nlin0505024/PC.tex
1: 
2: \documentclass[english]{iopart}
3: 
4: \usepackage{epsfig}
5: 
6: 
7: \begin{document}
8: 
9: \title[Chaotic non-attractors]
10: {\Large Post-critical set and non existence of preserved
11: meromorphic two-forms}
12:  
13: \author{ 
14: M. Bouamra$^\S$, S. Boukraa$^\dag$, S. Hassani$^\S$ and
15: J.-M. Maillard$^\ddag$}
16: \address{\S  Centre de Recherche Nucl\'eaire d'Alger, \\
17: 2 Bd. Frantz Fanon, BP 399, 16000 Alger, Algeria}
18: \address{\dag Universit\'e de Blida, Institut d'A{\'e}ronautique,
19:  Blida, Algeria}
20: \address{\ddag\ LPTMC, Universit\'e de Paris 6, Tour 24,
21:  4\`eme \'etage, case 121, \\
22:  4 Place Jussieu, 75252 Paris Cedex 05, France} 
23: \ead{maillard@lptmc.jussieu.fr, maillard@lptl.jussieu.fr, sboukraa@wissal.dz, bouamrafr@yahoo.com}
24: 
25: 
26: 
27: \begin{abstract}
28: 
29: We present a family of birational transformations 
30: in $\, CP_2$ depending on two, or three,
31: parameters which does not, generically, preserve 
32: meromorphic two-forms. With the introduction of the orbit of the critical set (vanishing condition
33: of the Jacobian), also called ``post-critical set'', we get some
34: new structures, some "non-analytic" two-form which reduce to meromorphic
35: two-forms for particular subvarieties in the parameter space.
36: On these subvarieties, the iterates of the critical set have a polynomial
37: growth in the \emph{degrees of the parameters}, while one has an exponential growth out
38: of these subspaces.
39: The analysis of our birational transformation in $\, CP_2$ is first carried out
40: using  Diller-Favre criterion in order to find the complexity
41: reduction of the mapping. The integrable cases are found.
42: The identification between the complexity
43: growth and the topological entropy is, one more time, verified.
44: We perform plots of the post-critical set, as well as calculations
45: of Lyapunov exponents for many orbits, confirming
46: that generically no meromorphic two-form can be 
47: preserved for this mapping. These birational transformations 
48: in $\, CP_2$, which,  generically, do not preserve any meromorphic two-form, 
49: are extremely similar to other birational transformations we previously studied,
50: which do preserve meromorphic two-forms. We note that these two sets of 
51:  birational transformations exhibit totally similar results as far as 
52:  topological complexity is concerned, 
53:  but drastically different results as far as a
54:  more ``probabilistic'' approach of dynamical systems is concerned
55:  (Lyapunov exponents). With these examples we see that the existence of a preserved
56: meromorphic two-form explains most of the (numerical) 
57: discrepancy between the topological and probabilistic approach of dynamical systems. 
58: 
59: 
60: 
61: 
62: \end{abstract} 
63: \vskip .5cm
64: 
65: \noindent {\bf PACS}: 05.50.+q, 05.10.-a, 02.30.Hq, 02.30.Gp, 02.40.Xx
66: 
67: \noindent {\bf AMS Classification scheme numbers}: 34M55, 47E05, 81Qxx, 32G34, 34Lxx, 34Mxx, 14Kxx 
68: 
69: 
70: \vskip .5cm
71:  {\bf Key-words}: Preserved meromorphic two-forms, invariant measures,
72: birational transformations, post-critical set, exceptional locus,
73: indeterminacy set, conservative systems, chaotic sets, complexity growth, 
74: Lyapunov exponents, topological category versus probabilistic category.
75: 
76: 
77: 
78: 
79: \vskip .1cm
80: 
81: 
82: 
83: \section{Introduction: Topological versus probabilistic methods in discrete
84:  dynamical systems}
85: \label{intro}
86: 
87: Two different approaches exist for studying discrete dynamical systems and
88: evaluating the complexity of a dynamical system: 
89: a topological approach and a probabilistic approach. A topological approach
90: will, for instance, calculate the topological
91:  entropy, the growth rate of the Arnold
92: complexity, or the growth rate of the successive degrees when iterating a
93: rational, or birational, transformation. This, quite algebraic, topological
94: approach is universal: one {\em counts integers}
95: (like some set of points, number of fixed points for the topological entropy,
96: number of intersection points
97:  for the Arnold complexity, or like the degrees of successive polynomials occurring
98:  in the iteration of rational
99: or birational transformations). This {\em universality} 
100: is a straight consequence
101: of the fact that integer counting remains invariant under any (reasonable)
102: reparametrization
103: of the dynamical system. Not surprisingly this (algebraic) topological
104: approach can be rephrased, or
105:  mathematically revisited (at least~\cite{bimero}
106:  in $\, CP_2$, and even~\cite{Kim} 
107: in $\, CP_n$), in the framework~\cite{bimero}
108: of a $\, H^{1,1}$ cohomology of curves in complex projective spaces
109: ($\, CP_2$, $\, CP_1 \times CP_1$).
110: In this topological approach, the dynamical systems are seen as dynamical
111: systems of {\em complex} variables
112: and, in fact, {\em  complex projective spaces}. 
113: 
114: The probabilistic (ergodic) approach, probably
115:  dominant in the study of dynamical
116: systems, is less universal, and
117: amounts to describing generic orbits, introducing some (often quite abstract)
118: positive invariant
119: measures, and other related concepts like the metric entropy (integral over
120: a measure of
121: Lyapunov exponents in a Pesin's formula~\cite{Pesin}). Roughly
122:  speaking, we might say that a phenomenological
123: approach consisting in the plot of as many real 
124: orbits as possible (phase portraits), or in the 
125: calculation of as many Lyapunov exponents as
126:  possible, in order to get some hint of the 
127: ``generic'' situation, also belongs to that probabilistic approach. In this
128:  probabilistic approach the dynamical systems 
129: are traditionally seen as dynamical systems
130:  of {\em real} variables,
131: dominated by {\em real functional analysis} (symbolic
132:  dynamics, Gevrey analyticity, ...), 
133: and differential geometry~\cite{Green} (diffeomorphisms, ...).
134: 
135: \vskip 0.2cm
136: \vskip 0.1cm
137:  
138: The fact that these two approaches, the ``hard'' one and the ``soft'' one, 
139: may provide (disturbingly)
140: different descriptions of dynamical systems is known by 
141: some mathematicians, but is hardly
142: mentioned in most of the graduate textbooks on discrete dynamical systems, 
143: which, for heuristic reasons,  try to avoid this question, implicitly 
144: promoting, in its most extreme form, the idea that most of the 
145: dynamical systems would be, up 
146: to strange attractors,  hyperbolic (or weakly 
147: hyperbolic) systems, the ``paradigm'' of 
148:  dynamical systems being the linearisable deterministic chaos 
149: of Anosov systems~\cite{Anosov,Anosov2}.  Of course, 
150: for such linearisable systems, these two
151:  approaches are equivalent. Along this line one should recall J-C. Yoccoz 
152:  explaining\footnote[1]{In his own address at 
153: the International Congress of Mathematicians in
154:  Zurich in 1994, or (in French) in~\cite{Yoccoz}.}
155: that the dynamical features that we are able to understand fall into two classes, 
156: hyperbolic dynamics and quasiperiodic dynamics: ``it may well happen, 
157: especially in the conservative case, that a system exhibits both 
158: hyperbolic and quasiperiodic features ...  we seek to extend these 
159: concepts, keeping a reasonable understanding of the dynamics, in order 
160: to account for as many systems as we can. The big question is then: 
161: Are these concepts sufficient to understand most systems'' ?  
162: 
163: The description of {\em conservative} cases (typically {\em area-preserving} maps, and, more generally,
164: mappings {\em preserving two-forms}, or $\, p$-forms) is clearly 
165: the difficult one, and the one for which the
166: distance between the two approaches, the ``hard'' one and the ``soft'' one, is 
167: maximum (in contrast with hyperbolic systems and, of course, linearisable Anosov systems). It is
168:  not outrageous to say that dynamical systems which are not
169:  hyperbolic (or weakly hyperbolic), or integrable (or quasiperiodic), but conservative, 
170:  preserving meromorphic two-forms (or $\, p$-forms), are {\em poorly understood, few tools, theorems, 
171: and results being available}.
172: 
173: In general for realistic reversible\footnote[9]{By reversible
174: we mean, flatly, invertible: the inverse map is well-defined, the number
175: of pre-image of a generic point being unique. Note that the word ``reversible'' is also used by
176: some authors~\cite{rob-qui-92,QuRo88} to say that the inverse 
177: map $\, K^{-1}$ is conjugate to the map itself $\, K$.
178: } mappings (which are far from being hyperbolic, 
179: or weakly hyperbolic, but closer to conservative systems), the equivalence of these two
180:  descriptions of {\em drastically different mathematical nature}
181:  is far from being clear.
182: 
183: This possible discrepancy between these two approaches (topological versus ergodic), is
184: well illustrated by the analysis of many discrete dynamical 
185: systems we have performed~\cite{BoMaRo93c,ab-an-bo-ma-2000,rearea,topo,complex}, corresponding to 
186:  iterations of (an extremely 
187: large class of) {\em birational} transformations.  These mappings
188:  have non-zero (degree-growth~\cite{complex} or Arnold growth rate~\cite{rearea})
189:  complexity,
190: or topological entropy~\cite{topo}, however, their orbits always look like (transcendental)
191: {\em curves}\footnote[2]{This
192: is the reason why we called these mappings ``Almost integrable'' in~\cite{BoMaRo93c}.} 
193: totally similar to the curves one would get with an integrable mapping,
194: and systematic calculations of the Lyapunov exponents of these orbits
195: give zero, or negative (for attracting fixed points), values.
196: To a great extent, the regularity of these
197: orbits, and, more generally,  the regularity of the whole phase portrait, seems to be related to 
198: the existence of preserved meromorphic two-forms (resp. $\, p$-forms)
199: for these birational transformations~\cite{ab-an-bo-ma-2000,rearea}. Could it
200:  be possible that (when being iterated) 
201: a birational transformation
202: could have a {\em non-zero topological entropy} and, in the same time, zero
203:  (or very small) metric (probabilistic)
204:  entropy, the previous ``almost-integrability'' being a consequence 
205: of preserved meromorphic two-forms (resp. $\, p$-forms) ?
206: 
207:  The existence of a preserved meromorphic two-form corresponds
208:  to a quite strong (almost algebraic)
209:  structure. Naively, one can imagine that a discrete dynamical system with a
210:  preserved meromorphic two-form should be ``less involved''
211:  than a discrete dynamical system
212:  without such differential structure. Should the
213:  existence of such exact differential structure
214:  be related to the ``hard'' topological, and algebraic, 
215: approach of discrete dynamical 
216:  systems (hidden K\"ahlerian structures\footnote[3]{
217: One may recall some exact algebraic 
218: (in their essence) results which are obtained
219: in some K\"alherian framework~\cite{Cantat,CantatFavre} (for instance,
220:  one inherits, immediately,
221:  a particular cohomology and strong differential structures~\cite{Green}). }
222:  for birational transformations, ...),
223:  or should it be related to the ``soft'' probabilistic (ergodic)  approach 
224: (possible relation between ``complex'' and ``real'' 
225:  invariant measures ...)? The answer to the previous question 
226:  will be fundamental to ``fill the
227:  gap'' between the two approaches or, at least, better
228:  understand the discrepancies between these two
229:  descriptions of birational dynamical systems. To answer this question,
230:  one would like to find two sets of birational
231:  transformations {\em as similar as possible}, but such that 
232:  one set {\em preserves} a meromorphic two-form, and the other
233:  set {\em does not preserve} a meromorphic two-form, in order
234:  to compare the topological and probabilistic approaches
235:  on these two sets. 
236:  
237:  Along this line, one should note that we found quite systematically,
238:  and surprisingly, preserved meromorphic
239:  two-forms (resp. $\, p$-forms) for an extremely large set of birational
240:  transformations in $\, CP_2$, and 
241:  in $\, CP_n$, $\, n >2$. Similar results were also found by 
242:  other groups\footnote[4]{J. Diller, (private communication).}
243:  for extremely large sets of birational transformations in $\, CP_2$.  
244:  Could it be possible that all birational transformations in $\, CP_2$
245:  preserve\footnote[5]{At first sight, such
246:  a strange result would present some similarity with the, still 
247:  quite mysterious, ``Jacobian conjecture'' of the 
248:  Smale's problems~\cite{Jacobconj}.} a meromorphic two-form ?
249:  We first need to find a first (and as simple as possible) 
250:  example of birational transformation in $\, CP_2$
251:  for which one can show, or at least get convinced of, a ``no-go'' result like
252:  the {\em non existence} of
253:  a meromorphic two-form (even very involved ...).
254: 
255:  The paper is organized as follows: we will first recall various ``complexity'' results 
256:  on a first set of birational transformations in $\, CP_2$, preserving meromorphic two-forms, 
257:  and we will also recall some results~\cite{bimero} of Diller and Favre
258:  on the topological approach of the complexity
259:  of these mappings. We will, then, introduce a {\em slightly}
260:  modified set of birational transformations in $\, CP_2$
261:  for which we will perform similar topological approach calculations. These calculations
262:  will provide, for this second set, subcases where meromorphic two-forms
263:  are actually preserved. This topological
264:  approach will yield us to introduce a fundamental tool, the {\em orbit of the critical
265:  set}\footnote[1]{Also called,
266:  by some mathematicians, {\em post critical set}, or, in short, ``PC''.
267:  Note that the general framework we consider here corresponds to birational
268:  transformations having a {\em non-empty indeterminacy set}, which is the
269:  natural framework when one considers {\em birational transformations} :
270:  the mathematician reader should forget all the theorems he knows on
271:  holomorphic transformations (toric monomial transformations, etc.)}, which 
272:  will give some strong numerical, and graphical, evidence
273:  that a meromorphic two-form {\em does not exist generically} 
274:  for this second set, outside the previous subcases. This non-existence of a  
275:  meromorphic two-form will be confirmed by a large set
276:  of Lyapunov exponents calculations, clearly exhibiting 
277:  non-zero positive Lyapunov exponents for this second set. We will, thus, 
278:  be able to conclude on the impact of the
279:  existence of a meromorphic two-form on the (apparent numerical) discrepancy between the
280:  topological and probabilistic (ergodic) approaches
281:  of discrete dynamical systems. 
282: 
283: 
284: 
285: \section{Two-forms versus invariant measures}
286: \label{2-formversus}
287: 
288: Let us first recall the birational transformation $\, k_{\epsilon}\, $
289:  in $\, CP_2$ we have 
290: extensively studied from a topological (almost algebraic) viewpoint, and, also,
291: from a measure theory (almost probabilistic) viewpoint~\cite{ab-an-bo-ma-2000}. It is a one parameter
292: transformation ($ \epsilon\,  \in\, C$, or $ \epsilon\,  \in\, R$) and it reads~\cite{topo,zeta}: 
293: \begin{eqnarray}
294: \label{keps}
295: (x, \, y) \, \rightarrow k_{\epsilon}(x, \, y)
296: \, =  \, \,  (x', \, y')\, =  \, \, 
297: \Bigl(y \cdot {{ x+\epsilon} \over {x-1}}, \,\, x+\epsilon-1\Bigr)
298: \end{eqnarray}
299: It was found \cite{zeta} that $\, k_{\epsilon}$, the 
300: $\, CP_2$ birational transformation 
301: (\ref{keps}), preserves\footnote[2]{Birational mapping (\ref{keps})
302: is a particular
303: case of a {\em two-parameter} dependent~\cite{zeta} 
304: birational mapping $\, k_{\epsilon,\alpha}$, which can also
305: be seen to preserve a meromorphic two-form~\cite{firthcoming}.} a 
306: meromorphic two-form~\cite{rearea}:
307: \begin{eqnarray}
308: \label{2form}
309: d\mu \, = \, \,\,{{ dx \cdot dy} \over { \rho(x, \, y)}}
310: \,  \, = \,\, \, {{ dx \cdot dy} \over { y \, -x\, +1}} 
311: \end{eqnarray}
312: The two-form (\ref{2form}) should not be called 
313: a ``measure'' since the denominator
314: $\, y \, -x\, +1$ can be negative.
315: The preservation of this  two-form corresponds to the following
316: identity between  the covariant $\, \rho(x, \, y)\, = \, \,y \, -x\, +1$
317: and the Jacobian of transformation $\, k_{\epsilon}$  :
318: \begin{eqnarray}
319: \label{jac}
320: J(x, \, y) 
321: \,\, = \,\,\, {{\rho(x', \, y')} \over {\rho(x, \, y)}}
322: \,\, = \,\,\, {{\rho(k_{\epsilon} (x, \, y))} \over {\rho(x, \, y)}}
323: \end{eqnarray}
324: The preservation of this two-form means that this birational 
325: mapping can be transformed, using a (non rational) change of variables, 
326: into an {\em area-preserving} mapping (see page 1475
327:  of~\cite{rearea}, or page 391 in~\cite{ab-an-bo-ma-2000}). As
328:  far as a ``down-to-earth'' visualization 
329: of the (real) orbits, and, more generally,
330: of the phase portraits, is concerned,
331: one sees that this  $\, k_{\epsilon}$-invariant
332:  two-form (\ref{2form}) can actually 
333: be ``seen'' on the phase portrait : 
334: near the straight line $\,  y \, -x\, +1\, = \,0$, corresponding
335:  to the vanishing of the denominator of  (\ref{2form}), the points of the  phase portrait
336: look like a ``spray'' of points ``sprayed'' near a wall
337:  corresponding to this straight line (see for instance Figure 2 right,
338: and Figures 3, 4, 6 and 7 in~\cite{rearea}).
339: 
340: This birational mapping was shown~\cite{topo,zeta} to have a non-zero
341: topological entropy and
342: a degree growth complexity (or growth rate of the Arnold complexity)
343: associated with a quadratic number (golden number), corresponding to the 
344: polynomial $\, 1-t-t^2$. 
345: However, the extensive Lyapunov exponents
346: calculations we performed, systematically, gave {\em zero values} 
347: for all the (numerous) orbits we considered
348:  (see Figure 3 right, or Figures 5, 8, 10, 21, and
349:  pages 403 to 419 of~\cite{ab-an-bo-ma-2000}). The orbits of this mapping
350:  look very much like {\em curves} and,
351:  thus, it is not surprising 
352: to get zero Lyapunov exponents (see paragraphs 4 and 5
353:  in~\cite{ab-an-bo-ma-2000}). This
354:  Lyapunov exponent viewpoint, as well as the down-to-earth visualization
355:  of the orbits,
356: suggests that the mapping is ``almost an integrable mapping'', in
357:  contradiction with the topological viewpoint. 
358: Recalling, just for heuristic reasons, some 
359: Pesin's like formula\footnote[3]{Such a birational mapping
360: is not a hyperbolic system, and the various other
361:  birational examples we have studied
362: are not even quasi-hyperbolic. Pesin's formula~\cite{Pesin} (see also 
363: pages 299 and 400 in~\cite{ab-an-bo-ma-2000}) is
364: certainly not valid here. We just recall it for heuristic reasons,
365: just as an analogy.}, considering
366: the entropy as the 
367: integral over ``some'' invariant measure $\, d\mu_{Lyap}$ 
368: of the Lyapunov exponents, 
369: it would be natural to ask where the non zero {\em positive} Lyapunov
370: exponents
371: are hidden? Where is this apparently ``evanescent'' invariant measure of
372: non zero {\em positive} Lyapunov exponents? It certainly does not
373: correspond
374: to any measure describing the previously mentioned ``spray'' of points
375: (which could be related to
376: the meromorphic two-form (\ref{2form})). For invertible mappings like
377: birational mappings,
378: the known way~\cite{Smillie} of building invariant measures as
379: successive
380:  pre-images\footnote[4]{Note that, for such non invertible cases,
381:  we found no contradiction
382: between the topological approach and the probabilistic (invariant measure)
383: approach:
384: for a non-invertible deformation of (\ref{keps}) we clearly found non
385: zero positive Lyapunov exponents for most of the orbits
386: (see paragraph 8 and Figures 27 and 28 in~\cite{ab-an-bo-ma-2000}).}
387:  of (almost) any point, simply does not work.  Bedford and 
388:  Diller~\cite{BedDill} showed how to build such invariant measure
389: $\, d\mu_{Lyap}$ corresponding to non-zero {\em positive} Lyapunov exponents,
390: for the (invertible) birational transformation (\ref{keps}).
391: Their method amounts to considering
392:  two arbitrary curves\footnote[5]{They might 
393: even be identical.} $\, \Gamma_1$ and $\, \Gamma_2$ 
394: (instead of an arbitrary point), iterate $\, \Gamma_1$ with $\, k_{\epsilon}$
395: and $\, \Gamma_2$ with $\, k_{\epsilon}^{-1}$, and consider the limit
396: set obtained as the intersection of these
397: two different iterated curves: 
398: the invariant measure emerges as a wedge product
399:  $\, \mu^{+} \wedge \mu^{-}$.
400: Such a wedge product construction is actually performed
401:  in detail in~\cite{BedDill} on mapping (\ref{keps}).
402:  The invariant measure built that way,
403: can be seen to correspond 
404: to an {\em extremely slim Cantor set, which is drastically different from
405: the meromorphic two-form} (\ref{2form}), or, more generally, from any invariant measure
406: one could imagine being associated with the previously mentioned 
407: spray of points. 
408: 
409: It is also worth recalling that Bedford and Diller were
410: also able~\cite{BedDill} on this very example, 
411: but only for $\, \epsilon \, < 0$ (where only saddle points occur), 
412: to build some {\em symbolic dynamics coding}, yielding a $\, 2 \times 2$
413:  matrix that actually identifies with 
414: some induced pullback $\, f^{*}$ on the cohomology group\footnote[1]{See the 
415: cohomological approach of Diller and Favre in~\cite{bimero}, to 
416: get the growth rate complexity.}  $\, H^2(P^1 \times P^1)$, 
417: thus filling, for $\, \epsilon \, < 0$, the gap between 
418: a {\em real analysis} approach of dynamical systems
419:  and an {\em algebraic projective complex analysis} of dynamical
420:  systems\footnote[2]{More recently
421: they have been able to generalize, very nicely~\cite{DillBed},
422: all these results to the birational mappings $\, k_{\epsilon, \alpha}$,
423: depending on two parameters~\cite{topo,zeta}. Mapping (\ref{keps}) is
424: obtained from $\, k_{\epsilon, \alpha}$ by setting $\alpha=0$.
425: This mapping \cite{topo,zeta}, $\, k_{\epsilon, \alpha}$,
426:  can also be seen to preserve a meromorphic two-form.
427: Paper \cite{DillBed} provides explicit examples of a 
428: $\, 5 \times 5$ matrix (linear map of the Picard group), and 
429: a $\, 4 \times 4$ matrix, encoding the symbolic
430:  dynamics, such that their characteristic polynomial
431: both contain a factor associated with the polynomial $\, 1-t-2\, t^2-t^3$, corresponding
432: to the (topological) complexities of our birational family analyzed in \cite{zeta}.}.
433: 
434: This provides a first answer to the discrepancy between the topological 
435: and probabilistic approach for such birational 
436: transformations (\ref{keps}) (at least\footnote[3]{For
437: $\, \epsilon \, > \, 0$, the situation is far from being so clear.} for
438: $\, \epsilon \, < \, 0$): as far as 
439: {\em computer experiments are concerned}, the regions where the 
440: chaos~\cite{Shilnikov,Shilnikov2,Shilnikov3,Shilnikov4} (Smale's horseshoe, 
441: homoclinic tangles, ...) is hidden, is concentrated in {\em extremely narrow regions}.   
442: 
443: 
444: \section{A first family of Noetherian mappings}
445: \label{Noether}
446: 
447: We have introduced in~\cite{Noether} a simple family of birational transformations in 
448: $\, CP_n$ ($n \, = \, 2, 3,  \cdots $)
449:  generated by the simple product
450: of the Hadamard inverse and (involutive) collineations. 
451: These birational transformations, we called Noetherian~\cite{Noether}
452:  mappings\footnote[4]{In reference to Noether's theorem of decomposition 
453: of birational transformations into products of quadratic transformations,
454:  like the Hadamard inverse,
455: and collineations~\cite{Noether}.},
456: present remarkable results for the growth-complexity, 
457: and the topological entropy,  in particular 
458: remarkable {\em complexity reductions}
459:  for some specific values of the
460:  parameters\footnote[5]{The parameters correspond to the entries of the collineation matrix.}
461:  of the mapping.
462:  These complexity reductions correspond
463: to a criterion, introduced by Diller and Favre~\cite{bimero}, based on the comparison between 
464: the  {\em orbit of the critical set, or even the exceptional locus},
465:  and the indeterminacy locus (see below (\ref{Diller})).
466: These mappings have similar properties compared to the
467: ones given for (\ref{keps}), namely a topological entropy, or 
468: a degree growth rate, associated with {\em algebraic numbers}, 
469: similar phase portraits, and the {\em existence of  preserved meromorphic two-forms}
470: for the transformations in 
471: $\, CP_2$, or, in $\, CP_n$,  preserved meromorphic $\, n$-forms, 
472:  together with $\, n-3$ algebraic invariants. 
473: In the following we will restrict ourselves to birational transformations 
474: in $\, CP_2$ : some of the results, we will display in the next
475:  sections, generalize, mutatis mutandis,
476: to  birational transformations in 
477: $\, CP_n$ ($n \, = \,  \, 3, \, 4, \, \cdots \, $) and some do not. 
478: 
479: \subsection{The mapping}
480: \label{themap}
481: 
482: Let us recall~\cite{Noether} the construction of the birational
483:  mapping $\, K$ product of a collineation $\, C$ and of a non-linear involution, 
484:  the Hadamard inverse,
485: $\, H $, acting on $\, CP_2$. We consider the standard quadratic homogeneous
486: transformation,  $\, H$,  
487: defined as follows on the three homogeneous variables
488: $\, (t,x,y)\, $ associated with $\, CP_2\,$: 
489: \begin{eqnarray}
490: \label{quadra}
491: H:\,\,\quad (t,x,y)\,\, \,\longrightarrow 
492: \,\,\,\,\,(x \; y,\,\, t \; y, \,\, t \; x)
493: \end{eqnarray}
494: We also introduce the following  $\, 3 \times 3\, $  matrix, acting on 
495: the three homogeneous variables $\, (t,x,y)\,$:
496: \begin{equation}
497: \label{C2}
498: M_C \, = \, \, 
499: \left [\begin {array}{ccc} 
500: a-1 &  b & c\\
501: a & b-1 & c\\
502: a & b & c-1
503: \end {array}\right ]
504: \end{equation}
505: and the associated collineation  $\, C$ which reads, in terms
506:  of the two inhomogeneous variables $\, u=\, x/t\, $ and $\, v=\, y/t$ :
507: \begin{eqnarray}
508: \label{colliC2}
509: && (u, v)\, \,\, \,\longrightarrow \,\,\, \, (u', v') \,\, = \, \, \\
510: &&   \quad \quad \quad \quad \quad \quad \quad \quad\, = \, \,
511: \Bigl( {{a\, + (b-1)\, u \, + \, c\,v } \over {
512: (a-1)\, \, + b\, u \,+ \, c\, v}}, \, \, \, 
513: {{a\, + b\, u \, + \, (c-1)\,v } \over {
514: (a-1)\, \, + b\, u \,+ \, c\, v }} \Bigr) \nonumber 
515: \end{eqnarray}
516: The birational mapping  $\, K \, = \, C \cdot H$,
517:  reads, in terms of the two inhomogeneous
518: variables $\, u=\, x/t\, $ and $\, v=\, y/t$:
519: \begin{eqnarray}
520: \label{BiC2}
521: &&  K : \quad (u, v)\, \,\longrightarrow \,\, (u', v') \, = \,  \\
522: &&   \quad \quad \quad \quad \quad \quad \quad\, = \, \,
523:    \Bigl({\frac {a\, uv\, +(b-1)\, v\, +cu}
524: {(a-1)\, uv \, + bv\, +cu}} , \, \, \, {\frac
525: {a\, uv\, +bv\, +(c-1)\, u}{(a-1)\, uv\, +b \, v\, +c\, u}} 
526: \Bigr) \nonumber 
527: \end{eqnarray}
528: 
529: This birational mapping (\ref{BiC2}) {\em conformally}\footnote[3]{This means 
530: that the two-form is preserved {\em up to a constant} $\xi$.}
531: preserves a two-form. Actually, if one considers the product
532:  $\, \rho(u,v) \, = \, (u-1)\, (v-1)\,(u-v)$,
533: a straightforward calculation shows that 
534:  $\, J(u,v)$,  the Jacobian of (\ref{BiC2}),
535: is actually equal to:
536: \begin{eqnarray}
537: \label{equalcov} 
538: & & J(u,v) \, = \,\,\,\, \xi  \cdot 
539:  {{\rho(u',v')} \over {\rho(u,v)}}\,
540: \, =
541: \,\,\,  \xi \cdot {{ u \, v}
542: \over { ((a-1) \, u \, v \, + c \, u\, + \, b\, v)^3}}
543: \end{eqnarray}
544: where $\, \xi \, = \, a+b+c-1$ and where $\, (u', v')$ is the image of $\, (u, v)$
545: by the birational transformation (\ref{BiC2}), or equivalently
546: \begin{eqnarray}
547: \label{conf2form} 
548:  {{ du'\cdot dv'} \over { (u'-1)\, (v'-1)\,
549: (u'-v')}} \, = \,\,\,\,  \xi \cdot  {{ du\cdot dv} \over { (u-1)\, (v-1)\,
550: (u-v)}}
551: \end{eqnarray}
552: For  $\,\xi\, = \, 1$ (i.e. det$(M_C)=1$),
553: the matrix $\, M_C$, as well as its associated
554: collineation $\, C$, are involutions, and 
555:  the two-form (\ref{conf2form}) is {\em exactly preserved}. 
556: 
557: \subsection{Diller-Favre criterion: complexity reduction 
558: from the analysis of the  orbit of the exceptional locus}
559: \label{Diller}
560: We recall, in this section, the Diller-Favre method~\cite{bimero},
561: in order to describe the singularities of the mapping, and deduce 
562: {\em complexity reductions} of the mapping. In particular 
563: we give, for mapping (\ref{BiC2}), the equivalent 
564: of Lemma 9.1 and Lemma 9.2 of~\cite{bimero}.   
565: 
566: We assume, here, that  condition $\, c \, = \, 2-a-b\, $ is
567: satisfied (i.e. $\xi=1$).
568:  The Jacobian $\, J(u,v)\, $ vanishes on
569: $\, u \, = \, 0$, on  $\, v\, = \, 0$, and 
570: becomes infinite when 
571: $v\, = \, -c \, u/((a-1)\, u \, +b)$.
572: 
573: 
574: Using the same terminology as in~\cite{bimero},
575: one can show that the exceptional locus\footnote[1]{Corresponding
576:  to the critical set $\, J(u,v)\, = \, \,0$, together with condition
577:  $\, J(u,v)\, = \, \infty$.} 
578: of  $\, K\, $ is given by
579: \begin{eqnarray}
580: \label{exceplocus}
581: {\cal E}(K) & = & 
582: \, \,\left \{ (u=0);\, (v=0) ; \, \Bigl(v=\, {{- c \, u} 
583: \over { (a-1)\, u \, + b}} \Bigr) \right \}  
584: \end{eqnarray}
585: and the indeterminacy locus~\cite{bimero} of $\, K\, $ is
586: given by:
587: \begin{eqnarray}
588: {\cal I}(K)\, =   \,\,
589:  \left \{ (0,0); \,\, \Bigl( {{b}\over{(b-1)}},1 \Bigr);\, \, 
590:  \Bigl( 1, {{c}\over{(c-1)}} \Bigr) \right  \} \nonumber 
591: \end{eqnarray}
592: Actually, for $\, (u, \, v) \, = \, \, (0,0)$, 
593: the $\, u$ and $\, v$ components of $\, K$ 
594: are, both, of the form $\, 0/0$, for $\,(u, \, v) \, = \, \, 
595: ( b/(b-1), \, 1)$, the  $\, v$-component of $\, K$ 
596: is of the form $\, 0/0$, and, for $\,(u, \, v) \, = \, \, 
597: (1, \, c/(c-1))$, the  $\, u$-component of $\, K$ 
598: is of the form $\, 0/0$.
599: 
600: As far as the three
601:  vanishing conditions (\ref{exceplocus})
602:  of the Jacobian, or its inverse, are concerned,
603: it is easy to see that their successive images
604:  by $\, K$ 
605: give respectively, when condition $\xi=1$ is satisfied:
606: \begin{eqnarray}
607: \label{0v}
608: &&(0, \, v) \quad \rightarrow \quad 
609: \Bigl( {{b-1 } \over {b}}\,, \, \, 1
610: \Bigr) \quad \rightarrow \quad 
611:  \cdots \quad \rightarrow \quad 
612: \Bigl( {{n\, (b-1) } \over {n\, b\, -(n-1)}}, \, 1\Bigr) \nonumber  \\
613: \label{0u}
614: &&(u, \, 0) \quad \rightarrow \quad
615:  \Bigl ( 1\, , \, \, {{c-1 } \over {c}}
616: \Bigr) \quad \rightarrow  \quad \cdots \quad \rightarrow  \quad
617:  \Bigl( 1, \; \; {{n (c-1)}\over{n c -(n-1)}} \Bigr) \nonumber \\
618: \label{ucu}
619: && (u \, , \, {{- c \, u} 
620: \over { (a-1)\, u \, + b}}) \, \,  \rightarrow \, \, 
621:  \Bigl( \infty \, , \,  \infty
622: \Bigr)  \, \,  \rightarrow  \, \,  \cdots  \\
623: && \qquad \qquad  \, \,  \rightarrow  \, \, 
624:  \Bigl( {{(n-1) a -(n-2)}\over{(n-1) (a-1)}} , \;\;
625: {{(n-1) a -(n-2)}\over{(n-1) (a-1)}} \Bigr) \nonumber
626: \end{eqnarray}
627: 
628: Do note that the iterates of ${\cal E}(K)$ for $n=\infty$ converge towards $(1,1)$
629: the fixed point of {\em order one} of mapping $K$.
630: 
631: \vskip 0.1cm
632: 
633: One has similar results~\cite{Noether} for  the successive images by
634: $\, K^{-1}$ of its exceptional locus.
635: 
636: 
637: At first sight it may look remarkable that 
638: the image  by $\, K$ of {\em curves} (like the three vanishing 
639: conditions (\ref{exceplocus}) of the Jacobian, or its inverse)
640: {\em actually blow down into points}.  
641: This is, in fact, a natural feature\footnote[2]{If one considers the set of 
642: points where the Jacobian vanishes, also called critical set, and 
643: assume that some
644: part of this critical set is not blown down into a point, then
645: the birational mapping would not be (locally) bijective.
646: Such points would have, at least, two preimages in contradiction with the
647: birational character of the transformation. 
648: This sketched proof remains valid for a birational 
649: transformation in $\, CP_n$ for  $\, n \, \ge 3$.}
650: of birational  transformations (even in $\, CP_n$). Such a phenomenon
651: of blow down {\em can only occur for transformations
652:  having a non empty indeterminacy set:}
653: for instance, {\em it cannot occur with holomorphic transformations}. 
654: 
655: One remarks that all these $\, n$-th iterates (by $\, K$ or $\,
656: K^{-1}$) belong (for $\, n \, \ge 2$) to the three $\, K$-invariant 
657: lines, namely 
658: $\, u\, = \, 1$,  $\, v\, = \, 1$, or $\, u \, = \, v$.  
659: 
660: Diller and Favre statement is that 
661: the  mapping $\, K$ {\em is analytically stable}~\cite{bimero} 
662: {\em if, and only if,} $\, \, \, K^n ( {\cal E}(K) ) \notin {\cal I}(K)\,\,  $
663: (respectively  $\, K^{(-n)} ( {\cal E}(K^{-1}) ) \notin {\cal I}(K^{-1})$)
664: for all $\, n \, \ge \,  1$. In other words the {\em complexity reduction},
665: which breaks the analytically stable character of the mapping,
666:  will correspond to situations where
667: some points of the orbit of the exceptional locus 
668: ($ K^n ( {\cal E}(K) )$) encounter
669: the indeterminacy locus  ${\cal I}(K)$.
670: Having an {\em explicit} description of these orbits (see
671:  (\ref{0v})) for this birational transformation,
672: one can easily deduce the complexity reduction situations 
673: associated with parameters $\, a$, $\, b$, or $c$, being of the form
674:  $\, (N-1)/N$, where  $\, N$ 
675: is any positive integer.
676: For instance, when $\, a=\, (M-1)/M$ ($M\, $ positive integer)
677:  and $\, b$ generic, one gets a complexity reduction. The
678: complexity~\cite{Noether} being associated with 
679: polynomial\footnote[4]{The degree generating function~\cite{rearea,zeta}
680: is a rational expression with polynomial (\ref{familN}) in its denominator.} 
681: \begin{eqnarray}
682: \label{familN}
683: P \, = \,\,\,\,  1\,-2\,t\,+{t}^{M+1}
684: \end{eqnarray}
685:  and, similarly, 
686: when $\, a=\, (M-1)/M$ and $\, b\, =\, (N-1)/N$ ($M\, $
687: and $N\, $  positive integers),
688: the complexity is associated~\cite{Noether} with polynomial :
689: \begin{eqnarray}
690: \label{familMN}
691: P_{M,N} \, = \,\,\,\, 1\,-2\,t\,+{t}^{M+1}\,+{t}^{N+1}\,-{t}^{M+N} 
692: \end{eqnarray}
693: 
694: 
695: \subsection{Beyond the involutive condition: $\xi =a+b+c-1 \ne 1$}
696: \label{mune1}
697: Let us show that the iterates of the exceptional locus have also explicit
698: expressions when $\, C$ is no longer involutive (namely $\xi \ne\, 1$).
699: The iterates of ${\cal E}(K)$ become:
700: \begin{eqnarray}
701: \label{0vc}
702: &&(0, \, v) \quad  \rightarrow  \quad 
703: \Bigl( {{b-1 } \over {b}}, \, \, \, 1
704: \Bigr) \quad \rightarrow \, 
705:  \cdots \, \rightarrow \quad (U_n, 1) \nonumber \\
706: \label{0uc}
707: &&(u, \, 0) \quad  \rightarrow  \quad 
708: \Bigl ( 1,\, \, \, {{c-1 } \over {c}}\Bigr)
709: \quad \rightarrow \,\,
710:  \cdots \,\, \rightarrow \quad (1, V_n) \nonumber \\
711: \label{ucuc}
712: &&\Bigl(u,  \, \, \, {{- c \, u} \over { (a-1)\, u \, + b}}\Bigr) \, \,  
713:  \rightarrow  \, \, 
714:  \Bigl( \infty \, , \, \, \infty
715: \Bigr)  \, \,  \rightarrow  \, \, \, \cdots  \,\, \,  \rightarrow  \, \, 
716:  \Bigl( X_n, X_n  \Bigr) \nonumber
717: \end{eqnarray}
718: with:
719: \begin{eqnarray}
720: \label{Unabc}
721: U_n(a,b,c) & = & \, \,\, \, {\frac{(b-1)\left((a+b+c-1)^n-1\right)}
722: {(b-1)(a+b+c-1)^n+(a+c-1)}}  \\
723: V_n(a,b,c) & = & \, \, \, U_n(a,c,b), \qquad \quad  X_n(a,b,c)\, =\,\, 1/U_{n-1}(b ,a, c) \nonumber 
724: \end{eqnarray}
725: 
726: 
727: Now, the iterates of ${\cal E}(K)$ in the $n=\infty$ limit, depend on the value
728: of $\xi=a+b+c-1$ and read:
729: \begin{eqnarray}
730: \vert  \xi \vert  < 1 &&  \qquad U_n\,  \rightarrow \,{\frac{1-b}{a+c-1}},
731: \,\,\,\,\,\, V_n \,\rightarrow\, {\frac{1-c}{a+b-1}},
732: \, \,\,\,\,\, X_n\, \rightarrow\, {\frac{b+c-1}{1-a}} \nonumber \\
733: \vert  \xi \vert  > 1 &&  \qquad U_n \,\rightarrow\, 1,\,\,\,\,
734: \,\, \,V_n \,\rightarrow\, 1,\,\,\,\,
735: \, \,\, X_n \,\rightarrow\, 1
736: \end{eqnarray}
737: 
738: The above limits are precisely the fixed point(s) of {\em order
739: one} of mapping K which read:
740: \begin{eqnarray}
741: \Bigl(1, 1 \Bigr), \quad \Bigl(1, {\frac{1-c}{a+b-1}} \Bigr), \quad
742: \Bigl({\frac{1-b}{a+c-1}}, 1 \Bigr), \quad
743: \Bigl({\frac{b+c-1}{1-a}}, {\frac{b+c-1}{1-a}} \Bigr) \nonumber
744: \end{eqnarray}
745: 
746: Again, one remarks that all these $\, n$-th iterates (by $\, K$ or $\,
747: K^{-1}$) belong (for $\, n \, \ge 2$) to the three $\, K$-invariant 
748: lines $\, u\, = \, 1$,  $\, v\, = \, 1$, or $\, u \, = \, v$, allowing 
749: a meromorphic two-form like (\ref{conf2form}) to be (conformally) preserved.  
750: 
751: For $\xi=1$, the four fixed points of  order one collapse to a only one.
752: For $\xi \ne 1$, the iterates of the exceptional locus converge to one, or more
753: than one, fixed point(s) of order one.
754: 
755: 
756: \section{A second family of Noetherian mappings}
757: \label{antisto}
758: Let us, now, introduce another set of birational transformations in $\, CP_2$,
759: built in a totally similar way as the Noetherian mappings~\cite{Noether}
760: of the previous section, 
761: namely as product of a collineation $C$ and the previous quadratic
762: transformation
763:   $H$ (Hadamard inverse (\ref{quadra})). Our only 
764: slight modification is that the $\, 3 \times 3$ matrix
765:  $M_C$, associated with this collineation, is now the {\em transpose}
766:  of matrix $M_C$
767: considered in~\cite{Noether} and previously given in (\ref{C2}). It is
768: straightforward to remark
769: that $\, \xi \, = \, a+b+c-1\, = \, 1$ is, again, the condition for
770: collineation $\, C$ to be an involution ($\det(M_C) =1$).
771: In that involutive case it is also straightforward to see that $\,K^N$,
772: and  $\,K^{-N}$, are conjugated :  $\,K^{-N} \, = \, \,$
773: $ C \cdot K^N \cdot  C $ $\,\, = \,C^{-1} \cdot K^N \cdot  C \,$
774:  $\,\, = \,H^{-1} \cdot K^N \cdot  H \,$
775:  $\,\, = \,H \cdot K^N \cdot  H$.
776: Thus transformations $\,K$ and  $\,K^{-1}$ have {\em necessarily
777: the same complexity}.    Most of the results
778: we will display in the following, will be restricted (for heuristic reasons) 
779: to this involutive condition 
780: $\, \xi \, = \, a+b+c-1\, = \, 1$, but it is important to keep 
781: in mind that many of these results can be 
782: generalized to the non-involutive case $\, \xi \, \ne \, 1$. 
783: 
784: The mapping $\, K\,=\,C \cdot H$, in terms of inhomogeneous
785: variables ($u=x/t$, $v=y/t$), reads:
786: \begin{eqnarray}
787: \label{defKanti}
788: && K: \,\, (u,v)
789: \,\,\longrightarrow \,\,  \\
790: &&  \quad \quad \quad \quad  \Bigl( {\frac{b\, uv\,+(b-1)\, v\,+b\, u} 
791:  {(a-1) \, uv\, +a\, (u+v)}},\,\,
792: {\frac{c \, uv \, + c\, v \, +(c-1) \, u}
793: {(a-1) \, uv +a\, (u+v) }} \Bigr) \nonumber
794: \end{eqnarray}
795: 
796: 
797: When written in a homogeneous way, it is clear, since the three 
798: homogeneous  variables, as well as 
799: the three  parameters  $\, (a, b, c)$, 
800:  are on the same footing, that transformation 
801: $\,K\, =\,C \cdot H$ must exhibit a symmetry with respect to the
802: group of permutations of the three (homogeneous) variables.
803: The symmetry, induced by this group of permutations of the
804:  three homogeneous variables,
805: leads to equivalence between mappings with different couple of 
806: parameters $a$ and $b$ (with $c=2-a-b$). 
807: The change $(a,b) \rightarrow (b,a)$ combined with
808: $(u,v) \rightarrow (1/u, v/u)$, and the change
809: $(a,b) \rightarrow (a, 2-a-b)$
810: combined with $(u,v) \rightarrow (v,u)$, leave the mapping $K$ unchanged.
811: Defining the two involutions
812: \begin{eqnarray}
813: P: (a,b) \,  \longrightarrow \,  (a,2-a-b), \qquad \qquad 
814: T: (a,b) \,  \longrightarrow \,  (b,a)  
815: \end{eqnarray}
816: the parameter plane $(a,b)$ is composed of six
817:  equivalent regions reached by five
818: transformations of one region. 
819: The five regions are reached from (e.g.) the region 
820: $\, 1\, -a/2\,  \le\,  b \, \le\,  a$
821: by the action of\footnote[3]{Note that $P\cdot T $ (or $ T\cdot  P$)
822:  is an order three symmetry.
823: } $P$, $T$, $P\cdot  T$, $T\cdot  P$ and $P \cdot T\cdot  P$.
824: It means that the mappings built with one of the matrices
825: $M_C$, $P\cdot M_C$, $T \cdot M_C$, $P\cdot T \cdot M_C$, $T\cdot P \cdot M_C$,
826: $P\cdot T\cdot P \cdot M_C$ are equivalent. As a consequence, if $(a, b)$
827:  gives the complexity
828: $\lambda$, so do $P(a, b)$, $T(a, b)$,
829: $P\cdot T(a, b)$, $T\cdot P(a, b)$, $P\cdot T\cdot P(a, b)$ for
830: the corresponding mapping.
831: The fixed points of the involutions $P$, $T$ and $\, P\cdot T\cdot P$ lie,
832: respectively, on three lines:
833: \begin{eqnarray}
834: \label{ablines}
835:   b=\, 1-a/2, \qquad \quad  b=\, a, \qquad \quad   b=\, 2-2a  
836: \end{eqnarray}
837: These three lines present interesting properties as will be seen in the 
838: following.
839: The fixed point of $\,P \cdot T$, or $\, T \cdot P$, 
840: correspond to a point $\, a=b=2/3\, $
841: in the $\, (a, \, b)$ parameter plane (we will see below that 
842:  this corresponds to an integrable mapping).  
843: As far as symmetries in the $\, (a, \, b)$ parameter plane are concerned,
844: another codimension-one subvariety pops out, namely the quadric
845: \begin{eqnarray}
846: \label{defC0}
847: C_0(a, \, b)\, =\,\,\, a^2\, +b^2\, +a b\, -2 (a\, +b)\, =\,\, 0
848: \end{eqnarray}
849: which is invariant under the five transformations
850: $P$, $T$, $P\cdot T$, $T\cdot  P$ and $P \cdot T \cdot P$.
851: Having a genus 0, curve (\ref{defC0}) has a rational parametrization.
852: 
853: 
854: Condition $C_0(a,b)=0$ occurs as a condition for $\, K$ {\em to be an 
855: order two} transformation not in the whole $(u, \, v)$ plane, but {\em on some singled-out 
856: curve} (see the algebraic curve (\ref{list2}) below).  Note that, an algebraic curve such that 
857: $\, K^2(u, \, v) \, = \, \, (u, \, v)$ is {\em necessarily a
858:  covariant} curve for $\, K$. 
859: 
860: 
861: \subsection{Diller-Favre complexity reduction analysis on
862:  the new Noetherian mappings}
863: \label{complexanaly}
864: In order to perform a complexity reduction analysis on (\ref{defKanti}),
865: similar to the one displayed in section (\ref{Diller}),
866:  based on the Diller-Favre criterion,
867:  let us calculate the  Jacobian of $\, K$, the birational 
868: transformation (\ref{defKanti}):
869: \begin{eqnarray}
870: \label{jact}
871: J(u, \, v)\, =\,\,\,\, 
872:  {\frac { \left( a+b+c-1 \right)\, uv  }
873: { \left((a-1)\,uv\,+a(u+v) \right) ^{3}}}
874: \end{eqnarray}
875: Denoting $J^{(-1)}$ the Jacobian of
876: $K^{-1}$,
877: one  easily verifies that (as it should):
878: \begin{eqnarray}
879:  J(K^{-1}(u, \, v))  \cdot J^{(-1)}(u, \, v)\, \,  = \, \ \, 
880:  J(u, \, v)  \cdot J^{(-1)}(K(u, \, v)) \,\,  =\,\, + 1 \nonumber 
881: \end{eqnarray}
882: 
883: The finite set of points of indeterminacy of the mapping, ${\cal I}(K)$, and  
884: the finite set of exceptional points of the mapping (critical set $\, J=0$, 
885: together with condition
886: $\, J=\infty$), ${\cal E}(K)$,
887: read: 
888: \begin{eqnarray}
889: {\cal I}(K) & = & \,\, \left \{ I_1, I_2, I_3 \right \}\,=\,\,
890: \left \{ (0,0); ({{ a} \over {b}}, {\frac{a}{1-a-b}});({\frac{a}{b-1}}, 
891: {\frac{a}{2-a-b}}) \right \} \nonumber \\
892: {\cal E}(K) & = \,\, & \left \{ V_1, V_2, V_3 \right \}\,=\,\,
893: \left \{ (u=0); (v=0);(u={\frac{-av}{v(a-1)+a}}) \right \} \nonumber 
894: \end{eqnarray}
895: Let us focus on the first iterates of one of the three vanishing conditions
896: of the  Jacobian $\, V_2$, namely $\, v \, = 0$:
897: \begin{eqnarray}
898: \label{singsing}
899: && (u_1, v_1) \, =\, \, \, 
900: \Bigl( {\frac{b}{a}},\,\, {\frac{1-a-b}{a}}\Bigr), \nonumber \\
901: && (u_2, v_2) \, =\, \, \,
902: \Bigl({\frac{(b-1)}{(a-1)}}{\frac{(C_2^{22}+b)}{(C_2^{22}+a)}},\,\,
903: {\frac{(1-a-b)}{(a-1)}}
904: {\frac{(C_2^{22}-a-b)}{(C_2^{22}+a)}}\Bigr), \nonumber \\
905: && (u_3, v_3) \, =\, \, \, \cdots 
906: \end{eqnarray}
907: The expression $C_2^{22}$ is given in (\ref{C222}) below.
908: Do note that, in contrast with the situation encountered
909: in the previous section (see (\ref{0v}), (\ref{Unabc})), the degree growth 
910: of (the numerator or denominator of) these successive expressions
911: in the parameters $a$ and $b$ is, now, {\em actually exponential, and, thus, 
912: one does not expect closed forms for the successive iterates} ($u_N$, $v_N$).
913: We will denote $\, \delta$ the degree growth rate (complexity)
914:  associated with the exponential degree growth
915:  $\, \simeq  \delta^N$ of these $u_N$'s 
916: and $v_N$'s (in the $(a, \, b)$ parameters).
917:  This degree growth rate (in the parameters $a$ and $b$)
918:  of the iterates of the vanishing
919:  conditions of the Jacobian depends on the values of $a$ and $b$.
920:  In the previous section (see (\ref{0v}), (\ref{Unabc}))
921: this degree growth rate was $\delta=1$ for generic values of the parameters.
922: 
923: Before performing any calculation, let us remark that, due to the
924:  previously mentioned permutation symmetry,
925: the nine ``Diller-Favre conditions''  $K^N( {\cal E}(K) ) \in I(K)$ 
926: for complexity reduction, are related
927: \begin{eqnarray}
928: \label{equiv}
929: & & K(V_1) \in I_1 \, \Longleftrightarrow \,  P \cdot K(V_2) \in I_1, 
930:  \quad \,\,
931: K(V_1) \in I_2  \, \Longleftrightarrow \,  K(V_2) \in I_3,  
932: \nonumber \\
933: & & K(V_2) \in I_2  \, \Longleftrightarrow \,  P \cdot K(V_1) \in I_3, 
934:  \quad \,\,
935: K(V_3) \in I_3  \, \Longleftrightarrow \,  P \cdot K(V_3) \in I_2 
936: \nonumber
937: \end{eqnarray}
938: 
939: The method in~\cite{bimero} amounts to solving $K^N(V_i) \in I_j$.
940: One obtains, for mapping (\ref{defKanti}),
941:  algebraic curves in the $(a, \, b)$-plane, with some singled-out
942:  $(a, \, b)$ points. These
943: algebraic curves appear, at {\em even} orders,
944: as common polynomials (gcd) in the components
945: of $K^N(V_1) \in I_3$, or  $K^N(V_2) \in I_2$ or
946:  $K^N(V_3) \in I_1$.
947: Let us call these algebraic curves associated with conditions 
948: $K^N(V_i) \in I_j$, respectively
949: $C_N^{13}$, $C_N^{22}$ and $C_N^{31}$ ($N$ being even).
950: For instance $C_2^{22}$ corresponds to $K^N(V_2) \in I_2$,
951:  that is $\, (u_2, \, v_2)\, = \, (a/b, \, a/(1-a-b))$, which reads 
952:  $\, (a^2+ab+b^2)-(a+b)\, = \,0$. 
953: These algebraic curves are $(a, \, b)$-subvarieties of complexity growth,
954: for (\ref{defKanti}),
955:  lower than the generic one ($\lambda=2$), and they are related
956: by $\,  P \cdot C_N^{13} = C_N^{22}$ and $\,  T \cdot C_N^{13} = C_N^{31}$.
957: They are polynomials in $a,b$ of degrees $2$, $6$, $12$, $26$,
958: $48$, $98$, $\cdots\, $ (for $N \, = \, 2, \, 4, \, 6, \, 8, \, 10, \, 12, \cdots$).
959: Since they are calculated from the $\, u_N$'s  and 
960: $\, v_N$'s (\ref{singsing}) which are rational expressions
961: in $(a, \, b)$ with corresponding polynomials
962: of degree growing exponentially like $\, \delta^N \sim 2^N$, it is not 
963: surprising to see the degree of these successive $(a, \, b)$ polynomials
964: growing exponentially, but with a {\em lower} rate (see Appendix A).
965: 
966: 
967: Note that the singularities of these algebraic curves
968: (from a purely algebraic geometry viewpoint: local branches, ...)
969: correspond to  points $(a, \, b)$, in the parameters
970: space, for which the birational transformation $\, K$ has actually {\em lower}
971: complexities (see Appendix A). 
972: Note that the singularities of the curves $\, C_N$'s contain those of
973: the curves of lower $N$. A detailed analysis of this set
974: of curves, their mutual intersections, and the relation between these
975: intersections, and  singled-out (singular) points of the curves, and
976: the associated further reduction of complexity, will not be performed here.
977: 
978: 
979: The polynomials $\,C_N^{22}$ appearing in this complexity reduction analysis,
980: are, of course, symmetric in $\, a$ and $\,b$.
981: Those of the first orders read:
982: \begin{eqnarray}
983: \label{C222}
984: && C_2^{22}  = \, (a^2+ab+b^2)-(a+b)  \\
985: && C_4^{22} = \,(a^2+ab+b^2)^3
986: -(a+b)(a^2+ab+b^2)(4a^2+7ab+4b^2) \nonumber \\
987: && \qquad  \quad  +(7a^4+26a^3b+36a^2b^2+26ab^3+7b^4) \nonumber \\
988: && \qquad  \quad  -(a+b)(6a^2+11ab+6b^2)+(2a^2+3ab+2b^2) \nonumber 
989: \end{eqnarray}
990: These polynomials  $\,C_{N}^{ij}$ ($ij=13,22,31$)
991: have been obtained up to  $\, N\, = \, 12$.
992: Some of their algebraic geometry properties (singularities, genus, ...) 
993: are summarized in Appendix A.
994: \begin{figure}
995: \psfig{file=Qcn3.ps,scale=0.53}
996: \caption{Polynomials $C_N$ in the $(a, \, b)$ parameter plane (upper right corner).}
997: \label{f:fig1}
998: \end{figure}
999: 
1000: Let us display these various algebraic curves
1001:  $\, C_N^{ij}$ in the $(a, \,\, b)$-parameter plane. One sees, on Figure \ref{f:fig1} (upper right corner), 
1002: that this accumulation of curves
1003: looks, a little bit, like a (discrete) ``foliation'' of the
1004: $(a, \,\,b)$-plane in curves similar to a linear pencil
1005:  of algebraic curves~\cite{Cremona}, the 
1006: ``base points'' of this linear pencil being, in fact, singular points
1007: of these $\, C_N$'s (see Appendix A) of lower 
1008: complexity and sometimes, $(a, \,\,b)$ points for which the 
1009: mapping becomes integrable.
1010: 
1011: 
1012: On these algebraic  curves
1013: $\,C_N^{ij}\, = \, 0$ $(N=2, 4, 6, \cdots)$, the
1014:  complexity is given by the inverse of
1015: the smallest root of:
1016: \begin{eqnarray}
1017: \label{shift}
1018: 1-2t+t^{N+2}\, =\, \, 0
1019: \end{eqnarray}
1020: As $N$ increases, the complexity reads $\lambda=1.8392, 1.9659,$
1021: $  1.9919, \cdots $
1022: One recovers a family of complexities (depending on $\, N$) 
1023: already seen for the Noetherian mappings~\cite{Noether}
1024: of the previous section and, 
1025: even,  for the  
1026: mapping (\ref{keps}) for $\, \epsilon \, = \, 1/N$ (see~\cite{zeta}). 
1027: Actually, one finds a shift of $\, +1$
1028:  between (\ref{familN}) and (\ref{shift}).
1029: 
1030: In contrast with the situation encountered with the
1031:  Noetherian mappings of the previous section (\ref{Diller}) (see also~\cite{Noether}),
1032: the complexity reduction conditions are now involved families
1033:  of polynomials (exponential degree growth in the $(a, \, b)$
1034:  parameters i.e. $\delta >1$),
1035:  instead of the previous extremely simple, and 
1036: {\em separated} conditions~\cite{Noether} 
1037:  in the $\, a$, $\, b$, $\,c$ 
1038: variables ($a \, = \, (N-1)/N$, ...). 
1039: 
1040: Recalling the complexity reduction scheme described in 
1041: section (\ref{Diller}) for mapping (\ref{BiC2}),
1042: we saw further complexity reductions on the intersections of
1043: two complexity reduction conditions $\, a=\, (M-1)/M$ and 
1044: $\, b\, =\, (N-1)/N$ ($M\, $
1045: and $N\, $  positive integers)
1046: and $\, c \, = \, 2-a-b\, $,
1047: namely families of complexities depending on the two integers
1048: $N\, $ and $\,M$
1049:  associated~\cite{Noether} with polynomials 
1050: $\,1-2\,t+{t}^{M+1}+{t}^{N+1}-{t}^{M+N}$. 
1051: 
1052: By analogy, it is natural to see if a similar complexity reduction scheme
1053: also occurs for mapping (\ref{defKanti}), by
1054: calculating the degree growth complexity when the parameters $a$, and $b$, 
1055: are restricted to the intersection of two conditions  $\,C_N^{ij}\, = \, 0$. 
1056: Actually, we have considered the intersection of
1057:  $\, C_2^{31}\, = \,0\, $ and $\, C_4^{22}\, = \, 0$,
1058: that we will denote symbolically 
1059: $\,  C_2^{31} \cap  C_4^{22}$, as well as the intersection
1060:  $\,  C_2^{13} \cap  C_4^{31}$.
1061: We obtained the following
1062: generating function in agreement with the successive degrees (up to $t^9$)
1063: in the corresponding iteration:
1064: \begin{eqnarray}
1065: \label{gener2}
1066: && G_{C_2^{13} \cap  C_4^{31}} \, = \, \,1+2\,t+4\,{t}^{2}
1067: +7\,{t}^{3}+13\,{t}^{4}
1068: +24\,{t}^{5}+43\,{t}^{6}+77\,{t}^{7}\nonumber \\
1069: && \qquad \quad \quad +138\,{t}^{8}+247\,{t}^{9}
1070: \, + \cdots   \, = \, \,{\frac 
1071: {1-{t}^{3}}{1-2\,t+{t}^{4}+{t}^{6}-{t}^{8}}}
1072: \end{eqnarray}
1073: 
1074: Keeping in mind the shift of $\, +1$
1075: between (\ref{familN}) and (\ref{shift}), one might expect a
1076: formula like (\ref{familMN}) for
1077: an intersection $\, C_M^{13} \cap  C_N^{31}$ (or $\, C_M^{31} \cap  
1078: C_N^{22}$)
1079: \begin{eqnarray}
1080: \label{familMNbis}
1081: Q_{M,N} \, = \,\,\, \,  1\,-2\,t\,+{t}^{M+2}\,+{t}^{N+2}\,-{t}^{M+N+2} 
1082: \end{eqnarray}
1083: This is actually the case with the  previous example
1084: (\ref{gener2})
1085: where one has $\, M\, = 2$ and $\, N \, = \, 4$. Another example, also in agreement with
1086: (\ref{familMNbis}),  corresponds to the intersection
1087:  $\,  C_4^{31} \cap  C_6^{13}$ 
1088: for which one gets a rational degree generating function with 
1089: denominator
1090: $1-2\,t+{t}^{6}+{t}^{8}-{t}^{12}$.
1091: 
1092: 
1093: Note that such formula seems to remain valid even when $\, M \, = \, N$. 
1094: For instance,
1095: for $C_4^{13} \cap  C_4^{31}$ the denominator 
1096: of the generating function reads $\, 1-2\,t+2\,{t}^{6}-{t}^{10}$,
1097: and for $C_6^{13} \cap  C_6^{31}$ the denominator 
1098: reads $\,1-2\,t+2\,{t}^{8}-{t}^{14}$,
1099: in agreement with (\ref{familMNbis}) for $\, M \, = \, N\, = \, 6$. 
1100: 
1101: One sees that one has exactly the same complexity reduction scheme, and 
1102: {\em the same family of complexity},
1103: as the one depicted in Section (\ref{Diller}) for~\cite{Noether}.
1104: 
1105: However, one does see a difference with the
1106:  intersection of three conditions.
1107: For mapping (\ref{BiC2}), we saw~\cite{Noether}
1108: that the intersection of three conditions
1109: $\, a=\, (N-1)/N$, $\, b=\, (M-1)/M$, $\, c \, = 2-(a+b)=\, P/(P+1)$, 
1110: yields
1111: systematically {\em integrable} mappings.
1112: Here the $(a, \, b)$ points
1113:  corresponding to  intersection of three conditions
1114:  $\,C_N^{ij}\, = \, 0$ when they exist,  may still yield an 
1115: exponential growth of the calculations of lower complexity:
1116: \begin{eqnarray}
1117: \label{gener7}
1118: && G_{C_6^{13} \cap  C_6^{31} \cap C_8^{22} } \, = \, \,
1119: 1+2\,t+4\,{t}^{2}+8\,{t}^{3}+14\,{t}^{4}+24\,{t}^{5}+40\,{t}^{6}+66\,{
1120: t}^{7} \nonumber \\
1121: && \qquad \quad \, +108\,{t}^{8}
1122: \, + \cdots \,  \, = \, \,{\frac {1+{t}^{3}}{ \left( 1-t \right)  
1123: \left( 1-t-{t}^{2} \right) }}
1124: \, = \, \,{\frac {1+{t}^{3}}{1-2\,t+{t}^{3}}}
1125: \end{eqnarray}
1126: We have a similar result for the intersection of the three curves 
1127: $\, C_2^{31} \cap  C_6^{31} \cap C_{10}^{31}$
1128: with a denominator reading $\,1-2\,t+{t}^{4}$.
1129: 
1130: One should remark, in contrast with most of the
1131: degree growth rate calculations we have performed for so
1132: many birational transformations~\cite{complex}, that one can hardly 
1133: find rational values for the two parameters $\, a$ and $\, b$, 
1134: lying on the various $\, C_N^{ij}$'s we have just considered,
1135: (and of course it is even harder for intersections of such algebraic curves), 
1136: such that one would deal with iterations of birational transformations with
1137: integer coefficients,
1138: and factorization of polynomials with integer coefficients. 
1139: Such $\, (a,\,  b)$ points on $\, C_N^{ij}$ algebraic curves or 
1140: intersections of such curves, are {\em algebraic numbers}. 
1141: The degrees of the successive iterates should correspond
1142: to factorizations performed in some field extension
1143: corresponding to these algebraic numbers and curves. In practice, results and
1144: series like the ones displayed above ((\ref{shift}), ..., (\ref{gener7})),
1145: cannot be obtained this way.
1146: To achieve these factorizations, we have introduced a ``floating''
1147: factorization method that is described in Appendix B.
1148: 
1149: 
1150: \subsection{Degree growth complexity versus topological entropy }
1151: \label{topol}
1152: 
1153: The topological entropy is related to 
1154: the growth rate of the number of fixed
1155: points of $\, K^N$ (see~\cite{rearea}).
1156: The counting of the number of primitive cycles of order $\, N$,
1157: for the generic case $[4, 1, 2, 3, 6, 9, 18, 30, \cdots \,\, ]$
1158: gives a {\em rational} dynamical zeta function~\cite{zeta}
1159: \begin{eqnarray}
1160: \label{generic}
1161:  \zeta_g(t) \, = \, \, \,     {{ 1} \over {\left( 1-2\,t \right)
1162:  \left( 1-t \right) ^{2} }} 
1163: \end{eqnarray}
1164: which is related to the homogeneous degree generating  
1165: function $G(K)$ by the identity:
1166: \begin{eqnarray}
1167: \label{rela1}
1168: {\frac{t}{\zeta_g}} \cdot {{d} \over {dt}} \zeta_g  \,\,  = \, \,  \, \,   
1169: 2\, G(K)(t) \, + \, \, {\frac {2\, t}{1-t}}
1170:  \,\,  = \, \,\,   \, {\frac {2\,t}{1-2\,t}}   + \, \,{\frac {2\, t}{1-t}}
1171: \end{eqnarray}
1172: 
1173: Restricted to the curve of complexity reduction
1174: $\, C_2^{22}(a, \, b) \, =\, 0$, the primitive fixed points become
1175: $[4, 1, 2, 2, 4, 5, 10, 15, 26, 42, \cdots\,\,]$ giving the rational dynamical zeta function:
1176: \begin{eqnarray}
1177: \label{C2zeta}
1178:  \zeta(t) \, = \, \, {\frac {1}{ \left(1-2\,t+{t}^{4} \right)  
1179: \left(1-t \right) ^{2}}}  
1180: \end{eqnarray}
1181: Again, note that this  dynamical zeta function is related to the homogeneous degree 
1182: generating  function $G(K)$
1183: (corresponding to $\, C_2^{22}(a, \, b) \, =\, 0$), by the identity:
1184: \begin{eqnarray}
1185: \label{rela2}
1186: {\frac{t}{\zeta}} \cdot {{d} \over {dt}} \zeta \,  \, = \, \,  \, \, 
1187: 2\,G(K)(t) \, + \, \, {\frac {2\, t}{1-t}} \,\,\, 
1188: = \, \, \, 
1189:  {{ 2\, t \cdot (1-2\, t^3) } \over { 1-2\,t+{t}^{4}}}
1190:  \, + \, \, {\frac {2\, t}{1-t}} 
1191: \end{eqnarray}
1192: 
1193: We thus see, with these two examples (and similarly to the results
1194:  obtained for the birational transformations~\cite{rearea,zeta}
1195: as well as the Noetherian mappings~\cite{Noether}), an {\em 
1196: identification
1197:  between the growth rate  of the number of fixed
1198: points of} $\, K^N$,  {\em and the growth rate of the degree of
1199:  the iteration} (previously studied (\ref{defKanti})), or
1200:  equivalently, the {\em  growth rate of the Arnold complexity}.  
1201: 
1202: Relations (\ref{rela1}) and (\ref{rela2}) are in agreement with a
1203: Lefschetz formula\footnote[4]{
1204: The Lefschetz formula is well defined 
1205: in the {\em holomorphic} framework (see page 419 in~\cite{Griffiths}),
1206: but  is much more problematic in the {\em non-holomorphic 
1207: case of birational transformations
1208: for which indeterminacy points take place}: in very simple words one could say
1209: that, in the Lefschetz formula, 
1210:  some fixed points are ``destroyed'' by the indeterminacy points.
1211:  A good reference is~\cite{Fourier}.}:
1212: \begin{eqnarray}
1213: \label{mudd}
1214:   \nu_N \, = \, \, \, d_N(K) \, + \, \,  d_N(K^{-1}) \, + \, \, 1 \, + \, \, 1
1215: \end{eqnarray}
1216: where $ \, \nu_N$ denotes the number of fixed points of $ \,K$ or  $ \,K^{-1}$,
1217:   $ \,d_N(K)$ denotes the degree of  $ \,K^N$,  $ \,d_N(K^{-1})$ the degree  $ \,K^{-N}$.
1218: This formula (\ref{mudd}) means that the  number of fixed points 
1219: is the sum of four ``dynamical degrees \cite{Fourier}'' 
1220: $ \, \delta_0 \, + \,  \delta_1 \, + \,  \delta_2 \, + \, 
1221:  \delta_3$. Dynamical degree $ \, \delta_0 $ is always equal to  $ \,+1$,  $ \, \delta_3 $ is
1222: the topological degree (number of preimages:  $ \, \delta_3 $ is equal to  $ \,+1$
1223: for a birational mapping), $ \, \delta_1 $ is the first dynamical degree (corresponding to
1224:  $ \,d_N(K)$) and $ \, \delta_2 $ is the second dynamical degree (corresponding to
1225:  $ \,d_N(K^{-1})$). 
1226: \vskip .1cm
1227: 
1228: 
1229: {\bf Remark 1:} Most of the physicists will certainly take for granted that
1230: the degree growth rate corresponding to the iteration of $\, K$ and its inverse $\, K^{-1}$
1231: identify: $ \,d_N(K) \simeq \, \lambda(K)^N,$ $ \, d_N(K^{-1}) \simeq \, \lambda(K^{-1})^N$,
1232: with $\,  \lambda(K) \, = \, \lambda(K^{-1})$. This is actually the case
1233: for all the birational transformations we have studied~\cite{zeta}.
1234: In the specific examples of this paper,
1235: this is, in the involutive case $\, \xi\, = \,a+b+c-1\,  = \,1$,
1236:  a straight consequence of the fact that $\, K$ and $\, K^{-1}$ are conjugated.
1237:  More generally,
1238: this fact can be proved for all birational transformations in $\, CP_2$,
1239: but certainly not
1240: for birational transformations in 
1241: $\, CP_n$, $\, n \, \ge 3$ (for instance birational transformations
1242: generated by products of more than two
1243:  involutions, or ``Noetherian'' mappings products of many
1244: collineations and Hadamard involutions~\cite{Noether}, such 
1245: that  $\, K$ and $\, K^{-1}$ are not conjugated).
1246: Appendix C provides a simple example of bi-polynomial transformation in $CP_3$ 
1247: such that $\,  \lambda(K) \, \ne  \, \lambda(K^{-1})$.
1248: \vskip .1cm 
1249: 
1250: {\bf Remark 2:} The very definition of the dynamical zeta function on
1251: $\, C_0(a, \, b)\, = \, 0$ is  a bit subtle, and problematic,  since the number
1252: of fixed points for $\, K^2$ (and thus $\, K^{2N}$) is actually {\em infinite} (one has 
1253: a {\em whole curve} (\ref{list2}) of {\em fixed points of order two}).
1254: Apparently, in that case where an infinite number of fixed
1255:  points of order two exist, one does not seem, beyond these
1256: cycles of order two,  to have primitive  cycles of even order.
1257: Introducing the dynamical zeta function as usual, from
1258: an infinite Weil product~\cite{zeta}
1259: on the cycles, and
1260: taking into account just the odd cycles, 
1261: one obtains (more details are given in Appendix D)  
1262: that this zeta function verifies a simple functional
1263: equation
1264: \begin{eqnarray}
1265:   \zeta(t^2) \, = \, \,
1266:  {{(1-2\,t) \cdot (1-t)^2} \over {(1+2\,t) \cdot (1+t)^2 }}
1267:  \cdot \zeta(t)^2
1268: \end{eqnarray}
1269: showing that, the complexity is still the generic
1270: $\, \lambda \, = \,2$ but,
1271: this time, with  an expression
1272:  which {\em  is not a rational function, 
1273: but some ``transcendental''expression}. 
1274: In order to have a Lefschetz formula (\ref{mudd}) remaining  
1275: valid, in such highly singled-out cases for dynamical zeta functions,
1276: one needs to modify the definition
1277: of the dynamical zeta function so that it is no longer 
1278: deduced from an infinite Weil product~\cite{zeta} formula
1279: on the {\em cycles}. To be more specific, this must be performed using 
1280: the so-called~\cite{Fulton} ``Intersection Theory'' which is a
1281: (quite involved) theory introduced to cope with isolated points, {\em as well
1282: as non-isolated points} (curves ...), introducing
1283: some well-suited (and subtle) concepts like the notion of {\em multiplicity}.
1284: All the associated counting of intersection numbers will, then, correspond
1285: to counting
1286: of {\em finite integers} (replacing the counting of cycles ...).
1287: This is far beyond the scope of this very paper. 
1288: 
1289: \section{Preserved meromorphic two-forms in particular subspaces $(a,b)$}
1290: \label{covar}
1291: 
1292: In Appendix E, we show that the degree growth (in the $(a, \, b)$ parameters) for
1293: the iterates of the \emph{three curves of the critical set} (resp.
1294: exceptional locus)
1295: when the parameters are restricted to $b=a$, $b=2-2a$, $b=\, 1-2a$,
1296: and $C_0(a,b)=0$, is {\em polynomial} ($\delta=1$).
1297: The iterates are found in closed expressions.
1298: Let us show that, in these cases, the mapping $K$ 
1299: preserves simple meromorphic two-forms.
1300: 
1301: On the three lines $\, b=a$, $b=\, 2\, -2a\,$, and $ b=1-a/2$,
1302:  one finds three preserved meromorphic
1303:  two-forms reading respectively:
1304: \begin{eqnarray}
1305: \label{list}
1306: && {{ du \cdot dv} \over
1307:  { (u-1) \cdot ( 2\, (2a-1)(u+v^2)\, +(5a-4) (1+u)\, v )}} 
1308: \, = \,\,  idem(u',\, v')  \nonumber \\
1309: &&  {{ du \cdot dv}
1310:  \over {(v-1) \cdot ( (5a-4) (1+v)u+2(2a-1)(v+u^2) )}}
1311:  \, =\,  \, idem(u', \, v')  \nonumber \\
1312: && {{ du \cdot dv} \over 
1313: { (v-u) \cdot ( 4(a-1) (1+uv)+(5a-2)(v+u) )}} \, = \, \, idem(u',\, v')
1314: \end{eqnarray}
1315: The second, and third, two-forms are obtained from the first one in (\ref{list})
1316: by respectively
1317: $(u,v)\rightarrow (v,u)$ for $b=2-2a$, and by
1318: $(u,v) \rightarrow (u/v,1/v)$ with $a \rightarrow 1-a/2$, for $b=1-a/2$.
1319: For the quadratic condition $C_0(a,b)=0$, the mapping preserves the following two-form,
1320: {\em up to a minus sign}:
1321: \begin{eqnarray}
1322: &&   {{ du \cdot dv} \over { \rho(u, \, v) }}\,
1323: = \, \, \, - \, {{ du' \cdot dv'} \over { \rho(u', \, v') }},
1324:  \quad \quad \quad \quad \quad  \hbox{where :}\nonumber \\
1325: \label{ellipt}
1326: && \rho(u, \, v)\, = \, \,\,\,
1327:  (b-a)(a^2+b^2+3ab)\,  (1+u^2)\, v \nonumber \\
1328: &&\quad \quad \quad -(2b+a)(a^2-b^2-ab)\, (1+v^2)\, u \label{list2} \\
1329: && \quad \, \,   \quad -(b+2a)(a^2-b^2+ab) \, (u^2+v^2)
1330: +2(b-a)(2a+b)(a+2b) \, uv \nonumber
1331: \end{eqnarray}
1332: Note that $\,\rho(u, \, v)\, = \, 0$ is an elliptic curve.  
1333: \vskip.2cm
1334: 
1335: Considering the 25 points $(a, \, b)$, listed in Appendix F,
1336: for which the mapping is integrable, one can see that they all belong
1337: to the codimension-one subvarieties
1338: of the $(a, \, b)$ plane, where preserved meromorphic
1339: two-forms are found, i.e. 
1340: the curve $C_0(a,b)=0$ and/or the lines $b=a$, $b=2-2a$, $b=1-a/2$
1341: (see Figure \ref{f:fig1}, lower left corner).
1342: 
1343: 
1344: Furthermore, when these codimension-one subvarieties
1345: intersect, the deduced $(a, \, b)$ points correspond to  integrability of
1346: the mapping. 
1347: The algebraic
1348: invariants corresponding to these  integrability cases, can easily
1349: be deduced from the fact that, at the intersection
1350: of two curves among $\,C_0(a,b)=0$, and the lines $b=a$, $b=2-2a$, $b=1-a/2$,
1351: {\em one necessarily has two simple two-forms preserved} (up to a sign).
1352: Performing the ratio of two such two-forms one immediately
1353: gets algebraic invariants of the integrable mapping.
1354: See Appendix F for examples of algebraic invariants deduced, for
1355: integrable points $(a, \, b)$, from 
1356: ratio of two preserved two-forms.
1357: 
1358: \vskip .1cm
1359:  {\bf Remark:} One may have the feeling that the exact results
1360: on preserved meromorphic two-forms, or in the previous sections
1361: on complexity reduction for (\ref{defKanti}), are consequences 
1362: of the fact that we restricted ourselves to $\, \xi \, = \, a+b+c-1\, = \, 1$,
1363: the condition for collineation $\, C$ to be involutive (yielding $\, K$ and $\, K^{-1}$ 
1364: to be conjugate). This is not the case. We give in Appendix G
1365: miscellaneous examples of exact results valid when this
1366: involutive condition on $\, C$ is not verified ($\, \xi \, = \, a+b+c-1\, \ne \, 1$).
1367: 
1368: \vskip .1cm
1369:  
1370: It is tempting, after such an accumulation
1371: of preserved two-forms, to see the previous results
1372: (\ref{list}), (\ref{list2}) as a restriction to these codimension-one
1373: subvarieties  in the ($a, \, b)$-plane,
1374: of a general (conformally preserved) meromorphic 
1375: two-form valid in the whole $(a, \, b)$-plane. In view of the
1376: expressions of the two-forms for the three lines on one side, and the 
1377: expression associated with the elliptic curve (\ref{ellipt})
1378: on the other side, one could expect, at first sight, this
1379:  meromorphic two-form to be quite involved.
1380:  Using a "brute force" method we
1381:  have tried to seek, systematically, for meromorphic two-forms
1382:  $\, d\mu(u, \, v)=du \cdot dv \,/\rho(u, \, v)$, with
1383: an algebraic (polynomial) covariant $\, \rho(u, \, v) $ in the form:
1384: \begin{eqnarray}
1385: \label{rho12}
1386: \rho(u, \, v) \, =  \, \, \, \,
1387: \sum_{i=0}^{n_1} \sum_{j=0}^{n_2} c_{ij}\, u^i \, v^j .
1388: \end{eqnarray}
1389: The existence of such a polynomial covariant curve is {\em ruled out}
1390: up to $n_1=n_2=18$.
1391: Formal calculations seem hopeless here, in particular if the final 
1392: result is a non existence of such an algebraic covariant
1393:  of (\ref{defKanti}) for generic $a, \, b$.
1394: One needs to develop another approach that
1395:  might be also valid to prove a ``no go''
1396: result like the {\em non existence} of an {\em algebraic} covariant
1397: $\, \rho(u, \,v) $, and beyond, the non existence of
1398: a ``transcendental'' covariant
1399: $\, \rho(u, \,v)$, corresponding to some  {\em analytic but not algebraic} 
1400: curve\footnote[5]{Along this line, one should recall the occurrence of a {\em transcendental
1401: invariant} for a birational mapping given by the ratio of
1402:  products of simple Gamma functions, providing
1403: an example of ``transcendental'' integrability (see equation (31), paragraph 7
1404: of~\cite{Noether}, or 
1405: equation (20), paragraph (8.3)
1406:  in~\cite{classif}, or equation (3.3) in~\cite{SIDEV}).
1407: }. 
1408: 
1409: 
1410: \section{Orbit of the critical set: algebraic curves versus chaotic sets}
1411: \label{from}
1412: 
1413: When a preserved (resp. conformally preserved)  meromorphic 
1414: two-form  $du.dv/\rho(u,\, v)$ exists,
1415: one has the following fundamental relation (\ref{jac}) between the 
1416: algebraic expression $\rho(u,\, v)$ and the Jacobian of transformation  $\,K $:
1417: \begin{eqnarray}
1418: \label{fundam}
1419: \rho \left( K(u,\, v) \right) \,\, = \, \,\, \xi \cdot J(u,\, v) \cdot
1420: \rho(u,\, v)
1421: \end{eqnarray}
1422: where $\, \xi $ is a constant. When $\, \xi \, = \, +1$ the two-form
1423:  is preserved.
1424: When there exists an integer $\, M$, such that $\, \xi^M \, = \, 1$,
1425: the transformation $\, K^M$, instead of $\, K$,  preserves a   
1426: two-form.  When $\, \xi \, \ne \, +1$
1427: (for any $M$, $\, \xi^M \,\ne \, +1\,$),
1428: it is just conformally preserved. 
1429: Let us restrict the previous fundamental relation (\ref{fundam})
1430: to a point $ (u,\, v)$ such that the Jacobian of transformation $\, K$
1431: vanishes,  $\, J(u,\, v)\, = \, 0$. The fundamental relation 
1432: (\ref{fundam}) necessarily yields for such a point:
1433: \begin{eqnarray}
1434: \label{fundam2}
1435:   \rho \left( K(u,\, v) \right) \,\, = \, \,\, 0
1436: \end{eqnarray}
1437: 
1438: 
1439: For birational transformations, the images of the 
1440: curves $\, J(u,\, v)\, = \, 0$
1441: are  not  curves but {\em blow down into set of points}. For mapping 
1442: (\ref{defKanti}), the vanishing condition 
1443: $\, J(u,\, v)\, = \, 0$  splits into three curves 
1444: $u=0$,$\, v=0$ and $u=-av/(v(a-1)+a)$. The image of these three curves
1445: {\em blow down} into three points $(u^{(1)}, \, v^{(1)})$,
1446: $(u^{(2)}, \, v^{(2)})$ and $(u^{(3)}, \, v^{(3)})$.
1447: Being covariant, $ \rho(u,\, v)$ not only vanishes at these points
1448: (i.e. $ \rho(u^{(i)},\, v^{(i)})=0$ for $i=1, \, 2, \, 3$),
1449: but {\em also on their orbits}:
1450: \begin{eqnarray}
1451: \label{fundam3}
1452: \rho \left( K^N(u^{(i)},\, v^{(i)}) \right) \,\, = \, \,\, 0, \qquad \quad \quad
1453:  N \, = 1, \, 2, \, \cdots, \,i=1, \, 2, \, 3  
1454: \end{eqnarray}
1455: One can thus construct a (generically) {\em infinite set} of points on
1456: $ \rho(u,\, v)=0$,
1457: as orbits of such ``singled-out'' points
1458:  $(u^{(i)},\, v^{(i)})$ and visualize them, whatever (the accumulation of) this set of 
1459: points is (algebraic curves, transcendental analytical curves, chaotic 
1460: set of points, ...). 
1461: 
1462: Before visualizing some orbits, let us underline that (\ref{fundam3})
1463: means that the iterates of the critical set, also called {\em  post-critical set},
1464: actually  cancel $\rho \left(u,\, v \right)$. These 
1465: iterates are known in closed
1466: forms for some subspaces. For instance, on the line $b=a$, the iterates
1467: are given in Appendix E in terms of Chebyshev polynomials.
1468: At these iterates $\left(u^{(i)}_N,\, v^{(i)}_N \right)$, with closed expressions,
1469: one has $\rho \left(u^{(i)}_N,\, v^{(i)}_N \right)=0$.
1470: 
1471: The meromorphic two-forms found in Section (\ref{covar}) (see (\ref{list})),
1472: actually correspond to situations such that the post-critical set (resp. the orbit of the
1473: exceptional locus) has $\, \delta\, = \, 1$, closed expressions 
1474: being available to describe all these points (Chebyshev polynomials, ...). 
1475: The generic exponential growth (in the parameters) of the $\left(u^{(i)}_N,\, v^{(i)}_N \right)$ 
1476: (namely $\, \delta \sim 2$), certainly excludes (even very involved)
1477: algebraic
1478: expressions (\ref{rho12}) for $\rho(u, \, v)$, but it may not exclude 
1479: {\em transcendental analytical curves} (like the transcendental curves (31)
1480:  in paragraph 7 of \cite{Noether}, or the transcendental curves (20)
1481:  in \cite{classif},  which are orbits of a birational transformation exhibiting some
1482: ``transcendental'' integrability~\cite{Noether,SIDEV}.)
1483: 
1484: \subsection{Visualization of post-critical sets}
1485: \label{visupost}
1486: \begin{figure}
1487: \psfig{file=NG2.ps,scale=0.6,bbllx=100bp,bblly=100bp,bburx=600bp,bbury=600bp}
1488: \vskip .9 cm
1489: \caption{Orbit of the critical set for $(a, \, b) \, = \, (0, \, 1.9)$}
1490: \label{f:fig19}
1491: \end{figure}
1492: 
1493: \begin{figure}
1494: \psfig{file=NG9.ps,scale=0.6,bbllx=100bp,bblly=100bp,bburx=600bp,bbury=600bp}
1495: \vskip .9 cm
1496: \caption{Orbit of the critical set for $(a, \, b) \, = \, (-.2, \, .7)$}
1497: \label{f:fig191}
1498: \end{figure}
1499: 
1500: 
1501: Let us visualize a few post-critical sets.
1502: In the cases  where a meromorphic two-form is actually preserved (see (\ref{list})),
1503: one easily  verifies that the orbit of $(u^{(1)}, \, v^{(1)})\, = \,(b/a, \, \, (1-a-b)/a)$,
1504: actually yields  the (whole) covariant condition $\, \rho(u, \, v)\, = \, 0$ 
1505: corresponding to the divisor of a meromorphic two-form 
1506: when such a meromorphic two-form has been found. Of course if one 
1507: performs iterations of other points (even very close) than
1508: the singled out points as 
1509: $\,(b/a, \, \, (1-a-b)/a)\,$, one will not get 
1510: such algebraic covariant curve $\, \rho(u, \, v)\, = \, 0$, but 
1511: more involved orbits. In contrast for parameters $(a, \, b)$ for which no
1512: meromorphic two-form
1513: was found, we see a drastically different situation, shown in
1514: Figure \ref{f:fig19}, corresponding to the 
1515: orbit of  $\, (b/a, \, \, (1-a-b)/a)$ (image by 
1516: transformation $\, K$ of one of the vanishing conditions for the Jacobian)
1517: for $(a, \, b) \, = \, (0, \, 1.9)$.
1518: 
1519:  This post-critical set (Figure \ref{f:fig19}) looks very much like a set of curves, 
1520: a ``foliation'' of the $\, (u, \, v)$-plane.
1521: Figure  \ref{f:fig191} shows the post-critical set corresponding to the
1522: case
1523: $(a, \, b) \, = \, (-.2, \, .7)$.
1524: With these two orbits it is quite clear that this set of points
1525: {\em cannot be} a
1526: simple algebraic curve $\, \rho(u, \, v)\, = \, 0$. 
1527: 
1528:  At this step the ``true'' nature of this set of points
1529:  is almost a ``metaphysical'' question: 
1530: is it a transcendental analytical curve infinitely winding, is it a chaotic 
1531: fractal-like set ... ?  In particular when one takes a larger frame for plotting the orbit, 
1532: the set of points becomes more fuzzy, and it becomes more and more difficult,
1533: to see if these points are organized in curves, like Figure \ref{f:fig19}
1534: which
1535: suggests an (infinite ...) accumulation of curves. 
1536: We have encountered many times such a situation (see paragraph 5.1
1537: and Figures 13 and 14 in~\cite{ab-an-bo-ma-2000}). A way to cope with the 
1538: fuzzy appearance of the orbit when the points go to infinity, is to perform
1539:  a change of variables (see paragraph 5 in~\cite{ab-an-bo-ma-2000}): 
1540: $(u, \, v) $ $\, \rightarrow \, $$(1/u, \, 1/v) $. 
1541: Again one has the impression to see some kind of ``foliation of curves''
1542: for the previously fuzzy points, but the 
1543: points that were seen in Figure \ref{f:fig19} as organized like
1544:  a ``foliation'' of curves, have now 
1545: (in some kind of ``push-pull game'') become fuzzy
1546:  sets. One way to avoid this ``push-pull'' problem,
1547:  and thus, ``see the global picture'' amounts to performing our plots 
1548: in the variables $\, u_c  \,= \,  u/(1+u+u^2)$ and 
1549: $\, v_c  \,= \,  v/(1+v+v^2)$. These variables are such
1550: that any orbit of {\em real} points
1551:  will be in the box $\,[ -1, \,1/3] \times [ -1, \,1/3]$.
1552: This trick {\em ``compactifies''automatically our orbits}. 
1553: \begin{figure}
1554: \psfig{file=NOrb3.ps,scale=0.6,bbllx=100bp,bblly=100bp,bburx=600bp,bbury=600bp}
1555: \vskip .9 cm
1556: \caption{The orbit of the critical set in the ``compact'' 
1557: variables $\, u_c$ and  $\, v_c$, for $\, (a, \, b)  \, = \,( -.2, \,  .72)$}
1558: \label{f:fig11}
1559: \end{figure}
1560: \begin{figure}
1561: \psfig{file=NG7.ps,scale=0.6,bbllx=100bp,bblly=100bp,bburx=600bp,bbury=600bp}
1562: \vskip .9 cm
1563: \caption{The orbit of the critical set in the  ``compact'' 
1564: variables $\, u_c$ and  $\, v_c$, for $\, (a, \, b)  \, = \,( -.2, \,  .999999999)$}
1565: \label{f:fig7}
1566: \end{figure}
1567: 
1568: Let us give two examples of such  ``compactified'' images of two 
1569: orbits of $\, (b/a, \, \, (1-a-b)/a)$ (image by 
1570: transformation $\, K$ of one of the vanishing conditions for the Jacobian).
1571: For $\, (a, \, b)  \, = \,( -.2, \,  .72)$ one gets 
1572: Figure \ref{f:fig11}, and for $\, (a, \, b)  \, = \,( -.2, \,  .999999999)$ one gets 
1573: Figure \ref{f:fig7}.
1574: 
1575: In the situation where preserved meromorphic two-forms exist,
1576: one sees that, even a very small deviation from the
1577:  $\, (b/a, \, \, (1-a-b)/a)$ point 
1578: (associated with the post-critical set), yields orbits that 
1579:  look {\em quite different} from the algebraic covariant curve 
1580: $\, \rho(u, \, v)\, = \, 0$. In contrast with this situation, we see, in the previous
1581: cases where no preserved meromorphic two-forms exist, that a slight modification
1582: of the  $\, (b/a, \, \, (1-a-b)/a)$ point (associated with the post-critical set)
1583: yields  orbits which are {\em extremely similar} to the post-critical set 
1584: of Figures (\ref{f:fig191})
1585: or (\ref{f:fig11}), or (\ref{f:fig7}). These orbits are ``similar'',
1586:  but {\em not converging towards 
1587: this post critical set}. They are, roughly speaking, ``parallel''
1588: to this post critical set. Therefore
1589:  the orbit of the critical set
1590: may be seen as a chaotic set, but it is a {\em  non attracting chaotic 
1591: set} in contrast with the well-known strange attractors of H\'enon bi-polynomial
1592: mappings~\cite{Bene,Vadim}.  
1593: 
1594: 
1595: \subsection{From preserved meromorphic two-forms and post-critical sets
1596: back to fixed points}
1597: \label{fromto}
1598: Denoting $(u',\, v')$, $(u'',\, v'')$, ..., $\, (u^{(n)}, \, v^{(n)})$, 
1599: the images of a point $ (u,\, v)$, by transformations
1600: $\, K$,  $\, K^2$, ...,  $\, K^n$,
1601: the preservation of a two-form yields
1602: ($ J[K^n](u, \, v)$ being the Jacobian of $\, K^n$):
1603: \begin{eqnarray}
1604: && {{du \cdot dv } \over {\rho(u, \, v)}} \, \, = \, \, \,
1605:   {{du' \cdot dv'} \over {\rho(u', \, v')}} \, \, = \, \,  \,
1606: \cdots \, \, = \,  \,\,  {{du^{(n)} \cdot dv^{(n)} } \over {\rho(u^{(n)}, \, v^{(n)})}} \nonumber \\
1607: &&J[K^n](u, \, v) \, \, = \, \, \,  {{ \rho(u^{(n)}, \, v^{(n)})} \over { \rho(u, \, v)}} 
1608: \, \, = \,  \,\, {{ \rho(K^n(u, \, v))} \over { \rho(u, \, v)}} 
1609: \end{eqnarray}
1610: From the previous relation, it is tempting to deduce (a little bit too quickly ...) 
1611: that the Jacobian of $\, K^n$ is equal to +1 when evaluated at the 
1612:  fixed point $(u_f, v_f)$ of $\, K^n$: 
1613: \begin{eqnarray}
1614: \label{toquick}
1615: J[K^n](u_f, \, v_f) \, = \, \, \, {{ \rho(K^n(u_f, \, v_f))} \over { \rho(u_f, \, v_f)}} 
1616:  \, = \, \, +1
1617: \end{eqnarray}
1618: We actually found such strong results for (\ref{keps}), and for 
1619: many other birational transformations
1620: (when, for instance, we evaluated
1621: precisely the number of $n$-cycles,
1622: to get the dynamical zeta function~\cite{zeta}),
1623: for
1624: which a meromorphic two-form was actually preserved. In fact, even when a 
1625: meromorphic two-form is preserved, relation (\ref{toquick}) (namely 
1626: the Jacobian of $\, K^n$ evaluated at a fixed point of  $\, K^n$,
1627:  is equal to +1), {\em may be ruled out} when the fixed 
1628: points of $\, K^n$ {\em correspond to divisors of the two-form}.
1629: If $\, \rho(u, \, v)$,
1630: corresponding to a preserved meromorphic two-form,
1631: is a rational expression $\, \rho(u, \, v)\, = \, P(u, \, v)/Q(u, \, v)$
1632: ($P(u, \, v)$ and $\, Q(u, \, v)$ are polynomials), the Jacobian of $\, K^n$,
1633: evaluated at a fixed point
1634: $\, (u_f, \, v_f)$ of  $\, K^n$, {\em can actually be different from} $\, +1$, if
1635: $P(u_f, \, v_f)\, = \, \, 0$, or $\, Q(u_f, \, v_f)\, = \, \, 0$. 
1636: Such ``non-standard'' fixed points of $\, K^n$ are such that $\, \rho(u_f, \, v_f) \, = \, \, 0$
1637: (resp. $\, \rho(u_f, \, v_f) \, = \, \, \infty$), and of course, since $\rho(u, \, v)$
1638: is typically a covariant of $\, K$ (see (\ref{fundam})), such that $\rho(u, \, v)$, evaluated at 
1639: all their successive images by $\, K^N$ (for any $N$ integer), {\em vanishes}
1640: (resp. is infinite):
1641: \begin{eqnarray}
1642: J[K^n](u_f, \, v_f)\, \ne \, +1, \quad 
1643: \Rightarrow \quad \rho(K^N(u_f, \, v_f)) \, = \, \, 0 \,\quad  \hbox{(resp. $\,\infty$) } 
1644: \end{eqnarray}
1645: Performing orbits of such ``non-standard'' fixed points could thus be
1646: seen as an {\em alternative way of visualization}
1647: of  $\rho(u, \, v)$ (whatever its ``nature'' is: polynomial, rational expression, analytic 
1648: expression, ... ), this alternative way being extremely similar to the 
1649: one previously described, associated with the visualization of post-critical
1650: sets.
1651: Finding by formal calculations a very large 
1652: accumulation of such  ``non-standard'' fixed points is not
1653: sufficient to
1654: prove the non-existence of meromorphic two-forms: one needs
1655: to be sure that this accumulation of points cannot be localized on some
1656: {\em unknown}
1657: highly involved algebraic curve. It is well known that
1658: proving  ``no-go'' theorems
1659: is often much harder than proving theorems that simply require to exhibit
1660: a structure.
1661: However, as far as this difficulty to prove a
1662:  non-existence is concerned, it can be seen as highly positive and effective,
1663:  as far as simple
1664: ``down-to-earth'' visualization methods are concerned. In contrast with the
1665: unique
1666: post-critical set, we can consider orbits of a large (infinite) number of
1667: such
1668: ``non-standard'' fixed  points of $\, K^n$.
1669: The relation between the post-critical set
1670: and such ``non-standard'' sets, is a very interesting one
1671: that will be studied elsewhere. 
1672: 
1673: Let us just consider the birational transformation (\ref{defKanti})
1674: for  $(a, \, b) =\, (-1/5, \, 1/2)$ (where no
1675: meromorphic two-form has been found).
1676: The primitive fixed points (cycles) and  the value
1677: of the  Jacobian of $\, K^n$ at the corresponding fixed points,
1678:  that we will denote $\, J$, are
1679:  given~\footnote[5]{In these tables $\vert J \vert=1$ means that,
1680: at the fixed point of $K^n$, the value of $J$  is complex and lying on the
1681:  unit circle. Similarly, $J \ne 1$ means that $\, J$ is real, $\vert J \vert \ne 1$
1682: that $\, J$ is not real.}
1683:  in Table 1.
1684: \vskip 5mm
1685: \centerline{
1686: \begin{tabular}{|l|l|l|l|l|l|l|l|l|}
1687: \hline
1688: $n$& 1& 2& 3& 4& 5& 6& 7& 8   \\ 
1689: \hline
1690: fix$(K^n)$& 4& 1& 2& 3& 6& 9& 18& 30   \\ 
1691: $J=1$& 0& 1& 0& 1& 0& 3& 0& 6   \\ 
1692: $\vert J \vert=1$& 2& 0& 2& 2& 4& 4& 6& 8   \\ 
1693: $J \ne 1$& 2& 0& 0& 0& 2& 2& 4& 4   \\ 
1694: $\vert J \vert \ne 1$& 0& 0& 0& 0& 0& 0& 8& 12   \\ 
1695: \hline
1696: \end{tabular}
1697: }
1698: \vskip 0.2cm
1699: \textbf{Table 1:} Counting of primitive cycles
1700:  for $a=-1/5,\,  b=1/2$. $\, J$ denotes $J[K^n](u_f, \, v_f)$.
1701: \vskip 0.3cm
1702: The number of cycles are in agreement with the Weil product expansion of 
1703: the known (see (\ref{generic}))
1704: exact expression of the dynamical zeta function:
1705: \begin{eqnarray}
1706: \label{weil}
1707: &&\zeta(t) \, \, = \, \,     {{ 1} \over {\left( 1-2\,t \right)
1708:  \left( 1-t \right) ^{2} }}\, \, = \, \,
1709: {\frac {1}{ \left(1-t \right)^{4} \left(1-{t}^{2} \right)
1710:  \left(1-{t}^{3} \right)^{2} \left(1-{t}^{4} \right)^{3}}} \, \times\nonumber \\
1711: &&\quad \quad \quad  \quad \times \,
1712:  {{1} \over {\left(1-{t}^{5} \right)^{6}
1713:  \left(1-{t}^{6} \right)^{9} \left(1-{t}^{7}
1714:  \right)^{18} \left( 1-{t}^{8} \right)^{30} \cdots }} 
1715: \end{eqnarray}
1716: To some extent, the situations where $\, J\, = \, -1$, or where $\, J$ 
1717: is an $N$-th root of unity, can be ``recycled'' into a  $\, J \, = \, 1$ situation, 
1718: replacing $\, K^n$ by $\, K^{2\, n}$ or $\, K^{N\, n}$.
1719: However, we see on Table 1, the beginning
1720: of a ``proliferation'' of  ``non-standard''points that cannot be reduced
1721: to $\, J\, = \, -1$
1722: or $\, J^N \, = 1$, strongly suggesting the non-existence of a 
1723: meromorphic two-form. These enumerations have to be compared with the ones 
1724: corresponding to   $(a, \, b) =\, (1/5, \, 1/5)$, for which a
1725: meromorphic two-form is actually preserved. We still have the same sequence $\, 4,\, 1, \, 
1726:  2, \, 3, \,  6, $$\,  9, \,  18,  \,  30 , \cdots$ of $\, n$-cycles, associated to the same
1727: dynamical zeta function (\ref{weil}), however (except for the fixed points of order one), 
1728: all the fixed points of order $\, n\, \ge 2$ are such
1729: that $\, J \, = \, 1$~\footnote[1]{Recall that mappings (\ref{keps}) and (\ref{BiC2}) for
1730: which a meromorphic two-form exists for generic values of the parameters,
1731: are such that $\, J \, = \, 1$ for all the fixed points we have computed.}.
1732: 
1733: Along this line, let us consider mapping $K$ on 
1734: curve $C_2^{22}(a,b)=0$, where, despite the complexity reduction,
1735: no meromorphic two-form
1736: has been  found. The calculations are performed
1737: for the (generic) values $a=12/13,\,  b=-3/13$ (the number of
1738: non generic $(a,b)$ on $C_2^{22}(a,b)$ is finite) and are given in Table 2.
1739: \vskip 5mm
1740: \centerline{
1741: \begin{tabular}{|l|l|l|l|l|l|l|l|l|l|l|}
1742: \hline
1743: $n$&1&2&3&4&5&6&7&8&9&10   \\ 
1744: \hline
1745: fix$(K^n)$&4&1&2&2&4&5&10&15&26&42   \\ 
1746: $J=1$&0&1&0&0&0&1&0&3&0&6   \\ 
1747: $\vert J \vert=1$&2&0&0&0&2&2&0&2&4&2   \\ 
1748: $J \ne 1$&2&0&2&2&2&2&6&6&10&14   \\ 
1749: $\vert J \vert\ne1$&0&0&0&0&0&0&4&4&12&20   \\ 
1750: \hline
1751: \end{tabular}
1752: }
1753: \vskip 0.3cm
1754: \textbf{Table 2:} Counting of primitive cycles
1755:  for $(a, b)$ such that $C_2^{22}(a,b)=0$. $\, J$ denotes $J[K^n](u_f, \, v_f)$.
1756: 
1757: \vskip 5mm
1758: The number of $ \, n$-cycles
1759: are of course,
1760: in agreement with the Weil product decomposition
1761: of the exact dynamical zeta function (\ref{C2zeta}). 
1762: We note the same proliferation of ``non-standard''
1763: fixed points of order $\, n$, in agreement with the non-existence
1764: of a preserved meromorphic two-form. This last result confirms 
1765: what we saw several times, namely the {\em disconnection between 
1766: the existence} (or non-existence) of a preserved meromorphic two-form
1767: and (topological) {\em complexity reduction} for a mapping. 
1768: 
1769: 
1770: 
1771: \subsection{Pull-back of the critical set : ``ante-critical sets'' versus post-critical sets}
1772: \label{antecrit}
1773: 
1774: 
1775: As far as visualization methods are concerned physicists, ``fortunately'', 
1776: perform iterations without being conscious of the potential 
1777: dangers: birational transformations have singularities and they 
1778: may proliferate\footnote[2]{Are our numerical iterations
1779: well-defined in some ``clean'' Zariski space, could ask mathematicians?} when performing iterations.
1780: More precisely,
1781: the critical set (vanishing conditions of the Jacobian) 
1782: is a set of curves, whose images, by transformation $\, \, K$,
1783: yield {\em points} and {\em not curves} (blow-down). 
1784:  In contrast, the images  by transformation  $\, \, K^{-1}\, $ (resp.  $\, \, K^{-N}$) 
1785: of these curves of the critical set (denoted in the following CS)
1786: give {\em curves}: we do not have any blow-down with $\, \, K^{-1}$ 
1787: (resp.  $\, \, K^{-N}$).
1788: This infinite set of curves obtained by iterating
1789: the critical set by $\, \, K^{-1}$, is such that, for some finite integer $\, N$,
1790: the image of these curves by  $\, \, K^{N}$ {\em will blow down into points,
1791:  after a finite number} $\, N$ of iterations. 
1792: Let us call this set, for obvious reasons, ``ante-critical set''.
1793: This ``ante-critical set'' is clearly
1794: a ``dangerously singular'' set 
1795: of points for the iteration of $\, K$. It is also a quite interesting
1796: set from the singularity analysis viewpoint~\cite{bimero}. In particular, among these 
1797: ``dangerous points'' associated with the infinite set of curves $\, \Gamma_N \, = \, K^{-N}(CS)$,
1798: some are singled-out (more ``singular'' ...): the points corresponding to 
1799: intersections of {\em two} (or more) such curves 
1800: $\, \Gamma_N$'s.  These singled-out points can, in fact, be obtained 
1801: by some simple ``duality'' symmetries from the points of the post-critical set. 
1802: Such ``ante-critical sets'', and their associated
1803:  consequences on the birational transformations $\, K$,
1804:  clearly require some further analysis that will be performed elsewhere.
1805: 
1806: 
1807: \section{Lyapunov exponents 
1808: and non-existence of meromorphic two-forms}
1809: \label{detector}
1810: \begin{figure}
1811: \psfig{file=lyapPTS.eps}
1812: \caption{Lyapunov exponents as a function of parameter $\, b$, for $\, a\, = \, 1/2$, the 
1813:  initial point being $\, (u, \, v) \, = \, \, (2, \, 3)$.}
1814: \label{f:fig4}
1815: \end{figure}
1816: \begin{figure}
1817: \psfig{file=sig.eps,scale=0.65,angle=-90,bbllx=50bp,bblly=50bp,bburx=600bp,bbury=450bp}
1818: \caption{Lyapunov exponents as a function of parameter $\, b$, for $\, a\, = \, 1/2$, the 
1819:  initial point being the image of the critical set.}
1820: \label{f:fig8}
1821: \end{figure}
1822: 
1823: The previous simple visualization approach can be confirmed by some Lyapunov
1824: exponents analysis.
1825: Let us consider 
1826: orbits of a given initial point 
1827: (for instance $\, (u, \, v) \, = \, \, (2, \, 3)$)
1828: under the iteration of birational transformation (\ref{defKanti})
1829: for parameter $\, a$ fixed (for instance  $\, a\, = \, 1/2$), 
1830: and for different values of the second parameter $\, b$, and let
1831: us calculate the corresponding Lyapunov exponent. One thus gets 
1832: the Lyapunov exponent (of what we can call a ``generic'' orbit) as a function 
1833: of parameter $\, b$. This simple analysis is an 
1834: easy down-to-earth way to detect
1835: drastic complexity reductions, the complexity being not the 
1836: topological complexity (like the topological entropy 
1837:   or the growth rate complexity) but a less universal (more probabilistic)
1838: complexity (like the metric entropy). 
1839: 
1840: 
1841: Figures (\ref{f:fig4}), (\ref{f:fig8}) show, quite clearly,  {\em non-zero and positive}
1842: Lyapunov exponents, such results being apparently valid, not only for the
1843: Lyapunov exponent
1844: corresponding to our singled-out orbit, the post-critical set (see Figure (\ref{f:fig8})), but, also, for 
1845: {\em every} orbit in the $(u, \, v)$-plane (see Figure (\ref{f:fig4})). With
1846:  this scanning in the $\, b$ parameter
1847: we encounter several times the singled-out cases where preserved meromorphic two-forms 
1848: exist ($a\, = b$, $\, C_0(a, \, b) \, = \, 0$, ..., see (\ref{list})), and we see that these
1849: specific points are singled-out on Figure (\ref{f:fig4}). 
1850: If instead of performing the orbit of an arbitrary point ($(u, \, v) \, = \, \, (2, \, 3)$)
1851: one calculates the Lyapunov exponent corresponding to the post-critical set
1852: one finds similar results with a quite high volatility (a value of $\, b$
1853: where the Lyapunov is a ``local'' maximum is quite close to a value where
1854:  the Lyapunov is almost zero). 
1855: 
1856: In order to better understand this volatility, we have performed specific Lyapunov exponents calculations
1857: restricted to the singled-out cases where preserved meromorphic two-forms 
1858: exist ($a\, = b$, $\, C_0(a, \, b) \, = \, 0$, ..., see (\ref{list})).
1859: In such cases we recover the situation we had~\cite{rearea}
1860:  with birational mapping (\ref{keps}), namely 
1861: the Lyapunov exponents are zero (or negative on the attractive fixed points)
1862: for all the orbits we have calculated (the positive non-zero 
1863: Lyapunov being possibly on some ``evanescent'' slim 
1864: Cantor set~\cite{BedDill,DillBed}, see section (\ref{2-formversus}), 
1865: that we have not been able to visualize numerically)
1866: and the orbits always look like curves. It is clear that computer
1867:  experiments like these, can hardly detect the slim
1868: and subtle
1869: Cantor sets corresponding to (wedge product) invariant measure described~\cite{BedDill,DillBed} by
1870: Diller and Bedford
1871: in such situations, associated with the narrow regions where non-zero positive Lyapunov 
1872: could be found: within such (extensive) computer experiments  
1873: we find, ``cum grano salis'',  that the  Lyapunov exponents are ``generically'' 
1874: ({\em as far as computer calculations are concerned ...}) zero. 
1875: 
1876: With this subtlety in mind, our computer experiments show clearly {\em non-zero positive} 
1877: Lyapunov exponents when there is {\em no preserved} meromorphic two-form and
1878: a total extinction of these Lyapunov exponents when such preserved meromorphic two-forms
1879: take place.
1880:  
1881: The occurrence of non zero positive Lyapunov exponents for hyperbolic systems, or 
1882: dynamical systems with strange attractors is well-known: this is not the 
1883: situation we describe here.
1884: 
1885: 
1886: \section{Conclusion}
1887: \label{concl}
1888: 
1889: 
1890: The birational transformations in $\, CP_2$,
1891: introduced in section (\ref{antisto}),
1892: which  generically do not preserve any meromorphic two-form,
1893: are extremely similar to other birational transformations we previously studied~\cite{Noether},
1894: which do preserve meromorphic two-forms. We note that these two sets of 
1895:  birational transformations exhibit totally
1896:  similar\footnote[3]{In fact identical
1897: results: one gets the same family of polynomials controlling the
1898:  complexity (see (\ref{shift}) or (\ref{familMNbis}) and
1899:  compare with~\cite{Noether}).} results as far as 
1900:  {\em topological complexity} is concerned (degree growth complexity, Arnold complexity
1901:  and  topological entropy), but {\em drastically different numerical} results as far as a
1902:  more ``probabilistic'' (ergodic) approach of
1903:  dynamical systems is concerned (Lyapunov exponents).
1904: With these examples we see that the existence, or non-existence,
1905: of a preserved meromorphic two-form explains most of the (disturbing) 
1906: apparent discrepancy, we saw, numerically,
1907: between the topological and probabilistic approaches of such
1908: dynamical systems.
1909: 
1910: The situation is as follows. When these birational mappings
1911: preserve a mero-morphic two-form (conservative reversible case)
1912: the (preliminary) results of  Diller and 
1913: Bedford~\cite{BedDill,DillBed} on mapping (\ref{keps}) give a 
1914: strong indication (at least in the region of the parameter
1915: $\, \epsilon \, < \, 0$ )
1916: that the regions where the chaos is concentrated, namely where 
1917: the Lyapunov exponents are non-zero and positive, are quite evanescent,
1918: corresponding to an extremely slim Cantor set
1919:  associated with an invariant measure given by some wedge product.
1920: This nice situation from a differential viewpoint (existence of a 
1921: preserved two-form), is the unpleasant
1922: one from the {\em computer experiments viewpoint}: it is extremely hard to see
1923: the ``chaos'' (homoclinic tangles, Smale's horseshoe, ...)
1924: from the analysis (visualization of the orbits, Lyapunov exponents
1925: calculations, ...) of even very large sets of {\em real} orbits. 
1926: 
1927: On the contrary, when the birational mappings
1928: do not preserve a meromorphic two-form, the regions 
1929: where the Lyapunov exponents are non-zero, and {\em positive},  can, then,
1930: clearly be seen on computer experiments. 
1931: 
1932: In conclusion, the existence, or non-existence, of preserved meromorphic
1933: two-forms
1934: has (curiously) no impact on the topological complexity of the 
1935: mappings, but drastic consequences
1936: on the {\em numerical appreciation} of the ``probabilistic'' (ergodic) complexity. 
1937: 
1938: 
1939: The introduction of the {\em post-critical set}, namely the orbit of
1940: the points obtained by the blow-down of the curves corresponding to the
1941: vanishing conditions of the Jacobian of the birational transformation,
1942: thus emerges
1943: as a fundamental concept, and tool (of topological and algebraic nature)
1944:   to understand the 
1945: probabilistic (and especially numerical) subtleties of the dynamics of such
1946:  reversible~\cite{rob-qui-92,QuRo88} mappings. 
1947: 
1948: 
1949: \vskip .4cm 
1950: 
1951: \textbf{Acknowledgments}  We thank C. Favre 
1952: for extremely useful 
1953: comments on analytically stable birational transformations,
1954: exceptional locus and indeterminacy locus, and its cohomology 
1955: of curves approach of growth rate complexity.  We thank J-C. Angl\`es d'Auriac
1956: and E. Bedford for many  discussions on birational transformations.
1957: We also thank  J-P. Marco for interesting discussions on invariant 
1958: measures.
1959: (S. B) and (S. H) acknowledge partial support from PNR3.
1960: 
1961: 
1962: \section{Appendix A: Algebraic geometry: singularities
1963:  of curves as candidates for complexity reduction}
1964: \label{appendC}
1965: The conditions of reduced complexity give the points $(a,b)$ that
1966: belong to the algebraic curves $C_N$. 
1967: These algebraic curves are such that one has a reduced
1968:  complexity for generic point $(a,b)$ on the curve.
1969: However, singularities of these algebraic curves
1970: (from a purely algebraic geometry viewpoint: local branches, ...)
1971: can actually be seen to correspond to  points $(a, \, b)$ in the
1972: parameter plane {\em yielding lower  complexities}
1973: for the birational transformation $\, K$. 
1974: 
1975: 
1976: On each curve $C_N$, the spectrum of complexity at the singularities is
1977: given by
1978: \begin{eqnarray}
1979: \label{formula}
1980: 1-2t+t^{p+2}\, =\,\, \, \,0, \qquad \quad p\, =\, 0,\, 1,\,  \cdots,\,  
1981: N/2-2
1982: \end{eqnarray}
1983: For example, a generic point on the curve $C_8^{22}$, has the complexity growth
1984: $\lambda=1.9980$. The singularities of this curve are non generic points
1985: and have complexity growth $\lambda \simeq 1, 1.6180, 1.8392$ given by
1986: (\ref{formula}) for $N=8$. The next curve $C_{10}^{22}$ with $\lambda \simeq 1.9995$,
1987: will inherit the last three values and adds (since $p$ goes now to 3)
1988: $\lambda \simeq 1.9275$. Note that for a given curve $C_N$, the largest value of
1989: complexity growth reached by its singularities is given by $1-2t+t^{N/2}$.
1990: 
1991: 
1992: Let us give the generating functions of the degrees $\, d_N$, and
1993: genus $\, g_N$, of the successive  $C_N(a,\, b)=0$ algebraic curves.
1994: Let us also introduce the generating function for $\, S_N$,
1995: the number of singularities of the algebraic curves  $C_N$:
1996: \begin{eqnarray}
1997: d_C(t)  =  \sum_{n=1}^{\infty} d_{2\, n} \cdot t^{2\, n}, \quad 
1998: g_C(t)  =  \sum_{n=1}^{\infty} g_{2\, n} \cdot t^{2\, n} , \quad 
1999: s_C(t)  =  \sum_{n=1}^{\infty} S_{2\, n} \cdot t^{2\, n} \nonumber
2000: \end{eqnarray}
2001: They read respectively (for $C_N^{22}$ up to $N=12$):
2002: \begin{eqnarray}
2003: \label{dgCN}
2004: && d_C(t) \, = \,  \, \, \, \, \, 
2005: 2 \, t^2\, + \, 6 \, t^4 \, + \, 12 \, t^6 \, + \, 26 \, t^8 \, + \, 
2006: 48 \, t^{10} \, + \, \, 98 \, t^{12} \, + \, \cdots  \nonumber \\
2007: && g_C(t) \, = \, \, \, \, \, 
2008: 0 \, t^2\, + 5 \, t^4 \, + \, 20 \, t^6 \, + \, 73 \, t^8 \, + \, 
2009: 182 \, t^{10} \, + \, \, 491 \, t^{12} \, + \, \cdots  \nonumber\\
2010: && s_C(t) \,  = \, \, \, \, \, 
2011: 0 \, t^2\, + 5 \, t^4 \, + \, 15 \, t^6 \, + \, 31 \, t^8 \, + \, 
2012: 53 \, t^{10} \, + \, \, 113 \, t^{12} \, + \, \cdots  
2013: \nonumber
2014: \end{eqnarray}
2015: The degrees  $\, d_N$, the genus $\, g_N$, and the  number of singularities $\, S_N$
2016: clearly grow exponentially like
2017: $\, \lambda^{2n}$ with  $\, \lambda \, < \,  2$.
2018: We have no reason to believe that these three generating
2019:  functions $\, d_C(t)$,  $\, g_C(t)$ and  $\, s_C(t)$,  
2020: could be rational expressions. Similarly, their
2021:  corresponding coefficients growth rates, $\, \lambda$,
2022: have no reason, at first sight, to be  algebraic numbers. 
2023: 
2024: A singularity of an algebraic curve is characterized by  
2025: the coordinates of the singularities in homogeneous 
2026: variables, the multiplicity $m$, the delta invariant $\delta$ and
2027: the number of local branches $r$.
2028: In general $m \ge r$ and $\delta \ge m(m-1)/2$. The equality holds for 
2029: all the singular points of $C_N$, however, as $N$
2030:  increases, some points do not satisfy the equality.
2031: These points are $(a=0, b=1)$, $(a=1, b=0)$, $(a=1, b=1)$ and $(a=0, b=0)$,
2032: $(a=0,b=2)$, $(a=2, b=0)$.
2033: 
2034: \section{Appendix B: Computing  complexity growth
2035: of points known in their floating forms }
2036: \label{appendQ}
2037: 
2038: Let us show how to compute the complexity growth
2039: of generic (algebraic) points on algebraic curves, and 
2040: how to compute the complexity growth
2041: of points known in their floating forms.
2042: 
2043: 
2044: To compute the complexity growth for the parameters
2045: $(a,b)$ belonging to a whole curve, e.g. $C(a,b)=0$, we fix $v$ (for
2046: easy iteration), and we iterate up to order $N$.
2047: We eliminate $b$ between the numerator of $u_N-X$ and the curve $C(a,b)=0$.
2048: We can obtain factorizable polynomials $P_1 \cdot P_2 \cdots$
2049: One counts the degree of $u$ in the polynomials depending on $X$, and  
2050: discards the polynomials $P_i$ that contain only $u$.
2051: Let us show how this works.
2052: One considers the curve $C_2^{22}$ given in (\ref{C222}) and computes
2053: the complexity for the parameters $a$ and $b$ such that $C_2^{22}(a,b)=0$. 
2054: Let us fix $v$, and eliminate $b$ between 
2055: $u_N-X$ and $C_2^{22}(a,b)$ ($u_N$ is the $N$-th iterated, one 
2056: may take $v_N$ instead). One gets for the first four iterations
2057: $P(X^2, u^2)$, $P(X^2, u^4)$, $P(X^2, u^8)$ and
2058: $P(u^2) \cdot P(X^2, u^{14})$,
2059: where $P(u^n)$, $P(X^n, u^p)$ denote polynomials in $X$ and $u$ with the 
2060: shown degrees.
2061: At step $4$, a polynomial in $u$ factorizes, which means that
2062: the sequence of degrees in this case is $[1, 2, 4, 8, 14, \cdots]$
2063: instead of the generic $[1, 2, 4, 8, 16, \cdots \, ]$. 
2064: 
2065: The degrees of the curves grow as the iteration proceeds, we may need, 
2066: then, to compute the growth complexity for points in the $(a,b)$-plane
2067: {\em only known in their floating form}. We introduce a float numerical method
2068: that deals with these points obtained as roots of polynomials of degree
2069: greater than five.
2070: The method
2071: starts with the parameters in their floating forms. The iteration
2072: proceeds to order $N$, where one solves the numerator, and the denominator,
2073: of the variable (say) $u_N$. We take away the common roots and so on.
2074: The computation is controlled by the number of digits used. The
2075: computation with the float numeric method is carried out on the
2076: homogeneous variables. Let us show how the method works.
2077: The parameters $a$ and $b$ are fixed, and known, as floating numbers (with the desired 
2078: number of digits). The iteration proceeds as (in the homogeneous variables 
2079: $(x,y,t)$, where we may fix the starting values of $y$ and $t$):
2080: \begin{eqnarray}
2081: x & \rightarrow &\,  x_1=P_1^x(x)\, \rightarrow\,
2082:  x_2=P_2^x(x)\, \rightarrow \,\cdots \nonumber \\
2083: y & \rightarrow & \,y_1=P_1^y(x)\, \rightarrow \,
2084: y_2=P_2^y(x)\, \rightarrow \,\cdots \nonumber \\
2085: t & \rightarrow &\, t_1=P_1^t(x) \,\rightarrow \,
2086: t_2=P_2^t(x)\, \rightarrow\, \cdots \nonumber 
2087: \end{eqnarray}
2088: At each step, solving in  float each expression, amounts to writing:
2089: \begin{eqnarray}
2090: P_i^x(x)= \Pi_{j=1}^{n_1} (x-\tilde{x}_j), \, \,
2091: P_i^y(x)= \Pi_{j=1}^{n_2} (x-\tilde{x}_j), \,  \, 
2092: P_i^t(x)= \Pi_{j=1}^{n_3} (x-\tilde{x}_j), \nonumber 
2093: \end{eqnarray}
2094: The common (up to the fixed accuracy) terms $(x-\tilde{x}_j)$ between
2095: $P_i^x(x)$, $P_i^y(x)$ and $P_i^t(x)$ are taken away and
2096: the degree of, e.g., $P_i^x(x)$  {\em is counted according to this 
2097: reduction}.
2098: 
2099: \section{Appendix C: Degree growth complexity and the ``arrow of time''}
2100: \label{appendZ}
2101: Let us consider (after V. Guedj and N. Sibony~\cite{Sibony,Sibony2}) 
2102: the following bi-polynomial transformation:
2103: \begin{eqnarray}
2104: K(x, \, y, \, z)\,\, = \, \, \, \, \Bigl(z
2105: , \, \, y  -z^{d}, \, \,x\, + \, y^2 -2 \, y \, z^{d} \Bigr)\nonumber
2106: \end{eqnarray}
2107: Its inverse reads: 
2108: \begin{eqnarray}
2109: K^{-1}(x,y,z)\, =\, \, \Bigl(z-y^2+x^{2d}, \, y+x^d, \, x  \Bigr)\nonumber
2110: \end{eqnarray}
2111: Written in the homogeneous variables $u,v,w,t$,  transformation $\, K$, and its inverse,
2112: become:
2113: \begin{eqnarray}
2114: &&K(u,v,w,t)\, =\,\, \Bigl(wt^d, \, vt^d-t w^d, \, ut^d+v^2t^{d-1}-2vw^d, \, t^{d+1}  \Bigr)
2115: \nonumber \\
2116: &&K^{-1}(u,v,w,t)\,\, = \,\,\nonumber \\
2117: && \qquad \Bigl(wt^{2d-1}-v^2t^{2d-2}+u^{2d}, \,t^d \, ( vt^{d-1}+u^d),
2118: \, ut^{2d-1}-2vw^d, \, t^{2d}  \Bigr) \nonumber
2119: \end{eqnarray}
2120: 
2121: Fixing $d=1$, for heuristic reason, 
2122: the successive degrees of $K^n(u,v,w,t)$ read
2123: \begin{eqnarray}
2124: deg_u\,=\,\,deg_v\,=\,\,deg_w\,=\,\,deg_t\,=\,\,
2125: [2,3,5,8,13,21,34,55,\,  \cdots\, ]\nonumber
2126: \end{eqnarray}
2127: giving the degree generating function
2128: \begin{eqnarray}
2129: G(K)(t)\,= \,\,{\frac{t\cdot (t+2)}{1-t-t^2}}\nonumber
2130: \end{eqnarray}
2131: while the  successive degrees of $(K^{-1})^n(u,v,w,t)$ read
2132: \begin{eqnarray}
2133: deg_u\,=\,\,deg_v\,=\,\,deg_w\,=\,\,deg_t\,=\,\,
2134: [2,4,8,16,32,64,\,  \cdots\, ]\nonumber
2135: \end{eqnarray}
2136: and give the degree generating function:
2137: \begin{eqnarray}
2138: \label{GKmoinsun}
2139: G(K^{-1})(t)\,=\,\, {\frac{2t}{1-2t}}\nonumber
2140: \end{eqnarray}
2141: Transformation $\, K$ has clearly a golden number complexity 
2142: different, and smaller, than the complexity $\, \lambda \, = \, 2$ of its inverse.
2143: 
2144: \section{Appendix D: A transcendental zeta function ?}
2145: \label{appendF}
2146: 
2147: In this appendix, we consider the dynamical zeta function for the parameters ($a,b$) on
2148: $\, C_0(a, \, b)\, = \, 0$. This is a bit subtle since the number
2149: of fixed points for $\, K^2$ (and thus $\, K^{2N}$) is infinite (a 
2150: whole curve (\ref{list2}) is a curve of fixed points of
2151:  order two). Apparently one does not seem to have
2152: even primitive cycles (except the infinite number of two-cycles). 
2153: Introducing the zeta functions as usual by 
2154:  the infinite Weil product~\cite{zeta}
2155: on the cycles, avoiding the two-cycles and taking 
2156: into account just the
2157: odd primitive cycles one could write:
2158: \begin{eqnarray}
2159: \label{what}
2160:  1/\zeta(t) \,\, = \, \,\,
2161:  ( 1-t )^{4} ( 1-{t}^{3})^{2} 
2162: ( 1-{t}^{5})^{6}
2163:  (1-{t}^{7})^{18} (1-{t}^{9})^{56}
2164:  (1-{t}^{11})^{186} \cdots \nonumber
2165: \end{eqnarray}
2166: 
2167: 
2168: Recalling the ``generic'' expression (\ref{generic}), 
2169: this expression $\zeta(t)$ is such that
2170: $\zeta(t)\, \zeta_g(-t)=\zeta(-t)\, \zeta_g(t)$,
2171: and verifies the following functional relation 
2172: \begin{eqnarray}
2173:  \zeta(t) \, = \, \, \,\, {{1+t} \over {1-t} } \cdot
2174:  \Bigl({{1+2t} \over {1-2t} } \Bigr)^{1/2} \cdot  \zeta(t^2)^{1/2}\nonumber
2175: \end{eqnarray}
2176: yielding an infinite product expression for $\,  \zeta(t)$:
2177: \begin{eqnarray}
2178: \label{infprod}
2179: \zeta(t) \, = \, \, {{1+t} \over {1-t} } \cdot
2180: \prod_{i=0}^{\infty} 
2181: \Bigl({{(1+2\, t^{2^i}) (1+t^{2^{i+1}})} 
2182: \over {(1-2\, t^{2^i})(1-t^{2^{i+1}})} } \Bigr)^{1/2^{i+1}} \nonumber
2183: \end{eqnarray}
2184: 
2185: For $n$ as upper limit of the above infinite product, the expansion
2186: is valid up to $t^{2^{n+1}-1}$.
2187: The ratio of the coefficients of (for example)
2188: $t^{1023}$ with $t^{1022}$ gives
2189: $\, \lambda \, \simeq \,1.9989099 $, 
2190: in agreement with a complexity  $\, \lambda \, = \, 2$, but 
2191: {\em with a dynamical zeta function that {\em is not} a rational expression, 
2192: but some ``transcendental'' expression}. 
2193: 
2194: 
2195: Of course one can always imagine that the ``true'' dynamical zeta function
2196: requires the calculation of all the ``multiplicities'' of Fulton's 
2197: intersection theory~\cite{Fulton}, and that this very zeta function
2198: is actually rational ...
2199: 
2200: 
2201: \section{Appendix E: The mapping on the lines $b= \pm a$ and $C_0(a,b)=0$}
2202: \label{appendB}
2203: 
2204: Along  the line $b=a$ (and similarly on its equivalents obtained 
2205: by the actions of $P$ and $T$),
2206: the growth of the degrees of the parameter $a$ in the iterates
2207: of the vanishing conditions of the Jacobian is polynomial ($\delta=1$). 
2208: One, then, expects the iterates to be
2209: given in closed forms. This is indeed the case as can be seen below.
2210: The iterates $K^n(V_1)$  are given by
2211: \begin{eqnarray}
2212: &&K^n(V_1) = \Bigl(u_n, \,\, v_n \Bigr)\qquad \hbox { with: } \qquad 
2213: \sigma_1 = {\frac{3a^2-4a+2}{2(2a-1)}} \nonumber \\
2214: &&u_n  = \, \,  {\frac{2(2a-1)\, T_{n}(\sigma_1)+a(5a-4)\,U_{n-1}(\sigma_1)-2(2a-1)}
2215: {2(2a-1)\,T_{n}(\sigma_1)+a(5a-4)\,U_{n-1}(\sigma_1)+2(2a-1) }}     
2216: \nonumber \\
2217: &&v_{2n}  = \, \,  
2218: {\frac{-2(2a-1)(5a-4)\,T_{n}(\sigma_1)-3a(3a-2)(a-2)\,U_{n-1}(\sigma_1)}
2219: {4(2a-1)^2\,T_{n}(\sigma_1)}}     \nonumber \\
2220: &&v_{2n-1}  =  \, \, 
2221: {\frac{2(2a-1)(a^2+2a-2)\,T_{n}(\sigma_1)-a(3a-2)(a-2)^2\,U_{n-1}(\sigma_1)}
2222: {-2a\,(2a-1)^2\,T_{n}(\sigma_1)+a(a-2)(3a-2)(2a-1)\,U_{n-1}(\sigma_1)}}     
2223: \nonumber
2224: \end{eqnarray}
2225: where $T_n, U_n$ are Chebyshev polynomials of order $n$ of, 
2226: respectively, first and second kind. 
2227: 
2228: We have very similar results for  the iterates $K^n(V_3)$. The iterates 
2229: $K^n(V_2)$  are quite simple and read:
2230: \begin{eqnarray}
2231: K^n(V_2)\, = \, \, \Bigl(1, \,\, {\frac{2(2a-1)U_{n-1}(\sigma_2)}
2232: {2T_n(\sigma_2)-(5a-4)U_{n-1}(\sigma_2) }} \Bigr), \quad \, 
2233: \sigma_2  =  (3a-4)/2 \nonumber
2234: \end{eqnarray}
2235: 
2236: For $(a, \, b)$ parameters such that $C_0(a,b)=0$, the iterates of the vanishing
2237: conditions of the Jacobian are also given in closed forms and the growth
2238: of the degrees of the parameters is {\em polynomial} ($\delta=1$).
2239: 
2240: Note that one finds similar results along the line $b=-a$
2241: (and similarly on its equivalents obtained 
2242: by the actions of $P$ and $T$) the growth of the degrees of the parameter $a$ in the iterates
2243: of the vanishing conditions  of the Jacobian is also polynomial ($\delta=1$)
2244: for $K^n(V_2)$. However, it is non-polynomial for $K^n(V_1)$ and $K^n(V_3)$
2245: ($1 < \delta \le 2$). The iterates $K^n(V_1)$ 
2246: and $K^n(V_3)$ are not given as closed expressions.
2247: Those $K^n(V_2)$  are given by:
2248: \begin{eqnarray}
2249: K^n(V_2) =\,   \Bigl(-1, \,\, {\frac{2U_{n-1}(\sigma)}
2250: {2T_n(\sigma)+aU_{n-1}(\sigma) }}  \Bigr), \quad \quad 
2251: \quad \sigma =  a/2 \nonumber 
2252: \end{eqnarray}
2253: $K^n(V_2) \in I_2$ gives the points 
2254: where the curves $C_N^{ij}$ are tangent to the line $b=-a$.
2255: 
2256: 
2257: \section{Appendix F: Cases of integrability}
2258: 
2259: The points $(a,b)$ for which the mapping $K$ defined in (\ref{defKanti})
2260: is integrable are shown in
2261: Figure \ref{f:fig1} (lower left corner). These points are lying on the lines (solid lines) $b=a$,
2262: $b=2-2a$ and $b=1-a/2$, and on the curve $C_0(a,b)=0$ (ellipse). The dashed 
2263: lines in Figure \ref{f:fig1} (lower left corner) are $b=-a$, $b=2$ and $a=2$.
2264: 
2265: 
2266: On the lines $b=a$ and $b=2-2a$, the integrable cases are:
2267: \begin{eqnarray}
2268: \label{intaeqb}
2269:  a\,=\,\,0, \,\,\,\,\, {{1}\over{3}}, \,\,\,\,\,
2270:  1-{{1}\over{\sqrt{3}}},  \,\,\,\,\,
2271: {{2}\over{3}}, \,\,\, \,\,1, \,\,\,\,\, {{4}\over{3}}, 
2272:  \,\,\,\,\, 1+{{1}\over{\sqrt{3}}}
2273: \end{eqnarray}
2274: On line $b=1-a/2$, the integrable cases, obtained by applying $T \cdot P$,
2275: are given by $(2-2a, a)$ from (\ref{intaeqb}).
2276: The point $(a=2/3, b=2/3)$ is common to three lines and
2277: corresponds to a matrix of the stochastic form (\ref{C2})
2278: and the ``antistochastic'' form (transpose)  in the same time. 
2279: 
2280: From these 19 points $(a,b)$, the following six are also on the curve $C_0(a,b)=0$
2281: \begin{eqnarray}
2282: \label{C0intab}
2283:  (-2/3, 4/3), \quad (4/3, -2/3), \quad (4/3, 4/3), \quad (0,0), \quad 
2284: (0,2), \quad (2,0)  \nonumber 
2285: \end{eqnarray}
2286: The curve $C_0(a,b)=0$ has six other integrable cases:
2287: \begin{eqnarray}
2288: \label{C0intC0}
2289: \Bigl({{1 \, \pm \sqrt{5}}\over{2}}, 1 \Bigr), \,\, \quad 
2290: \Bigl({{1\, \pm \sqrt{5}}\over{2}}, {{1 \, \mp \sqrt{5}}\over{2}} 
2291: \Bigr), \,\,\quad 
2292: \Bigl(1, {{1\, \pm \sqrt{5}}\over{2}} \Bigr), \,\,
2293: \end{eqnarray}
2294: One has a total of 25 values of $(a,b)$ for which the mapping $\, K$ is 
2295: integrable.
2296: 
2297: The integrable points common to $C_0(a,b)=0$ and the lines $b=a$,
2298: $b=2-2a$, and $b=1-a/2$,
2299: can be understood from the
2300: existence of the two preserved two-forms.
2301: Let us consider, for instance, the point
2302:  $\, (a, \, b) \, = \, (0, \, 2)$ intersection of $\, C_0(a, b) \, = \, 0$
2303:  and $\, b \, = 2-2\, a$.
2304: Transformation $\, K$ for $\, (a, \, b) \, = \, (0, \, 2)$ 
2305: preserves
2306:  two two-forms respectively associated with $\, b\, = \, 2 -\, 2\, a$ in (\ref{list}),
2307: and $\, C_0(a, b) \, = \, 0$ (see  (\ref{list2})), namely:
2308: \begin{eqnarray}
2309: \label{bothtwoform}
2310: {{ du \cdot dv} \over { (1-v) \cdot ((v+{u}^{2})\, +2\,u (1+v)) }}, \quad \quad 
2311: {{ du \cdot dv} \over { (1+v) \cdot ((v+{u}^{2})\, +2\,u (1+v)) }} \nonumber
2312: \end{eqnarray}
2313: corresponding to the fact that  $\, K$ has (up to a sign)
2314:  $\, Inv \, = \, \, (1+v)/(1-v)$,  as an invariant. This is indeed the case since:
2315: \begin{eqnarray}
2316: \label{K2homo}
2317: K^2(u, \, v) \, = \, \, \, \Bigl( -{{ (4 \, +7\,v\, +4\,{v}^{2}) \cdot u\,
2318:  +2\,v \left( 1+v \right) }
2319:  \over { 2\, (1+v) \cdot u\, + \, v }}
2320: , \, \,v \Bigr)
2321: \end{eqnarray}
2322: We have similar results for the two other integrable points
2323:  $\, (a, \, b) \, = \, (0, \, 0)$ and  $\, (a, \, b) \, = \, (2, \, 0)$. They 
2324: also correspond to $\, K^2$ being a homographic
2325:  transformation ($(a, \, b) \, = \, (0, \, 0)$
2326: preserves the $\, u$ coordinate, and $\, (a, \, b) \, = \, (2, \, 0)$
2327:  preserves the ratio $\, u/v$).
2328:  Note that for the point $\, (a, \, b) \, = \, \, (1, \, 1)$, as well
2329:  as $\, (1, \, 0)$ and $(0, \, 1)$,
2330: the mapping  $\, K$ is of order six,
2331: $\, K^6 \, = \, identity$.
2332: 
2333: The mapping $\, K$, 
2334: for the integrable point  $\, (a, \, b) \, = \, \, (4/3, \, -2/3)$
2335:   preserves two  two-forms:
2336: \begin{eqnarray}
2337: {{du \cdot dv} \over 
2338: {\left( v-1 \right)  \left( 4\,u(1+v)+5\,(v+u^2) \right) }}, 
2339: \qquad  \qquad 
2340:   {{du \cdot dv} \over {
2341: \left( v-1 \right)  \left( v-  {u}^{2} \right) }} \nonumber
2342: \end{eqnarray}
2343: their ratio giving the algebraic $\, K$-invariant (up to sign):
2344: \begin{eqnarray}
2345: Inv \, = \, \, 
2346: {\frac {  v-{u}^{2} }
2347: { 4\,u(1+v)+5\,(v+u^2) }}
2348: \end{eqnarray}
2349: 
2350: 
2351: \section{Appendix G: miscellaneous exact results for $\, \xi \, = \, a+b+c-1\, \ne \, 1$}
2352: \label{appendX}
2353: 
2354: Let us provide here a set of exact results, structures (existence of meromorphic two-forms ...) 
2355:  valid in the more general framework where 
2356:  $ c \, \ne \, 2-a-b$ ($\, K^N$ and  $\, K^{-N}$
2357: are no longer conjugate).
2358: 
2359: When  $ c \, \ne  \, 2-a-b$,
2360: the resultant in $\, u$ of the two conditions of
2361:  order two of birational transformation (\ref{defKanti}),
2362: namely $\, K^2(u, \, v) \, = \, \, (u, \, v)$,
2363: yields the following condition (reducing to condition
2364:  $\, C_0(a, \, b)\,=\,0$ previously 
2365: written, when  $ c \, =  \, 2-a-b$):
2366: \begin{eqnarray}
2367: \label{newcond}
2368: a\,b\,+b\,c\,+c\,a  \, =  \,\, \, \,0
2369: \end{eqnarray}
2370: associated with the (quite symmetric) homogeneous $\, K$-covariant ($K^2$-invariant) 
2371: in the $(x, \, y, \, t)$ homogeneous variables:
2372: \begin{eqnarray}
2373: \label{covorder2}
2374: &&cov(x, \, y, \, t) \, = \, \, 
2375: \,bc \cdot t \left( {y}^{2}-{x}^{2} \right) +\,ac \cdot x \left( {t}^{2}-{y}^{2}
2376:  \right) +\,a b\cdot y \left( {x}^{2}-{t}^{2} \right)
2377: \nonumber \\
2378: &&\quad +\, \left( yt+xt+xy \right)  \left(  \left( c-b \right) bc\cdot t\, + \left( b-
2379: a \right) a b \cdot y \, + \, ac \cdot\left( a-c \right)\cdot x \right)\nonumber 
2380: \end{eqnarray}
2381: 
2382: 
2383: One easily finds that, restricted to (\ref{newcond}), the following
2384: meromorphic two-form is preserved up to a minus sign:
2385: \begin{eqnarray}
2386:  {{ dx' \cdot dy' } \over { cov(x', \, y', \,1)}} \,\, \,  =  \,\,\, \,  (-1) \times  
2387: {{ dx \cdot dy } \over { cov(x, \, y, _,1)}}
2388: \end{eqnarray}
2389: 
2390: 
2391: \subsection{For $\, b \, = \, c$, 
2392: when $ c \, \ne  \, 2-a-b$: more two-forms. }
2393: 
2394: Keeping in mind the simple results (\ref{list})
2395: for meromorphic two-forms (\ref{rho12}), let us restrict to the case
2396: where the $\, K$-covariant $\, \rho(u, \, v)$
2397: in a meromorphic two-form like (\ref{rho12}), is a {\em polynomial}, instead 
2398: of a rational (algebraic, ...)
2399: expression. 
2400: Let us remark that when $\, c\, = \, b$ 
2401: but $ c \, \ne  \, 2-a-b$, $\, u-v$ is a covariant of transformation $\, K$
2402: with cofactor $\, 1/((a-1) \, u\, v\,  +a (u+v))$. Recalling expression (\ref{jact}) 
2403: of the Jacobian of (\ref{defKanti}), it becomes quite natural, when $\, b=c$, 
2404:  to make an ``ansatz'' 
2405: seeking for covariant polynomials $\, \rho(u, \, v)$ of the form $\,\rho(u, \, v)\, 
2406: = \, (u-v) \cdot Q(u, \, v)$, where  $\, Q(u, \, v)$ 
2407: will be a  $\, K$-covariant quadratic polynomial
2408: with cofactor  $\, \xi \cdot u\, v/((a-1) \, u\, v\,  +a (u+v))^2$. After
2409:  some calculations, one finds
2410: that the quadratic polynomial $Q(u, \, v)$ must be of the form:
2411: \begin{eqnarray}
2412: Q(u, \, v) \, = \,\,\,\,
2413: A \, {a}^{2}\,uv\,\,  + \,B \,{a}^{2}\cdot  (u+v)
2414:  -a \, (2\,b-1)\, B\, -b
2415: \, (b-1)\,  A \nonumber
2416: \end{eqnarray}
2417: the $(a, \, b)$ parameters being necessarily such that :
2418: \begin{eqnarray}
2419: &&\left( b-a \right)  \left( a+b+c-2 \right) 
2420:  \left( ab+bc+ac \right)
2421: \, \, = \, \, 0 \qquad \hbox{and :} \nonumber \\
2422: &&{a} \left( b+c-1 \right)  \left( b+a \right)  
2423: \left( {b}^{2}+ab+{a}^{2}-a-b-c+1 \right) \nonumber 
2424: \, \, = \, \, 0 
2425: \end{eqnarray}
2426: $\, \bullet \, $ Conditions $\, b \, = \, c \, = \, -a\, $
2427:  yields $A \, = \, B$, and the conformally preserved two-form
2428: reads ($\xi=a+b+c-1$):
2429: \begin{eqnarray}
2430: &&{{ du' \cdot dv'} \over { (u'-v') \cdot (u'+1)  \cdot (v'+1) }}
2431: \,   = \, \,\,\, \, \xi \cdot 
2432: {{ du \cdot dv} \over { (u-v) \cdot (u+1)  \cdot (v+1) }}
2433:  \nonumber 
2434: \end{eqnarray}
2435: $\, \bullet \, $ Conditions $\, b \, = \, c \, = \, a\, $
2436:  yields the conformally preserved two-form:
2437: \begin{eqnarray}
2438: \label{aegbegc}
2439: {{ du' \cdot dv'} \over {(u'-v') \cdot (u'-1) \cdot (v'-1)}}\,\, = \, \, \,\, \,
2440: \xi \cdot 
2441: {{ du \cdot dv} \over {(u-v) \cdot (u-1) \cdot (v-1)}} \nonumber
2442: \end{eqnarray}
2443: 
2444: 
2445: \subsection{For  $ c \, \ne  \, 2-a-b$: more complexity reductions }
2446: \label{miscell}
2447: Condition $K^2(V_2) \in \, I_2$ amounts to writing
2448: \begin{eqnarray}
2449:   K^2\Bigl(  u, \, \,0 \Bigr)  \,\,
2450:  = \,\,K \Bigl( {{b} \over {a}} , \, \,{{ c-1} \over {a}} \Bigr) 
2451:  \, = \, \, \Bigl( {{a} \over {b}} , \, \,{{a} \over {1-a-b}} \Bigr)
2452:  \nonumber
2453: \end{eqnarray}
2454: which yields several algebraic curves,  in particular the rational curve 
2455:  $\, (c, \, b) \, =$ 
2456: $\, \, (1+1/2\,{a}^{2}, \,-a)$, for which one can verify 
2457: that a reduction of the degree
2458: growth rate complexity $\, \lambda \, \simeq \,1.839 $ takes place. The 
2459: degree generating function reads:
2460: \begin{eqnarray}
2461: G_{b=-a,c=1+1/2\,{a}^{2}} \,\,  = \, \, \,
2462:  {{1} \over { 1-t-t^2-t^3}} \,\,  = \,\,  \,
2463: {{ 1-t} \over { 1-2\, t+t^4}}
2464: \nonumber 
2465: \end{eqnarray}
2466: Similarly   $K^4(V_2) \in \, I_2$ yields several algebraic curves, in particular the rational curve 
2467:  $\, (c, \, b) \, =$ $\, \, (1+a^2/3, \,-a)$, for which one can verify 
2468:  a reduction of the degree
2469: growth rate complexity $\, \lambda \, \simeq \,1.965 $, the degree generating function reading:
2470: \begin{eqnarray}
2471: &&G_{a=-b,c= 1+a^2/3} \, = \, \, \, {{1-t} \over {1\, -2\,t\, +t^6}}  \, = \, \, \, 
2472: 1+t+2\,{t}^{2}+4\,{t}^{3} \nonumber \\
2473: && \quad  \quad \quad \quad +8\,{t}^{4}+16\,{t}^{5}+31\,{t}^{6}+61\,{t}^{
2474: 7}+120\,{t}^{8}+236\,{t}^{9} \, + \cdots \, \, \,  \nonumber
2475: \end{eqnarray}
2476: 
2477: 
2478: \vskip .2cm 
2479: 
2480: This is just a set of results for $\, \xi\, \ne 1$, among many 
2481: others that can be easily established.
2482: 
2483: \vskip 2cm 
2484: 
2485: 
2486: 
2487: 
2488: \begin{thebibliography}{1}
2489: 
2490: 
2491: \bibitem{bimero} J. Diller and C. Favre, \emph{Dynamics of bimeromorphic maps of surfaces}, 
2492: Amer. J. Math., 123 (6) (2001) 1135- 1169
2493: 
2494: \bibitem{Kim} E. Bedford and K. Kim, \emph{On the degree growth of birational mappings in higher dimension.}, 
2495: J. Geom. Anal., {\bf 14} (2004) 567-596 and arXiv: math>DS/0406621 
2496: 
2497: 
2498: \bibitem{Pesin}
2499:   Y. B. Pesin,
2500:   Russ. Math. Survey {\bf 32} (1977) 55
2501: 
2502: 
2503: \bibitem{Green} T-C Dinh and N. Sibony, {\em Green currents for holomorphic
2504:  automorphisms of compact K\"ahler manifolds}, 
2505: J. Amer. Math. Soc. {\bf 18}, (2005), 291-312.
2506: 
2507: 
2508: \bibitem{Anosov} D.V. Anosov, {\em Geodesic Flow on Closed Riemannian Manifolds of Negative Curvature},
2509: Trudy Mat. Inst. Steklov, {\bf 90}, (1970) 1-209.
2510: 
2511: 
2512: \bibitem{Anosov2} S. Smale,  {\em Differentiable Dynamical Systems}, 
2513: Bull. Amer. Math. Soc, {\bf 73}, (1967) 747-817.
2514: 
2515: 
2516: 
2517: \bibitem{Yoccoz} J-C. Yoccoz, {\it Id\'ees g\'eom\'etriques en syst\`emes dynamiques}, in 
2518: 	{\it Chaos et d\'eterminisme }, A. Dahan Dalmenko, J-L. Chabert and K. Chemla editors. 
2519:         Editions du Seuil, (1992), Collection In\'edit Sciences, pp. 18-67
2520: 
2521: 
2522: \bibitem{rob-qui-92}
2523: J.A.G. Roberts, G.R.W. Quispel,
2524: {\it Chaos and time-reversal symmetry. Order and chaos in reversible dynamical
2525: systems}, Phys. Rep. {\bf 216} (1992)63-177
2526: 
2527: 
2528: \bibitem{QuRo88}
2529: 	G.R.W. Quispel and J.A.G.Roberts, 
2530: 	{\em Reversible mappings of the plane}.
2531: 	Phys.Lett {\bf A132} (1988), pp. 161--163.
2532: 
2533: \bibitem{BoMaRo93c}
2534:          S.~Boukraa, J-M. Maillard and  G.~Rollet, 
2535: 	{\em Almost integrable mappings}.
2536:         \newblock Int. J. Mod. Phys. {\bf B8} (1994), pp. 137--174
2537: 
2538: 
2539: \bibitem{ab-an-bo-ma-2000}
2540: 	N. Abarenkova, J.-C. Angl\`es d'Auriac, 
2541: 	S. Boukraa and J.-M. Maillard, 
2542: 	{\it Real topological entropy versus metric entropy for birational
2543: 	 measure-preserving transformations}. 
2544: 	Physica D {\bf 144} (2000) 387-433
2545: 
2546: 
2547: \bibitem{rearea} 
2548: 	N. Abarenkova, J-C. Angl\`es d'Auriac, S. Boukraa, 
2549: 	S. Hassani and  J-M. Maillard,
2550: 	{\em Real Arnold complexity versus real topological
2551: 	  entropy for birational transformations}.
2552:  	J. Phys  {\bf A 33 } (2000)  1465-1501
2553: 	and:  chao-dyn/9906010
2554: 
2555: 
2556: \bibitem{topo} 
2557: 	 N. Abarenkova, J-C. Angl\`es d'Auriac, 
2558: 	S.~Boukraa, S. Hassani and  J-M. Maillard,
2559: 	{\em Topological entropy and complexity for discrete dynamical
2560: 	systems},  Phys. Lett. {\bf A 262} (1999)  44-49
2561: 	and : chao-dyn/9806026 .
2562: 
2563: 
2564: \bibitem{complex} 
2565:          N. Abarenkova,  J-C. Angl\`es d'Auriac,
2566:         S. Boukraa and J-M. Maillard,
2567:         \emph{  Growth-complexity spectrum of some 
2568: discrete dynamical systems},  Physica {\bf D 130}, 27--42, (1999) 
2569: 
2570: 
2571: 
2572: \bibitem{Cantat} S. Cantat,
2573: {\em Dynamique des Automorphismes des surfaces K3}.
2574: Acta Math. {\bf 187}:1 (2001), 1-57
2575: 
2576: \bibitem{CantatFavre} S. Cantat and C. Favre, {\em Sym\'etries birationnelles
2577:  des surfaces feuillet\'ees}, J. Reine Ange. Math. (2003),  $n^{o}\, 561$,
2578: 199-235
2579: 
2580: 
2581: 
2582: 
2583: \bibitem{Jacobconj} Jacobian conjecture:
2584: http://www.math.columbia.edu/~woit/blog/archives/000105.html, 
2585: http://mathworld.wolfram.com/InvertiblePolynomialMap.html
2586: 
2587: 
2588: \bibitem{zeta} 
2589:          N. Abarenkova,  J-C. Angl\`es d'Auriac, S. Boukraa,
2590:          S. Hassani and J-M. Maillard ,
2591:         \emph{   Rational dynamical zeta
2592:  functions for  birational transformations}, Physica {\bf A 264},(1999)  264-293
2593: 
2594: 
2595: \bibitem{firthcoming} 
2596:              N. Abarenkova, et al. In preparation. 
2597: 
2598: \bibitem{Smillie} E. Bedford and J. Smillie,
2599:  {\em Polynomial diffeomorphisms of $\, C^2$: currents, equilibrium
2600:  measure and hyperbolicity
2601: } Inventiones Math. {\bf 103}, (1991) 69-99
2602: 
2603: 
2604: \bibitem{BedDill} E. Bedford and J. Diller, {\em Real and Complex dynamics
2605:  of a Family of birational maps of the plane: the golden mean subshift}, 
2606: arKiv: math.DS/0306276 v1 
2607: 
2608: 
2609: \bibitem{DillBed} E. Bedford and J. Diller, {\em Dynamics of a Two Parameter Family of Plane 
2610: birational Maps: I. Maximal Entropy }, 
2611: in preparation
2612: 
2613: \bibitem{Shilnikov} S.V. Gonchenko, L.P. Shil'nikov and D.V. Turaev, {\em On
2614:  models with non-rough Poincar\'e homoclinic curves}, Physica {\bf D62} (1993) 1-14 
2615: 
2616: \bibitem{Shilnikov2} S.V. Gonchenko, L.P. Shil'nikov and D.V. Turaev, 
2617: {\em Homoclinic tangencies 
2618: of an arbitrary order in Newhouse domains}, J. Math. Sci. (NY) {\bf 105} 1738-1778
2619: 
2620: \bibitem{Shilnikov3}  S.V. Gonchenko, L.P. Shil'nikov and D.V. Turaev, {\em Complexity 
2621: in the bifurcation structure of homoclinic loops to a saddle-focus},
2622:  Nonlinearity {\bf 10}, (1997) 409-423
2623: 
2624: \bibitem{Shilnikov4}  S.V. Gonchenko, L.P. Shil'nikov and D.V. Turaev, 
2625: {\em Dynamical phenomena in systems with
2626: structurally unstable Poincar\'e homoclinic orbits}, Nonlinearity {\bf 10}, (1997) 409-423
2627: 
2628: \bibitem{Noether} S. Boukraa, S. Hassani and J-M. Maillard, {\em Noetherian mappings}, 
2629: Physica {\bf D 185}, (2003) 3-44 
2630: 
2631: 
2632: \bibitem{Cremona} S. Boukraa, S. Hassani and J-M. Maillard, {\em New integrable cases of a Cremona
2633: 	  transformation: a finite order orbit analysis}, Physica {\bf A 240}, (1997), 586
2634: 
2635: 
2636: \bibitem{Griffiths} P. Griffiths and J. Harris , {\em  Principles
2637:  of Algebraic Geometry}, J. Wiley and Sons, 1978. M. Reid.
2638: 
2639: \bibitem{Fourier} C. Favre, {\em Points p\'eriodiques
2640:  d'applications birationnelles}, Annales de l'Institut Fourier. 
2641: (Grenoble) {\bf 48}, (1998), $n^{o}\, 4$,, 999-1023
2642: 
2643: \bibitem{Fulton} W. Fulton, {\em Intersection theory}, Erghebnisse 2, 1984, Springer.
2644: 
2645: \bibitem{classif}
2646: 	J-C. Angl\`es d'Auriac, J-M. Maillard and C. M. Viallet, 
2647: 	{\em A classification of four-state spin edge Potts models},
2648:   J. Phys. A 35 (2002) pp. 9251--9272,
2649: 	and  cond-mat/0209557
2650: 
2651: \bibitem{SIDEV}
2652: 	 J-M. Maillard, 
2653: 	{\em Polynomial growth for birational mappings from four-state spin edge
2654:  models},  Journal of Nonlinear Mathematical Physics, 
2655: 	 {\bf 10}, Supplement 2, (2003) 119-132
2656: 
2657: 
2658: \bibitem{Bene} M. Benedicks and L. Carleson {\em The dynamics of
2659:  the H\'enon map}, Ann. Math. {\bf 133},(1991)  73-169
2660: 
2661: \bibitem{Vadim} V. S. Anishchenko, T. E. Vadivasova, G. I. Strelkova and A. S. Kopeikin,
2662:  Discrete Dynamics in nature and Society,
2663: {\em Chaotic attractors of Two-Dimensional
2664: Invertible Maps},  {\bf 2},(1998) 249-256 
2665: 
2666: 
2667: \bibitem{Sibony} V. Guedj and N. Sibony, {\em Dynamics of polynomial automorphisms
2668: of $\, C^k$}, Ark. Mat. {\bf 40}, (2002), 207-243.
2669: 
2670: 
2671: \bibitem{Sibony2}  N. Sibony, {\em Dynamique des applications rationnelles de $\, P^k$},
2672:   Panoramas et Synth\`eses, {\bf 8}, (1999), 97-185
2673: 
2674: 
2675: 
2676: \end{thebibliography}
2677: 
2678: 
2679: 
2680: \end{document}
2681: 
2682: