a}:s^{d-b}t^b:\ell^d]$ has type IV.
1978: We may assume that $a>d-b$ and thus $b>d-a$.
1979: We apply the standard Cremona transformation $[yz:xz:xy]$ and remove the common factor of
1980: $s^{d-b}t^{d-a}$, to get the type I curve, 
1981: \[
1982:    [t^{a+b-d}\ell^d\,:\,s^{a+b-d}\ell^d\,:\, s^at^b]\,.
1983: \]
1984: 
1985: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1986: %
1987: %
1988: \subsubsection{Curves of type V}
1989: Suppose that $\gamma=[s^at^b\ell^{d-a-b}:t^d:\ell^d]$ has type V.
1990: If we apply the standard Cremona transformation $[yz:xz:xy]$, we get the type I curve,
1991: \[
1992:    [t^d\ell^d\,:\,s^at^b\ell^{2d-a-b}\,:\, s^at^{d+b}\ell^{d-a-b}]\ =\ 
1993:    [t^{d-b}\ell^{a+b}\,:\, s^a\ell^d\,:\, s^at^d]\,.
1994: \]
1995: 
1996: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1997: %
1998: %
1999: \subsubsection{Curves of type VI}
2000: 
2001: Suppose that $\gamma=[s^at^{d-a}:t^d:\ell^d]$ has type VI.
2002: If we apply the standard Cremona transformation $[yz:xz:xy]$, we get the type I curve,
2003: \[
2004:    [t^d\ell^d\,:\,s^at^{d-a}\ell^d\,:\, s^at^{2d-a}]\ =\ 
2005:    [t^a\ell^d\,:\, s^a\ell^d\,:\, s^at^d]\,.
2006: \]
2007: 
2008: 
2009: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2010: %
2011: %
2012: \subsubsection{Curves of type VII}
2013: 
2014: Suppose that $\gamma=[s^d:t^d:\ell^d]$ has type VII. 
2015: If we apply the standard Cremona transformation, we get the type I curve,
2016: \[
2017:    [t^d\ell^d\,:\,s^d\ell^d\,:\, s^dt^d]\,.
2018: \]
2019: 
2020: 
2021: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2022: %
2023: \section{Singularities of polynomials}\label{S:technical}
2024: %
2025: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2026: 
2027: Fix a form $F$ with prime factorization 
2028: $F=F_{1}^{n_{1}}F_{2}^{n_{2}}\cdots F_k^{n_k}$.
2029: Write $\sqrt{F}=F_{1}F_{2}\cdots F_{r}$ for its square free part, the product of its prime
2030: factors.  In the following theorem we do not distinguish between the
2031: curves $F=0$ and $\sqrt {F}=0$.\medskip 
2032: 
2033: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2034: \noindent{\bf Theorem~\ref{Th:singularities}.}
2035: {\it 
2036:   If a form $F$ coprime to $xyz$ defines a toric polar
2037:   Cremona transformation, then the curve $F=0$ has at most one singular point
2038:   outside the coordinate lines, and if there is such a point, then all factors of $F$ are
2039:   contracted. Furthermore, if this singular point is ordinary,
2040:  then $F$ has at most two distinct factors through the singular point.
2041: }\medskip
2042: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2043: 
2044: We prove Theorem~\ref{Th:singularities} by studying the resolution of base points of the
2045: toric polar linear system $T(F)$ at a singular point $p$ on $\sqrt {F}=0$ not lying on the
2046: coordinate lines.  
2047: In the resolution, there is a tree of exceptional rational curves lying over $p$.
2048: We show that the leaves of this tree are exceptional curves above $p$
2049: that are not contracted by the lift of the 
2050: toric polar map, but are components of the lift of $\sqrt{F}=0$.
2051: This implies that there is at most one such leaf and its exceptional
2052: curve has multipicity 1, and that all other
2053: components of this lift, including the strict transforms of 
2054: the components of $\sqrt{F}=0$, are contracted by the toric polar map.
2055: Thus there is at most one such singular point, and if it is ordinary,
2056: then $F$ has two branches at this point.
2057: 
2058: Any common factor in $T(F)=\langle xF_{x},yF_{y}, zF_{z} \rangle$ is a multiple component of $F$.   
2059: Indeed, if
2060: $F=F_{1}^{n_{1}}F_{2}^{n_{2}}\cdots F_{r}^{n_{r}}$ is the prime factorization of $F$, 
2061: then 
2062: \[
2063:    \Blue{G}\ :=\  \gcd(xF_x,yF_y,zF_z)\ =\ 
2064:     F_{1}^{n_{1}-1}F_{2}^{n_{2}-1}\cdots F_{r}^{n_{r-1}}\,.
2065: \]
2066: is a common factor of $xF_{x},yF_{y}, zF_{z}$.
2067: Removing this factor we get the vector space 
2068: %
2069:  \begin{equation}\label{Eq:RLS}
2070:    \Blue{\sqrt{T(F)}}\ :=\ \langle F^x,F^y,F^z\rangle\ :=\ 
2071:    \left\langle \frac{xF_x}{G},\frac{yF_y}{G},\frac{xF_x}{G}\right\rangle,
2072:  \end{equation}
2073: %
2074: where $\Blue{F^{x}}:=n_{1}xF_{1,x}\dotsb F_{r}+\dotsb+n_{r}xF_{1}\dotsb F_{r,x}$,
2075: and the same for $F^y$ and $F^z$.
2076: Notice that 
2077: \[
2078:    F^x+F^y+F^z\ =\ (n_{1}\deg(F_1)+\dotsb+n_{r}\deg(F_r))F_1 \dotsb F_r
2079:    \ =\ \deg(F)\sqrt{F}\,.
2080: \]
2081: In particular any common factor for form in $\sqrt{T(F)}$ is a factor of $\sqrt{F}$
2082: and is thus one of the $F_{i}$. 
2083: But no $F_{i}$ is a common factor, so forms in $\sqrt{T(F)}$ are coprime.
2084: 
2085: %
2086: %   We do not use T(sqrt(F)), so why mention it?
2087: %
2088: %Notice that $\sqrt{T(F)}$ is in general different from $T(\sqrt{F})$,
2089: %although they both contain the curve $\sqrt{F}=0$. 
2090: 
2091: Let $p$ be a multiple point of the curve $\sqrt{F}=0$ outside the coordinate lines.  It is  
2092: a common zero of the forms in $F^x,F^y,F^z$, as well as all partial
2093: derivatives of $\sqrt{F}$, and is therefore a base point for
2094:  $\sqrt{T(F)}$.
2095: Resolving this base point and possibly infinitely near base points
2096: gives a tree of exceptional rational curves lying over $p$.
2097: We are only concerned with leaves of this tree so we assume that
2098: $p=p_{0}, p_{1},\dotsc, p_{r}$ are
2099: successive infinitely near base points that we blow up   
2100: to resolve the base locus of $\sqrt{L_{F}}$ lying over $p$.  
2101: In particular we assume that $p_{1}$ lies on the exceptional curve over $p$, the point
2102: $p_{2}$ lies on the exceptional curve over $p_{1}$, and etc.  and that there are no base points infinitely near to $p_{r}$.  
2103: Thus there is a unibranched curve that is smooth at $p_r$ and passes through all these 
2104: infinitely near points. 
2105: These are some, but not necessarily all of the infinitely near base points at
2106: $p$.
2107: 
2108: We denote by $\pi_{1}\colon S_{1}\to S_{0}=\P^2$ 
2109: the blow up of the point $p_{0}$, and by $E_{0}$ the exceptional curve of this map.
2110: Inductively we denote by $\pi_{i}:S_{i}\to S_{i-1}$ the blowup of the point $p_{i-1}\in
2111: S_{i-1}$ and by $E_{i-1}$ the exceptional curve of this map. 
2112: Write $E_{i}$ also for the total transform in $S_k$ for $k>i$ of the exceptional curve
2113: $E_i$ of $\pi_{i+1} \colon S_{i+1}\to S_{i}$.
2114: The map $\pi:S_{r+1}\to \P^2$ is then the composition of the blowups $\pi_{i}$
2115: for $i=1,\dotsc,r+1$.   
2116: 
2117: Let $\mu_0(\sqrt{T(F)})$ be the minimal multiplicity at $p$ of a curve in $\sqrt{T(F)}$.
2118: Then the linear system $\sqrt{T(F)}_{(1)}$  on $S_1$ is generated by the strict transforms of curves
2119: in $\sqrt{T(F)}$ having multiplicity $\mu_0(\sqrt{T(F)})$ at $p$.
2120: Set $\mu_1(\sqrt{T(F)})$  to be the minimal multiplicity at $p_1$ of a curve
2121: in $\sqrt{T(F)}_{(1)}$.
2122: Then the linear system $\sqrt{T(F)}_{(2)}$  on $S_2$ is generated by the strict transforms of
2123: curves in $\sqrt{T(F)}_{(1)}$ having multiplicity $\mu_1(\sqrt{T(F)})$ at $p_1$. 
2124: Inductively, we obtain linear systems $\sqrt{T(F)}_{(i)}$ with multiplicities
2125: $\mu_i(\sqrt{T(F)})$ at $p_i$.
2126: 
2127: For any curve $C$ in $\sqrt{T(F)}$, define the 
2128: \Blue{{\sl virtual transform} $C_{(i)}$} in $\sqrt{T(F)}_{(i)}$ for $i=1,...,r$, to be unique member of 
2129: this linear system that is mapped by $\pi_{1}\circ\dotsb\circ\pi_{i}$ to $C$ in $\P^2$.
2130: Thus the virtual transform $C_{(i)}$ of $C$ on $S_{i}$ is the sum of the strict
2131: transform of $C_{(i-1)}$ and $\bigl(\mu_{i-1}(C_{(i-1)})-\mu_{i-1}(\sqrt{T(F)})\bigr)E_{i-1}$, where
2132: $\mu_{i-1}(C_{(i-1)})$ is the multiplicity of $C_{(i-1)}$ at $p_{i-1}$. 
2133: 
2134: We consider the virtual transform $\sqrt{F}_{(r+1)}$ in
2135: $\sqrt{T(F)}_{(r+1)}$, and claim that it contains the leaf $E_{r}$ as a component.
2136: For this we follow the line of argument in \cite {CA}, Section 8.5. We 
2137: compare multiplicities and show that the inequality
2138: \[
2139:   \mu_{r}\bigl(\sqrt{F}_{(r)}\bigr)\ \geq\  \mu_{r}\bigl(\sqrt{T(F)}\bigr)
2140: \]
2141: is strict.
2142: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2143:  We reduce this to a local calculation at $p$. 
2144:  First, let $L_P$ be the polar linear system of $F$ defined by the partial derivatives
2145:  $\langle F_x,F_y,F_z \rangle$. 
2146:  Let $\sqrt{L_{P}}$ be the linear system obtained by removing the fixed component of
2147:  $L_{P}$. 
2148:  By linearity, $F_{\ell}=aF_{x}+bF_{y}+cF_{z}$ is the partial derivative of $F$ with
2149:  respect to the linear form $\ell=ax+by+cz$. 
2150:  In the Euler relation $dF=xF_{x}+yF_{y}+zF_{z}$ locally at $p$, the
2151:  coordinates $x,y,z$ are units.  
2152:  Therefore, locally at $p$, a general form in the toric polar linear system 
2153:  $\langle xF_{x},yF_{y},zF_{z}\rangle$  is a linear combination of $F$ and its partial
2154:  derivative $F_{\ell}$ with respect to some linear form $\ell$ that vanishes at $p$.
2155:  In particular, such a general form has the same multiplicities as $F_{\ell}$ 
2156:  (compare~\cite {CA} Section 7.2 and in particular Remark 7.2.4).
2157:  So we may compute  $\mu_{i}(T(F))$ and $\mu_{i}(\sqrt{T(F)})$ locally at $p$, replacing
2158:  $F=0$  and $\sqrt{F}=0$ with their germs $f$ and $\sqrt{f}$ at $p$,  and considering
2159:  their partials derivatives (polars) with respect to linear forms that vanish at $p$. 
2160:  In particular $\mu_{i}(T(F))= \mu_{i}(f_{\ell})$ for a general polar $f_{\ell}$ with
2161:  respect to a linear form $\ell$ that vanishes at $p$.  
2162:  We also change coordinates so that $x,y$ are local coordinates at $p$, and
2163:  let $f_{x}, f_{y}$ be the germs of the polars with respect to $x$ and $y$.  
2164: 
2165:  We now analyze these germs.  
2166:  For any two germs $g,\gamma$ of curves at $p$, we write \Blue{$[g,\gamma]_{p}$} for their local
2167:  intersection multiplicity at $p$. 
2168:  If $\gamma$ is unibranched, then this is simply the order of vanishing of the pullback of
2169:  $g$ along a local parameterization of $\gamma$.
2170:  Let $f=f_{1}^{n_{1}}f_{2}^{n_{2}}\cdots f_{k}^{n_{k}}$ be the irreducible factorization of
2171:  the germ $f$ of the curve $F=0$ at $p$.  
2172:  Set $g=f_{1}^{n_{1}-1}f_{2}^{n_{2}-1}\cdots f_{k}^{n_{k}-1}$ and let  
2173:  $\sqrt{f}=f_{1}f_{2}\cdots f_{r}=\frac fg$, and $\overline{f_{x}}=\frac {f_{x}}g$ and
2174:  $\overline{f_{y}}=\frac {f_{y}}g$. 
2175: 
2176: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2177: \begin{lemma}\label{L:germ_mult}
2178:  Let $f$ be the germ of the curve $F=0$ at $p$, and let $\gamma$ be any smooth germ
2179:  through $p$. 
2180:  Then we have $[f,\gamma]_{p}> \min\{[f_{x},\gamma]_{p},[f_{y},\gamma]_{p}\}$.  
2181:  Furthermore, 
2182: \[
2183:    [\sqrt{f}, \gamma]_{p}\ >\ 
2184:    \min\{[\overline{f_{x}},\gamma]_{p},[\overline{f_{y}},\gamma]_{p}\}\,.
2185: \]
2186: \end{lemma}
2187: 
2188: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2189: 
2190: 
2191: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2192: \begin{proof}
2193: Let $t\mapsto (x(t), y(t))$ be a local parameterization of $\gamma$.  
2194: If $[f,\gamma]_{p}=n$, then $f(x(t),y(t))=t^nu$ for some invertible series $u$.  
2195: Taking the derivative, we have 
2196: \[
2197:     nt^{n-1}u+t^n \frac{du}{dt}\ =\ 
2198:     f_x(x(t),y(t))\frac{dx}{dt} + f_y(x(t),y(t))\frac{dy}{dt}
2199: \]
2200: so 
2201: \[
2202:     n-1\ \geq\  \min\{[f_{x},\gamma]_{p},[f_{y},\gamma]_{p}\}\,.
2203: \]
2204: Now, if $m_{i}=[f_{i}, \gamma]_{p}$, then $n=\sum_{i=1}^rm_{i}n_{i}$ and 
2205: %
2206:  \begin{eqnarray*}
2207:    f_x(x(t),y(t))&=& n_{1}f_{1,x}f_{1}^{n_{1}-1}f_{2}^{n_{2}}\dotsb
2208:    f_{k}^{n_{k}}+\dotsb+n_{k}f_{1}^{n_{1}}f_{2}^{n_{2}}\dotsb f_{k}^{n_{k}-1}f_{k,x}\\
2209:   &=& 
2210:    f_{1}^{n_{1}-1}f_{2}^{n_{2}-1}\dotsb f_{k}^{n_{k}-1}(n_{1}f_{1,x}f_{2}\dotsb
2211:    f_{k}+\dotsb+n_{k}f_{1}f_{2}\dotsb f_{k,x})\,.
2212:  \end{eqnarray*}
2213: %
2214: Similarly for $f_{y}$, so
2215: \[
2216:    n-1\ \ge\ \min\{[f_{x},\gamma]_{p},[f_{y},\gamma]_{p}\}\ =\ \sum_{i=1}^k m_{i}(n_{i}-1)
2217:      +\min\{[\overline{f_{x}},\gamma]_{p},[\overline{f_{y}},\gamma]_{p}\}\,.
2218: \]
2219:  Therefore the strict inequality 
2220:  $[\sqrt{f}, \gamma]_{p}> \min\{[\overline{f_{x}},\gamma]_{p},[\overline{f_{y}},\gamma]_{p}\}$ also holds
2221:  at $p$.   
2222: \end{proof}
2223: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2224: 
2225: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2226: \begin{lemma}\label{L:inequality}
2227: Let $L= \langle g,h \rangle$ be a linear system of germs of curves on
2228: a smooth surface $S$ and assume that $p=p_{0},...,p_{r}$ is a sequence
2229: of infinitely near  base points for the linear system. 
2230:  Let $f$ be a germ of a curve in $L$ whose virtual transform $f_{(i)}
2231: $ in $L_{(i)}
2232: $ has multiplicity $\mu_{i}(f_{(i)}
2233: )$ at the point $p_{i}$, 
2234:  and let $\mu_{i}(L)$ be the multiplicity of the linear system $L_{(i)}
2235: $ at $p_{i}$.
2236:  Assume that  for any smooth germ $\gamma$ on $S$ through $p$ the local intersection numbers satisfy:
2237:   $$[f, \gamma]_{p}\ >\ 
2238:    \min\{[g,\gamma]_{p},[h,\gamma]_{p}\}\,.$$
2239: Then we have strict inequalities
2240:  $\mu_{i}(f_{(i)})>\mu_{i}(L)$ for each $i=0,1,\dotsc,r$.
2241: \end{lemma}
2242: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2243: 
2244: \begin{proof}
2245: The effective multiplicities at the points $p=p_{0}, p_{1},\dotsc,p_{r}$ of the strict
2246: transforms of the polar germs coincide for all but a finite number of members in the pencil
2247: $\langle g, h\rangle$. 
2248: Changing variables if necessary, we may assume that the multiplicity
2249: sequence for the germs $g$ and $h$ coincide and are equal to that of the 
2250: linear system: $\mu_{i}:=\mu_{i}(L)$ for $i=0,\dotsc,r$. 
2251: These multiplicities are by definition the virtual multiplicities of $f$ with respect to
2252: the linear system $L$.   
2253: At $p=p_{0}$ the multiplicity $\mu_{0}$  differs from the multiplicity $e_{0}(f)$ of $f$.
2254: If $\gamma$ is a smooth germ at $p$ that avoids all tangent directions of $f$, then  
2255: $[f,\gamma]=e_{0}(f)$, and by assumption,
2256: %
2257:  \[
2258:     \mu_{0}(F)\ =\ e_{0}(f)\ =\ [f,\gamma]\ >\ \min\{[g,\gamma],[h,\gamma]\}
2259:      \ =\ \mu_{0}(L)\,.
2260: \]  
2261: Inductively, consider the virtual transform $f_{(i)}$ of $f$ on $S_{i}$ and choose a
2262: unibranched germ $\gamma$ through the sequence of points $p_{0},\dotsc,p_{i}$ that is smooth
2263: at $p_{i}$ and avoids all the tangent directions of $f_{(i)}$ at $p_{i}$.  
2264: Let $e_{j}(\gamma)$ be the multiplicity of the strict transform $\gamma_{j}$ of
2265: $\gamma$ at the $p_{j}$, for $j=0,\ldots,i$. 
2266: Then, by assumption, 
2267: \[
2268:    [f,\gamma]\ >\ \min\{[g,\gamma],[h,\gamma]\}\ \geq\ \sum_{i=0}^{i-1}
2269:    \mu_{j}e_{j}(\gamma)+\mu_{i}\,.
2270: \]
2271:  On the other hand, if $e_{j}(f)$ is the multiplicity of the strict transform $f_{j}$ on
2272:  $S_{j}$ of $f$ at $p_{j}$  and $\mu_{i}(f_{(i)})$ is the multiplicity of the virtual
2273:  transform $f_{(i)}$ of $f$ at $p_{i}$, then  
2274: \[
2275:    f_{(i)}\ =\ f_{i}-\sum_{j=0}^{i-1}(\mu_{j}-e_{j}(f))E_{j}
2276: \]
2277: while $e_{j}(\gamma)=[E_{j},\gamma_{j}]$, so
2278: %
2279:  \begin{eqnarray*}
2280:    [f,\gamma] &=&\sum_{j=0}^{i-1}e_j(f) e_{j}(\gamma)+[f_{i},\gamma_{i}]\\
2281:   &=& \sum_{j=0}^{i-1} \mu_{j}e_{j}(\gamma)+
2282:    \sum_{j=0}^{i-1}(e_{j}(f)-\mu_{j})e_{j}(\gamma)+[f_{(i)},\gamma_{i}] 
2283:    +\sum_{j=0}^{i-1} (\mu_{j}-e_{j}(f))[E_{j},\gamma_{j}]\\
2284:    &=& \sum_{j=0}^{i-1}\mu_{j}e_{j}(\gamma)+\mu_{i}(f_{(i)})\,.
2285: \end{eqnarray*}
2286: % 
2287: In particular, $\mu_{i}(f_{(i)}) > \mu_{i} = \mu_{i}(L)$. 
2288: \end{proof}
2289: %%%%%%%%%%%%%%%%%%%%%%%%
2290: 
2291: Lemma~\ref{L:inequality} applied to the curve $\sqrt{F}$ in the linear system
2292: $\sqrt{T(F)}$ yields the following corollary.
2293: 
2294: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% 
2295: \begin{cor} \label{C:leaves}
2296:  Let $p$ be a multiple point of $\sqrt{F}$ outside the coordinate
2297:  lines, and let $p_{r}$ be a base point of $\sqrt{T(F)}$ infinitely
2298:  near to $p$, such that $\sqrt{T(F)}$ has no base points infinitely
2299:  near to $p_{r}$. 
2300:  Then the virtual transform of $\sqrt{F}=0$ in the linear system
2301:  $\sqrt{T(F)}_{(r+1)}$ on the blowup of $S_{r}$ at $p_{r}$ contains
2302:  the exceptional curve  $E_{r}$ as a component.
2303:  Furthermore, the restriction of the linear system  $\sqrt{T(F)}_{(r+1)}$ to the exceptional curve $E_{r}$ has 
2304:  degree equal to $\mu_{r}\bigl( \sqrt{T(F)} \bigr)$, the multiplicity of $\sqrt{T(F)}$ at $p_{r}$.
2305: \end{cor}
2306: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% 
2307: \begin{proof} 
2308:  Since the virtual multiplicity of $\sqrt{F}$ at $p_{r}$ is strictly greater than the multiplicity of the linear system by 
2309:  Lemmas \ref{L:germ_mult} and \ref{L:inequality}, the first part follows.  
2310:  The multiplicity of $\sqrt{T(F)}_{(r)}$ at $p_{r}$ is precisely the number of intersection points between
2311:  the general member of $\sqrt{T(F)}_{(r+1)}$  and the exceptional curve $E_{r}$, so the second part also follows.
2312:  \end{proof}
2313:  
2314: 
2315: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% 
2316: \begin{proof}[Proof of Theorem $\ref{Th:singularities}$]
2317: The toric polar linear system $T(F)$ is equivalent to $\sqrt{T(F)}$
2318: and $\sqrt{F}=0$ belongs to the later system. Assume that $p$ is a
2319: singular point of $\sqrt{F}=0$ outside the coordinate lines. The point
2320: $p$ is then a base point of $\sqrt{T(F)}$. The set of infinitely near
2321: base points of $\sqrt{T(F)}$ at $p$ is finite, so it has at least one
2322: point $p_{r}$ without further infinitely near base points. 
2323:  By Corollary~\ref{C:leaves}, the exceptional curve $E_{r}$ on
2324:  $S_{r+1}$ of this point is a component of the virtual transform
2325:  $\sqrt{F}_{(r+1)}$ on $S_{r+1}$.  
2326:  Since the linear system $\sqrt{T(F)}$ defines a birational map, the
2327:  restriction of its base point free lift $\sqrt{T(F)}_{(r+1)}$ to
2328:  $E_{r}$ must have degree $0$ or $1$.  But this degree is $\mu_{r}>0$,
2329:  so $\mu_{r}=1$ and $E_{r}$ must be mapped isomorphically to a line. 
2330:  Therefore all other components must be contracted and there can be no further multiple
2331:  points of $\sqrt{F}$ outside the coordinate lines.   
2332:  At an ordinary multiple point $p$ of $\sqrt{F}$ the multiplicity of
2333:  $\sqrt{T(F)}$ is one less than the multiplicity of $\sqrt{F}$, so the
2334:  last part follows. 
2335:  \end{proof} 
2336: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% 
2337: 
2338: \providecommand{\bysame}{\leavevmode\hbox to3em{\hrulefill}\thinspace}
2339: \providecommand{\MR}{\relax\ifhmode\unskip\space\fi MR }
2340: % \MRhref is called by the amsart/book/proc definition of \MR.
2341: \providecommand{\MRhref}[2]{%
2342:   \href{http://www.ams.org/mathscinet-getitem?mr=#1}{#2}
2343: }
2344: \providecommand{\href}[2]{#2}
2345: \begin{thebibliography}{10}
2346: 
2347: \bibitem{CA}
2348: E.~Casas-Alvero, \emph{Singularities of plane curves}, London Mathematical
2349:   Society Lecture Note Series, vol. 276, Cambridge University Press, 2000.
2350: 
2351: \bibitem{CLO}
2352: D.~Cox, J.~Little, and D.~O'Shea, \emph{Ideals, varieties, and algorithms},
2353:   third ed., Undergraduate Texts in Mathematics, Springer, New York, 2007.
2354: 
2355: \bibitem{Cox03}
2356: D.~Cox, \emph{What is a toric variety?}, Topics in algebraic geometry and
2357:   geometric modeling, Contemp. Math., vol. 334, Amer. Math. Soc., Providence,
2358:   RI, 2003, pp.~203--223.
2359: 
2360: \bibitem{Do00}
2361: I.~V. Dolgachev, \emph{Polar {C}remona transformations}, Michigan Math. J.
2362:   \textbf{48} (2000), 191--202, Dedicated to William Fulton on the occasion of
2363:   his 60th birthday.
2364: 
2365: \bibitem{GS}
2366: L.~ Garc\'ia-Puente and F.~Sottile, \emph{Linear precision for parametric
2367:   patches}, 2009, Advances in Computational Math., to appear.
2368: 
2369: \bibitem{Gold}
2370: R.~Goldman, \emph{Pyramid algorithms: A dynamic programming approach to curves
2371:   and surfaces for geometric modeling}, Morgan Kaufmann Publishers, Academic
2372:   Press, San Diego, 2002.
2373: 
2374: \bibitem{Hart}
2375: R.~Hartshorne, \emph{Algebraic geometry}, Springer-Verlag, New York, 1977,
2376:   Graduate Texts in Mathematics, No. 52.
2377: 
2378: \bibitem{KK00}
2379: K.~Kar{\v{c}}iauskas and R.~Krasauskas, \emph{Comparison of
2380:   different multisided patches using algebraic geometry}, Curve and Surface
2381:   Design: Saint-Malo 1999 (P.-J. Laurent, P.~Sablonniere, and L.L. Schumaker,
2382:   eds.), Vanderbilt University Press, Nashville, 2000, pp.~163--172.
2383: 
2384: \bibitem{Kr07}
2385: R.~Krasauskas, \emph{Minimal rational parametrizations of canal surfaces},
2386:   Computing \textbf{79} (2007), no.~2-4, 281--290.
2387: 
2388: \bibitem{KR02}
2389: R.~Krasauskas, \emph{Toric surface patches}, Adv. Comput. Math.
2390:   \textbf{17} (2002), no.~1-2, 89--133, Advances in geometrical algorithms and
2391:   representations.
2392: 
2393: \bibitem{Kr06}
2394: \bysame, \emph{B\'ezier patches on almost toric surfaces}, Algebraic geometry
2395:   and geometric modeling, Math. Vis., Springer, Berlin, 2006, pp.~135--150.
2396: 
2397: \bibitem{PW01}
2398: H.~Pottmann and J.~Wallner, \emph{Computational line geometry},
2399:   Mathematics and Visualization, Springer-Verlag, Berlin, 2001.
2400: 
2401: \bibitem{So03}
2402: F.~Sottile, \emph{Toric ideals, real toric varieties, and the moment map},
2403:   Topics in algebraic geometry and geometric modeling, Contemp. Math., vol.
2404:   334, Amer. Math. Soc., Providence, RI, 2003, pp.~225--240.
2405: 
2406: \bibitem{Stu}
2407: B.~Sturmfels, \emph{Solving systems of polynomial equations}, CBMS Regional
2408:   Conference Series in Mathematics, vol.~97, Amer. Math. Soc., 2002.
2409: 
2410: \end{thebibliography}
2411: 
2412: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2413: \end{document}
2414: 
2415: 
2416: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2417: \bibliographystyle{amsplain}
2418: \bibliography{bibl}
2419: 
2420: \end{document}
2421: 
2422: