1: \input hmacro

2: \input epsf.sty

3: \magnification\magstep1

4: %\headline{\rm\hfil February 12, 2004}

5: 

6: \centerline{\bf Real Polynomial Diffeomorphisms with Maximal Entropy:}

7: \centerline{\bf II. Small Jacobian}

8: \smallskip

9: \centerline{Eric Bedford and John Smillie}

10: \bigskip

11: \section 0.  Introduction

12: 

13: The problem of understanding the dynamical behavior of  diffeomorphisms has played a central

14: role in the field of dynamical systems. One way of  approaching this question is to ask about

15: generic behavior in the space of diffeomorphisms.  Another way to approach it is to ask about

16: behavior in some specific parametrized family.  The family of diffeomorphisms of $\R^2$

17: introduced by H\'enon has often played the  role of such a test case. This is a two parameter

18: family given by the formula 

19: $$f_{a,b}(x,y)=(a-x^2-by,x)$$ for $b\ne0$. There are regions of parameter space which are well 

20: understood. If we fix $b$ then for $a\ll0$ the nonwandering set of $f_{a,b}$ is empty. For 

21: $a\gg0$, it is shown in [DN] that the restriction of $f_{a,b}$ to its nonwandering set is

22: hyperbolic and topologically conjugate to the full two-shift. Such  diffeomorphisms are called

23: ``horseshoes''. How the dynamics changes between these two extremes  has been the subject of

24: much investigation.  The case $b=0$ is an  interesting special case. In this case the map

25: $f_{a,b}$ is not a diffeomorphism; in fact the  dynamical behavior is essentially one

26: dimensional.  The dynamical  complexity of

27: $f_{a,0}$ increases monotonically with $a$ (see [MT]).  For other values of $b$ no  such

28: results are known. In fact [KKY] show that in some respects the behavior should  not be

29: expected to be monotone. One way of measuring the topological complexity is through the

30: topological entropy,

31: $h_{top}(f_{a,b})$. This is a continuous real  valued function of the parameters which takes

32: on values in the interval $[0,\log 2]$.  The case $a\ll0$ corresponds to  $h_{top}=0$. The

33: case $a\gg0$ corresponds to 

34: $h_{top}=\log 2$. In this paper we study the set of parameters $(a,b)$ for which 

35: $h_{top}(f_{a,b})$ takes on its maximal value. We say that $f_{a,b}$ has maximal entropy if 

36: $h_{top}(f_{a,b})=\log 2$. We analyze the ``maximal entropy locus'' when the Jacobian parameter

37: $b$ is small. We show:

38: 

39: \proclaim Theorem 1. For each $b$ with $|b|<.08$ there is a unique 

40: $a=a_b$ so that

41: $h_{top}(f_{ab})<\log 2$ for $a<a_b$ and $h_{top}(f_{ab})=\log 2$ for 

42: $a\ge a_b$.

43: Further, we have:

44: \item{(1)} If $a>a_b$, $f_{ab}$ is a hyperbolic horseshoe.

45: \item{(2)} If $a=a_b$, $f_{ab}$ has a

46: quadratic tangency between stable and unstable manifolds of fixed 

47: points.  This tangency

48: is homoclinic when $b>0$ and heteroclinic when $b<0$.

49: 

50: The next result discusses properties of the function $b\mapsto a_b$ defined in Theorem 1.

51: 

52: \proclaim Theorem 2.  The function $b\mapsto a_b$ is continuous on the interval

53: $(-.08,.08)$.  It is analytic on the subintervals $(-.08,0)$ and $(0,.08)$ but not

54: differentiable at $b=0$.  Furthermore, there is a generic

55: unfolding of the homoclinic tangency at the parameter $(a_b,b)$, i.e., at

56: the point of  tangency, the stable and

57: unstable manifolds move past one another with positive speed with respect to  $a$.

58: 

59: The terminology ``generic unfolding'' will be explained in greater detail in \S5.

60: 

61: Part of Theorem 1 follows from a more general analysis of polynomial  diffeomorphisms of

62: maximal entropy in degree $d\ge 2$, which was carried out in  [BS8] and [BS1]. In  particular

63: we proved in this more general context that a maximal entropy polynomial diffeomorphism  is

64: either hyperbolic or has a quadratic tangency between stable and unstable manifolds of 

65: periodic points. The contribution of this paper is to describe the set of parameter values 

66: corresponding to these two types of behavior.

67: 

68: Though these results are stated for the diffeomorphisms 

69: $f_{a,b}:\R^2\to\R^2$ our methods give us very complete information about the corresponding

70: complex extensions

71: $f_{a,b}:\C^2\to\C^2$ for maximal entropy parameter values. In fact it  is the analysis of

72: these complex extensions which allows us to obtain information about  the real H\'enon

73: diffeomorphisms. In particular we take advantage of theory of  intersections of complex

74: manifolds to analyze the complex extensions of the of the real stable  and unstable manifolds.

75: 

76: In addition to proving Theorems 1 and 2, a goal of this paper is to develop the technique of

77: crossed mappings as a method of more general applicablilty in the analysis of families of

78: polynomial diffeomorphisms of ${\bf C}^2$.  These techniques are explored further in [BS3].

79: 

80: We note that this is not the first time that complex methods have  been used to address

81: similar questions.   J.H. 

82: Hubbard and R.~Oberste-Vorth [O] used complex methods to improve the result of

83: Devaney-Nitecki.  And Forn\ae ss and  Gavosto [FG1,2] have

84: used complex methods to show that there is a generic unfolding of a  complex tangency for

85: $f_{a,b}$ for certain parameters $(a,b)$.

86: 

87: 

88: \section 1. The Quadratic Horseshoe Locus: First Approximation

89: 

90: We consider mappings of the form

91: $$f_{a,b}(x,y) = (a-x^2-by,x)\eqno(1.1)$$

92: with $b\ne0$.  Note that $f$ may also be written in the form

93: $\chi^{-1}\circ f\circ\chi=(y,y^2-a-bx)$, where we set 

94: $\chi(x,y)=(-y,-x)$.  We

95: say that

96: $f_{a,b}$ is a {\it (complex) horseshoe} if

97: $f_{a,b}$ is hyperbolic on $J=J(f_{a,b})$ and if $f|J$ is 

98: topologically conjugate

99: to the full 2-shift.  If, in addition, $J\subset\R^2$, we say that 

100: $f_{a,b}$ is a

101: {\it real horseshoe}.  J.H.\ Hubbard and R.

102: Oberste-Vorth have obtained estimates

103: on the (complex) horseshoe locus; see [O] and [MNTU, 

104: Proposition 7.4.6].  These are summarzied in the following:

105: 

106: \proclaim Theorem 1.1.  If $b\ne0$, and if

107: $|a|>2(1+|b|)^2$, then $f_{a,b}$ is a (complex) horseshoe.  If, in addition,

108: $b\in\R$, and $a>0$, then $f_{a,b}$ is a real horseshoe.

109: 

110: Since horseshoes have entropy equal to $\log 2$, the following result 

111: gives a large

112: region of parameter space where there are no horseshoes.  This is the region to

113: the left in Figure~1.

114: 

115: \proclaim Theorem 1.2.  Define $\sigma^-(b)=2-{13\over 8} b-{7\over 

116: 64} b^2$ and

117: $\sigma^+(b)=2+{7\over 4}b + {5\over 16}b^2$.  If $(a,b)\in\R^2$ 

118: satisfy $b\ne0$,

119: $|b|\le1$ and

120: $a\le\max(\sigma^-(b),\sigma^+(b))$, then the entropy of

121: $f|\R^2$ is less than

122: $\log2$.

123: 

124: \give Proof.  First we note that a fixed point of $f_{a,b}$ has the form

125: $(x,y)$, where

126: $$x=y=-{1\over 2}\left[b+1 \pm\sqrt{(b+1)^2 + 4a}\right]. \eqno(1.2)$$

127: Now we recall some results from [BS1].  If $f_{a,b} $ is a quadratic

128: diffeomorphism of $\R^2$, and if $f|\R^2$ has entropy $\log2$, then $f_{a,b}$

129: has two fixed points, which must be saddles.  Further, one of these points must

130: be unstably one-sided, and the unstable eigenvalue of $Df$ at this fixed point

131: is (strictly) greater than 4.  The other fixed point has a negative eigenvalue

132: in the unstable direction, and this eigenvalue must be less than $-2$.

133: 

134: The differential is given in $(x,y)$-coordinates as

135: $$Df_{a,b}=\pmatrix{-2x & -b\cr 1& 0\cr}.$$

136: The product of the eigenvalues is $b$, so we may write them as $\lambda$ and

137: $b/\lambda$.  Thus the trace of the differential is

138: $$-2x=\lambda+b/\lambda.$$

139: If $|b|\le1$, then $\lambda\mapsto\lambda+b/\lambda$ is strictly increasing in

140: $\lambda$ for $|\lambda|>1$.  Thus the condition that there is an eigenvalue

141: greater than 4 gives us the inequality

142: $$-2x>4+{b\over 4},\eqno(1.3)$$

143: and the condition that there is an eigenvalue less than $-2$ gives the

144: inequality

145: $$-2x<-(2+{b\over2}).\eqno(1.4)$$

146: (Note that the inequalities (1.3) and (1.4) refer to different fixed 

147: points and thus

148: involve different values of $x$.)

149: 

150: Now we substitute expression (1.2) into (1.3) and obtain

151: $$\eqalign{ b+1 \pm \sqrt{(b+1)^2+4a} & > 4+{b\over 4}\cr

152:    \sqrt{(b+1)^2+4a} & > 3 - {3\over 4}b\cr

153: b^2 +2b+1+4a > 9 - {9\over2}b + {9\over16} b^2\cr}$$

154: which gives $a>\sigma^-(b)$.

155: 

156: 

157: Similarly, we substitute (1.2) into (1.4) and find

158: $$\eqalign{ b+1 \pm \sqrt{(b+1)^2+4a} & < -2 - {b\over 2}\cr

159: \pm \sqrt{(b+1)^2+4a} & < -3 -{3\over 2}b\cr

160: b^2 +2b + 1 + 4a & > 9 + 9b + {9\over 4}b^2,\cr}$$

161: where the last inequality is reversed because the quantities being squared are

162: negative.  Thus $a>\sigma^+(b)$.  The case that one of these inequalities fails

163: happens exactly when we have $a\le\max(\sigma^+(b),\sigma^-(b))$, and 

164: in this case

165: the entropy is not equal to $\log2$.\qed

166: 

167: \epsfxsize4.0in

168: \centerline{\epsfbox{fig3_1.eps}}

169: 

170: \centerline{Figure 1}

171: 

172: Figure 1 shows the information on parameter space that is given by Theorems 1.1 and

173: 1.2.  This figure considers only parameters $|b|\le1$.  In fact,

174: we restrict our attention without loss of generality to the case

175: $|b|\le1$ throughout this paper.  Each of the

176: items discussed in the theorem: maximal entropy, the horseshoe property, and generic

177: unfolding, will hold for $f$ if and only if it holds for $f^{-1}$.  Thus the fact that

178: $f_{a,b}^{-1}$ is conjugate to $f_{{a\over b^2}, {1\over  b}}$ means that the regions

179: that define these dynamical behaviors are invariant under the involution

180: $(a,b)\mapsto (ab^{-2},b^{-1})$.  In particular, this gives versions of these Theorems

181: corresponding to the case $|b|>(.08)^{-1}$. 

182: 

183: \section 2.  Complex Boxes and Crossed Mappings

184: 

185: In order to study the system $f|K:K\to K$, we will introduce an open cover by

186: ``boxes'' $B_i$ and study a family of ``crossed mappings'' 

187: $f_{i,j}:B_i\to B_j$.

188: We start by working with $p(z)=2-z^2$ and a covering of its Julia set $[-2,2]$.

189: The Green function for $[-2,2]$ is

190: $$G(z)=\log\left|{z+\sqrt{z^2-4}\over 2}\right|.$$

191: For $\lambda>0$, $p$ induces a 2-fold branched covering

192: $p:\{G<\lambda\}\to\{G<2\lambda\}$. The level sets $\{G=\lambda\}$ are ellipses

193: with foci at $\pm2$, and the gradient lines (the orthogonal 

194: trajectories) are given

195: by the family of hyperbolae with foci at $\pm2$.

196: 

197: 

198: Let us fix $c={1\over2}(\sqrt{17}-1)/2\sim 1.5615 $ and

199: $d={1\over2}(1+\sqrt{7+2\sqrt{17}})\sim2.4523

200: $.  Let

201: $E\subset\C$ be the domain bounded by the ellipse with foci at $\pm2$ 

202: and passing

203: through $\pm d$.  It follows that

204: $p(E)$ is the ellipse with foci at $\pm2$ and passing through 

205: $\pm(d^2-2)$.  We set

206: $D_0:=\{\zeta\in\C^2:\Re(\zeta)<0\}\cap E$ and

207: $D_2:=\{\zeta\in\C^2:\Re(\zeta)>0\}\cap E$ as in Figure 2.  It follows that

208: $p(D_0)=p(D_2)=p(E)-[2,d^2-2)$.  Let

209: $D_1$ denote the region in $E$ lying between the hyperbolae with foci 

210: at $\pm2$ and

211: which pass through $\pm c$ as in Figure 3.  Thus $p(D_1)$ is the region of the

212: ellipse

213: $p(E)$ to the right of the hyperbola with foci at $\pm2$ and passing through

214: $2-c^2$.   We have the following inclusions:

215: $$D_0\cup D_1\subset p(D_0)=p(D_2), \quad D_2\subset p(D_1).$$

216: We may also compute certain distances related to these inclusions:

217: $$\eqalign{ dist(\partial p(D_0),D_0) & = d^2-2-d\cr

218: dist(\partial p(D_0),D_1) & = 2-c\cr

219: dist(\partial p(D_1),D_2) &=c^2-2;\cr}\eqno(2.1)$$

220: While calculating the distances between ellipses can be difficult in general, these

221: calculations are straightforward because the relevant ellipses are in a confocal

222: family.  Thus the minimal distances between these ellipses are realized by points on the

223: real axis.  By the choices of

224: $c$ and $d$, we have

225: $$\Delta:=d^2-d-2=2-c=c^2-2\sim.4384.\eqno(2.2)$$

226: 

227: 

228: \epsfxsize4.0in

229: \centerline{\epsfbox{definitionD0_2.eps}}

230: 

231: \centerline{Figure 2}

232: 

233: \epsfxsize4.0in

234: \centerline{\epsfbox{definitionD1_2.eps}}

235: 

236: \centerline{Figure 3}

237: 

238: Now choose $e>d$ and set $B_j=\{(x,y)\in\C^2:x\in D_j, 

239: |y|<e\}=D_j\times\{|y|<e\}$

240: for $j=0,1,2$.  Thus $ B_0\cup B_1\cup B_2 = E\times\{|y|<e\}$.  We 

241: introduce the set

242: $$\cA:=\{a,b\in\C,\ \ b\ne0, \ \ |a-2|+e|b|<\Delta\}\approx \{|a-2|+2.4|b|<.43\}. 

243: \eqno(\ddag)$$

244: 

245: \proclaim Proposition 2.1.   If $(a,b)\in\cA$, then $K\subset 

246: B_0\cup B_1\cup B_2$.

247: 

248: \give Proof.  By [MNTU, p.\ 238], we know that $K$ is contained in the bidisk

249: $\{|x|,|y|<R\}$, where $R$ is the larger root of the equation 

250: $t^2-(1+|b|)t-|a|=0$.

251: By the condition $|a-2|+e|b|<\Delta$, we conclude that we may take 

252: $e$ sufficiently

253: close to $d$, then we have

254: $$R\le {1+\Delta/e +\sqrt{(1+\Delta/e)^2+4(2+\Delta)}\over2}\sim 2.25845$$

255: Recall that $pE$ is an ellipse with foci at $\pm2$ and major axis of

256: length $d^2-2\sim 4.01378$.  We then compute that its minor axis has length

257: $\sqrt{(d^2-2)^2-4}\sim3.48$.

258: 

259: To prove the Proposition, we need to show that if $(x,y)\in\{|x|,|y|<R\}$, and

260: if $x\notin E$, then $(x,y)\notin K$.   For such $(x,y)$, the $x$-coordinate of

261: $f(x,y)$ satisfies:

262: $$|\pi_v(f(x,y))-p(x)|< |a-2| +|by|<|a-2| +R|b|<\Delta 

263: +(R-e)|b|<1.03465\Delta$$

264: since $|b|<\Delta/e$.  Now let $D:=\{\zeta\in pE:dist(\zeta,\partial

265: (pE))>1.03465\Delta\}$.  Since $x\notin E$, it follows that $px\notin 

266: pE$, and so

267: the

268: $x$-coordinate of $f(x,y)$ does not belong to $D$.  On the other 

269: hand,  the minor

270: axis of

271: $pE$ is 3.48, so that $D$ contains the disk of radius 

272: $3.48-1.03465\Delta\sim3.0264

273: >R$.  Thus $f(x,y)\notin K$.

274: \qed

275: 

276: The vertical and horizontal

277: components of the boundaries are defined to be

278: $$\partial_vB_j=(\partial

279: D_j)\times\{|y|\le e\},\quad\partial_hB_j=\bar D_j\times\{|y|=e\}.$$

280: We set $\cG=\{(0,0),(0,1), (1,2), (2,1), (2,0)\}$, and we interpret $\cG$ as a

281: graph on the vertices $\{B_0,B_1,B_2\}$ as in Figure 4.

282: 

283: \epsfxsize4.0in

284: \centerline{\epsfbox{graph0_1.eps}}

285: 

286: \centerline{Figure 4: Graphs $\cG$ for $f$ (on left) and $\cG^{-1}$ 

287: for $f^{-1}$

288: (on right)}

289: \proclaim Proposition 2.2.   If $(\ddag)$ holds, then 

290: $f_{a,b}(\partial_vB_i)\cap

291: \bar B_j=\emptyset$ and $f_{a,b}(\bar B_i)\cap \partial_hB_j=\emptyset$ for all

292: $(i,j)\in\cG$. 

293: 

294: \give Proof.  By estimates (2.1) and (2.2) and the fact that $p(\partial

295: D_0)=p(\partial D_2)$, we have

296: $$dist(p(\zeta),\partial D_j)\ge\Delta$$

297: for $\zeta\in\partial D_i$ and $(i,j)\in\Gamma$.  Thus if 

298: $x\in\partial D_i$ and

299: $|y|<e$, the first coordinate of

300: $f_{a,b}(x,y)$ satisfies

301: $$|a-x^2-by-p(x)|\le|a-2|+|by|<\Delta.$$

302: This gives $a-x^2-by\notin\bar D_j$, so

303: $f(\partial_vB_i)\cap\bar B_j=\emptyset$.

304: 

305: Note that $\partial_hB_j\subset\{|y|=e\}$.  The second coordinate

306: of $f$ is $x$, so the second condition follows from the fact that $\bar

307: D_j\cap\{|y|=e\}=\emptyset$, independently of $a$ and $b$.

308: \qed

309: 

310: Let $\pi_v(x,y)=x$ and $\pi_h(x,y)=y$ denote the projections in the 

311: vertical and

312: horizontal directions.  We let $f_{i,j}$ denote the mapping

313: $f:B_i\cap f^{-1}B_j\to B_j$.  Following [HO], we say that $f_{i,j}$ is a

314: {\it crossed mapping} if for each $y\in\{|y|<e\}$,

315: $$\pi_v\circ f:(D_i\times\{y\})\cap f^{-1}B_j\to D_j\eqno(2.4)$$

316: is proper.  Given $(i,j)\in\cG$, then it follows from Proposition 2.2 

317: that $f_{ij}$

318: is a crossed mapping.  We say that the degree of

319: $f_{ij}$ as a crossed mapping is the mapping degree of the map in 

320: (2.4) (which is

321: independent of

322: $y$).  Similarly, we say that

323: $f^{-1}:B_j\cap f B_i\to B_i$, which we denote by $f^{-1}_{ji}$, is a crossed

324: mapping if for each

325: $x\in D_j$,

326: $$\pi_h\circ f^{-1}:(\{x\}\times\{|y|<e\})\cap fB_i\to \{|y|<e\}\eqno(2.5)$$

327: is proper.  As was observed in [HO], $f_{ij}$ is a crossed mapping 

328: if and only if

329: $f_{ji}^{-1}$ is.  And

330: the degree of

331: $f^{-1}_{ji}$ as a crossed mapping is defined as the mapping degree 

332: of the map in

333: (2.5) (which is independent of $x$).  This, in turn, is the same as 

334: the degree of

335: $f_{ij}$.  We will say that $(\cB,\cG)$ is a system of crossed mappings, if $f_{i,j}$

336: induces a crossed mapping from $B_i$ to  $B_j$ for each $(i,j)\in\cG$.   The

337: following Corollary is a consequence of Proposition 2.2.

338: 

339: \proclaim Corollary 2.3.   If $(\ddag)$ holds, then $(\cB,\cG)$ is a system of crossed mappings.

340: 

341: We define an {\it orbit} in a system of crossed mappings as a bi-infinite sequence

342: $(p_j,i_j)_{j\in\Z}$ such that for all $j\in\Z$, $p_j\in B_{i_j}$, $(i_j,i_{j+1})\in\cG$, and

343: $f(p_j)=p_{j+1}$.  Next we give conditions that guarantee that every $f$-orbit

344: $(p_j)_{j\in\Z}$ in $K$ can be lifted to an orbit of the system of crossed mappings.

345: 

346: 

347: \proclaim Proposition 2.4.  Suppose that $K\cap (B_0\cup B_1)\subset f(B_0\cup

348: B_2)$ and $K\cap B_2\subset f(B_1)$.  Then for $q\in K$ there is an admissible

349: sequence $I=(i_n)_{n\in\Z}$ such that $f^nq\in B_{i_n}$ for all $n\in\Z$.

350: 

351: \give Proof.  Let us start by making a sequence $J_M=\{j_n: -M\le n\le M\}$ of

352: finite length.  If we have determined $j_n$ already, then $f^n(q)\in 

353: B_{j_n}\cap

354: K$.  If

355: $j_n=0$ or 1, then by hypothesis $f^{n-1}q\in (B_0\cup B_2)\cap K$. 

356: Thus we may

357: choose

358: $j_{n-1}\in\{0,2\}$ such that $f^{n-1}q\in B_{j_{n-1}}$, and in 

359: either case we have

360: $(j_{n-1},j_n)\in\cG$.  Similarly, if $j_n=2$, then $f^nq\in B_2\cap K$, and by

361: hypothesis we have $f^{n-1}q\in B_1$.  Thus we set $j_{n-1}=1$ and

362: $(j_{n-1},j_n)=(1,2)\in\cG$.  Starting at $n=M$, we continue 

363: backwards and generate

364: an admissible sequence $J_M$.

365: 

366: Now we have admissible sequences $J_1,J_2,\dots$ of increasing 

367: length.  For each

368: $M$, there is a sequence $I_M$ that is a subsequence of infinitely 

369: many sequences

370: $J_{k_m}$.  We may make $M$ increasingly large and thus obtain an 

371: infinite sequence

372: $I$.\qed

373: 

374: \proclaim Proposition 2.5.  If $a,b\in\R$ and if (\ddag) holds, then

375: $$(\bar B_{0,r}\cup\bar B_{1,r})\cap f(B_0\cup B_1\cup B_2)\subset 

376: f(B_0\cup B_2)$$

377: $$\bar B_{2,r}\cap  f(B_0\cup B_1\cup B_2)\subset f(B_1).$$

378: 

379: \give Proof.  We note that $(B_0\cup B_1\cup B_2 -B_0\cup B_2)\cap\R^2 =

380: \{0\}\times (-e,e)$.  Thus to prove the first inclusion, it suffices 

381: to show that

382: $(\bar B_{0,r}\cup\bar B_{1,r})\cap f(\{0\}\times(-e,e))=\emptyset$.  But $\bar

383: B_{0,r}\cup\bar B_{1,r}=[-d,c]\times[-e,e]$, and the $x$-projection of the

384: $f$-image of this set is

385: $$\pi_v\circ f(\{0\}\times(-e,e))=\{a-x^2-by:x=0,|y|<e\}\subset

386: (2-\Delta,2+\Delta).$$

387: On the other hand, $\bar B_{0,r}\cup\bar B_{1,r}=[-d,c]\times[-e,e]$. 

388: Thus $(\bar

389: B_{0,r}\cup\bar B_{1,r})\cap f(\{0\}\times(-e,e))=\emptyset$ since 

390: $c+\Delta=2$,

391: which proves the first inclusion.

392: 

393: For the second inclusion, we note that

394: $$(B_0\cup B_1\cup B_2-B_1)\cap\R^2=((-d,-c)\cup(c,d))\times(-e,e).$$

395: The $x$-projection of the $f$-image of this set is

396: $$\{a-x^2-by:c<|x|<d,|y|<e\}\subset(2-d^2-\Delta,2-c^2+\Delta).$$

397: On the other hand, the $x$-projection of $\bar B_{2,r}$ is $[c,d]$, which is

398: disjoint from $(-\infty,2-c^2+\Delta)$ since $2-c^2+\Delta=c$.\qed

399: 

400: 

401: 

402: Let $V\subset B_i$ be a complex subvariety.  We say that $V$ is a 

403: {\it horizontal

404: multi-disk} (resp.\ {\it vertical multi-disk}) if each component of $V$ is

405: conformally equivalent to a complex disk, and if 

406: $\partial_hB_i\cap\bar V=\emptyset$

407: (resp.\ $\partial_vB_i\cap\bar V=\emptyset$).

408: {With this terminology the union of horizontal (resp.\ vertical) multi-disks is

409: again a horizontal (resp.\ vertical) multi-disk.}   We denote the set 

410: of horizontal

411: (resp.\ vertical) multi-disks by

412: $\cD_h(B_i)$ (resp.\

413: $\cD_v(B_i)$).  If $V\in\cD_h(B_0)$ (resp.\ $V\in\cD_v(B_i)$), then 

414: $\pi_v:V\to D_i$

415: (resp.\

416: $\pi_h:V\to\{|y|<e\}$) is proper, and we let

417: $\delta(V)$ denote the degree of the corresponding projection.  By 

418: $\cD_h^{m}(B_i)$

419: (resp.\

420: $\cD_v^{m}(B_i)$) we denote the set of horizontal (resp.\ vertical) 

421: multi-disks $V$

422: such that for each component $W$ of $V$, the degree $\delta(W)$ is no

423: greater than $m$.  We note the following:

424: $${\rm If \ }V'\in\cD_{v}(B_i){\rm\  and\ } V''\in\cD_h(B_i), {\rm\ then\ }

425: \#(V'\cap V'')=\delta(V')\delta(V'').\eqno(2.6)$$

426: If $f_{ij}$ is a crossed map, and if $V\subset B_i$ is a subvariety for which

427: $\partial_hB_i\cap\bar V=\emptyset$, then $\tilde f_{ij}(V):=f(V)\cap 

428: B_j$ is closed

429: in

430: $B_j$ and satisfies $\partial_hB_j\cap\bar{\tilde V}=\emptyset$, and is thus a

431: horizontal subvariety.  If $\deg (f_{ij})$ denotes the degree as a 

432: crossed map, then

433: we have

434: $$\deg(f_{ij})\delta(V)=\delta(\tilde f_{ij}(V)).\eqno(2.7)$$

435: 

436: \proclaim Proposition 2.6.  If $(\ddag)$ holds, it follows that

437: $$\tilde

438: f_{12}:\cD_h^m(B_i)\to\cD_h^{2m}(B_j){\rm\ and\ }\tilde

439: f^{-1}_{21}:\cD_v^m(B_j)\to\cD_v^{2m}(B_i),$$

440: and if $(i,j)\in\cG$, $(i,j)\ne(1,2)$, then

441: $$\tilde

442: f_{ij}:\cD_h^m(B_i)\to\cD_h^m(B_j){\rm\ and\ }\tilde

443: f^{-1}_{ji}:\cD_v^m(B_j)\to\cD_v^m(B_i).$$

444: 

445: \give Proof.  We will show that $\tilde f_{ij}(V)$ is conformally 

446: equivalent to a

447: disk; the degree is given by (2.7).  Suppose

448: $V$ is a horizontal disk in

449: $B_i$.  Then, taking boundary inside $\C^2$, we have $\partial

450: V\subset\partial_v(B_i)$.  By Proposition 2.2,

451: $f(\partial V)\cap \partial B_j=\emptyset$.  Thus each component of 

452: $f(V)\cap B_j$

453: is closed in $B_j$.  The second part of Proposition 2.2 implies that 

454: $\pi_v:f(V)\cap

455: B_j\to D_j$ is proper.  Finally, we need to show that each component $W$ of

456: $f(V)\cap B_j$ is conformally equivalent to the disk.  Since $V$ is a 

457: disk, there

458: is a conformal equivalence $\varphi:\Delta\to V\subset\C^2$.  Now the 

459: components of

460: $fV\cap B_j$ correspond to the components of

461: $\{\zeta\in\Delta:f\circ\varphi(\zeta)\in B_j\}=\{\zeta\in\Delta:\pi_v\circ

462: f\circ\varphi(\zeta)\in D_j\}$.  Since $\bar D_j$ is simply 

463: connected, there is a

464: subharmonic function $s$ on $\C$ such that $\bar D_j=\{s\le0\}$.  It 

465: follows from

466: the maximum principle that each component of $(\pi_v\circ 

467: f\circ\varphi)^{-1}(\bar

468: D_j)=\{s\circ f\circ\varphi\le0\}$ is simply connected.  Finally, 

469: since $\pi_v\circ

470: f\circ\varphi$ is an open mapping, each component of $(\pi_v\circ

471: f\circ\varphi)^{-1}D_j$ is simply connected.

472: \qed

473: 

474: Since $f_{00}$ is a crossed mapping from $B_0$ to itself, there is a 

475: saddle fixed

476: point $p_0\in B_0$.  Let us define $W^{s/u}_0$ to be the connected component of

477: $W^{s/u}(p_0)\cap B_0$ which contains $p_0$.  It follows that

478: $W^{s/u}_0\in\cD^1_{v/h}(B_0)$.  Note that

479: $$W^u_0=\bigcap_{n\ge0}f^nB_0,\quad W^s_0=\bigcap_{n\ge0}f^{-n}B_0.\eqno(2.8)$$

480: There is also a saddle point $p_1\in B_1\cap B_2$.  We let $W^u_1$ denote the

481: component of of $W^u(p_1)\cap B_1$ that contains $p_1$.  We show in Proposition 4.3 that if

482: (\ddag) holds, then it is a horizontal disk of degree 1.

483: 

484: Let us say that a sequence $I=i_0i_1\cdots i_n$ is {\it admissible} if

485: $(i_k,i_{k+1})\in\cG$ for all $k$.  We will sometimes also say that a sequence

486: $J=j_0j_1\cdots j_m$ is admissible if $(j_k,j_{k+1})\in\cG^{-1}$ for 

487: all $k$.  It

488: will be clear from context whether we mean $\cG$ or $\cG^{-1}$.  For admissible

489: sequences $I$ (for $\cG$) and $J$ (for $\cG^{-1}$), we use the notation

490: $$W^u_I=W^u_{i_0i_1\cdots i_n}=\tilde f_{i_{n-1}i_n}\tilde 

491: f_{i_{n-2}i_{n-1}}\cdots

492: \tilde f_{i_0i_1}(W^u_{i_0})\eqno(2.9)$$

493: $$W^s_J=W^s_{j_0j_1\cdots j_n}=\tilde f^{-1}_{j_{n-1}j_n}\tilde

494: f^{-1}_{j_{n-2}j_{n-1}}\cdots

495: \tilde f^{-1}_{j_0j_1}(W^s_{j_0}).\eqno(2.10)$$

496: It follows that $W^{s/u}_0$ are vertical/horizontal disks of degree 1 in $B_0$, and $W^s_{02}$

497: is a vertical disk of degree 1 in $B_2$.  By Proposition 2.6,

498: $W^u_{01}$ are vertical/horizontal disks of degree 1; and $W^u_{012}$ is a

499: horizontal disk of degree 2.  This last statement includes two possibilities:

500: $W^u_{012}$ might consist of two disjoint disks of degree 1 or one disk on which

501: $\pi_v$ has degree 2.  In either case, $W^u_{012}$ intersects $W^s_{02}$ in

502: $B_2$ with multiplicity two, which means that either $W^u_{012}\cap W^s_{02}$

503: consists of two distinct points, or the intersection is tangential.

504: 

505: 

506: 

507: \section 3.  Mappings of Real Boxes

508: 

509: Here we work under the additional condition that $f_{a,b}$ is a real 

510: mapping.  In this section, we will restrict our attention to the real parameter region

511: $$\cA_r:=\cA\cap\R^2.$$

512: Let

513: $\tau$ be the involution of $\C^2$ defined by

514: $\tau(x,y)=(\bar x,\bar y)$.  The fixed point set of $\tau$ is $\R^2$.  The

515: condition that $a,b\in\R$ is equivalent to the condition that 

516: $f_{a,b}$ commutes

517: with $\tau$.  We say that a set $S\subset\C^2$ is {\it real} if $\tau 

518: S=S$.  For instance $\tau

519: B_i=B_i$, so in this terminology $B_i$ is real.  Let

520: $\cD_{h/v,r}(B_i)$ denote the set of horizontal/vertical disks in 

521: $\cD_{h/v}(B_i)$

522: which are real.  If $(a,b)\in\cA_r$, then Proposition 2.6 applies 

523: to real disks

524: to yield

525: $$\tilde

526: f_{12}:\cD_{h,r}^m(B_i)\to\cD_{h,r}^{2m}(B_j){\rm\ and\ }\tilde

527: f^{-1}_{21}:\cD_{v,r}^m(B_j)\to\cD_{v,r}^{2m}(B_i),$$

528: and

529: $$\tilde

530: f_{ij}:\cD_{h,r}^m(B_i)\to\cD_{h,r}^m(B_j){\rm\ and\ }\tilde

531: f^{-1}_{ji}:\cD_{v,r}^m(B_j)\to\cD_{v,r}^m(B_i)$$

532: for $(i,j)\in\cG$, $(i,j)\ne(1,2)$.

533: 

534: We set $B_i^r:=B_i\cap\R^2$, which

535: is a rectangle in $\R^2$ with sides parallel to the axes.

536: \proclaim Proposition 3.1.  If $V\in\cD_{h,r}(B_i)$, then

537: $V\cap B_{i,r}$ consists of a nonempty, connected, one-dimensional 

538: curve.  In fact,

539: there is a conformal uniformization $h:\Delta\to V$ such that

540: $h(\bar\zeta)=\tau\circ h(\zeta)$.

541: 

542: \give Proof.  Let $\varphi:\Delta\to V$ be a conformal uniformization 

543: of $V$.  It

544: follows that

545: $\kappa:\Delta\ni\zeta\mapsto\varphi^{-1}\circ\tau\circ\varphi(\zeta)\in\Delta$ 

546: is

547: an anti-conformal involution of $\Delta$. It follows that $\kappa$ is an

548: orientation-reversing isometry for the Poincar\'e metric, so the 

549: fixed point set

550: $\gamma:=\{\zeta\in\Delta:\kappa(\zeta)=\zeta\}$ is a Poincar\'e geodesic.  Let

551: $\psi$ be a conformal automorphism of $\Delta$ which maps the real axis

552: $(-1,1)\subset\Delta$ to $\gamma$.  It follows that 

553: $\psi^{-1}\circ\kappa\circ\psi$

554: is an isometric involution of $\Delta$ which fixes $(-1,1)$, so it is 

555: simply the

556: map $\zeta\mapsto\bar\zeta$.  Thus

557: $h=\varphi\circ\psi$ is the desired uniformization.

558: \qed

559: If $f$ is a real map,

560: then for $(i,j)\in\cG$, $f_{ij}$ is a crossed mapping of the pair

561: $(B_i^r,B_j^r)$.

562: 

563: \proclaim Proposition 3.2.  If $a,b\in\cA_r$, then $B_{0,r}\cap

564: fB_0$ lies below $B_{0,r}\cap fB_2$ inside $B_{0,r}$, and 

565: $B_{1,r}\cap fB_0$ lies

566: below

567: $B_{1,r}\cap fB_2$ inside $B_{1,r}$.  In particular, let 

568: $I=0i_1\cdots i_n00$ and

569: $J=0j_1\cdots j_m20$ be admissible sequences.  Then $W^u_I$ lies below $W^u_J$

570: inside $B_{0,r}$.  Similarly, if $K=0k_1\cdots k_n01$ and 

571: $L=0l_1\cdots l_m21$ are

572: admissible sequences, then $W^u_K$ lies below $W^u_L$ inside $B_{1,r}$.

573: 

574: \give Proof.  The $y$-coordinate of $f$ is $\pi_h\circ f=x$.  Since 

575: $B_0$ lies to

576: the left of $B_2$, it follows that the $y$-coordinate of $fB_{0,r}$ 

577: is less than

578: that of $fB_{2,r}$.  Thus it lies below.

579: 

580: For the assertions about the pieces of unstable manifolds, we note that if

581: $I$ is a sequence that ends in $ij$, then $W^u_I\subset fB_i\cap 

582: B_j$.  Thus for a

583: sequence $I$ which ends in $00$ and a sequence $J$ which ends in 

584: $20$, we will have

585: $W^u_I\subset B_{0,r}\cap fB_0$ which lies below $W^u_J\subset 

586: B_{0,r}\cap fB_2$.

587: \qed

588: 

589: If $(i,j)\in\cG$, $(i,j)\ne(1,2)$, then the crossed mapping $f_{ij}$ 

590: has degree 1.

591: This means that real, horizontal curves in $B_{i,r}$ which run from 

592: left to right

593: are taken to real, horizontal curves in $B_{j,r}$ which run either from left to

594: right or from right to left.  If the left-to-right direction is preserved, we

595: assign the symbol $\epsilon_u=+$ to $f$.  Otherwise, we set $\epsilon_u=-$.

596: Similarly, real, vertical curves in $B_{j,r}$ which run from bottom to top are

597: mapped under

598: $f^{-1}$ to real, vertical curves which either run from bottom to top 

599: or from top to

600: bottom.  If the run from bottom to top, then we assign the symbol 

601: $\epsilon_s=+$ to

602: $f$.  Otherwise, $\epsilon_s=-$.

603: 

604: \epsfxsize4.5in

605: \centerline{\epsfbox{graph1_1.eps}}

606: \centerline{Figure 5:  Graph induced by $f$ (orientation-preserving)}

607: 

608: \epsfxsize4.5in

609: \centerline{\epsfbox{graph2_1.eps}}

610: \centerline{Figure 6:  Graph induced by $f$ (orientation-reversing)}

611: 

612: \proclaim Proposition 3.3.  If $a,b\in\cA_r$, then the signs

613: $(\epsilon_s,\epsilon_u)$ are given as in Figures 5 and 6.

614: 

615: \give Proof.  First we consider the degenerate case $b=0$.  The map

616: $a-x^2=\pi_v\circ f_{a,0}(x,y)$ is increasing on $D_0\cap\R=(-d,0)$ 

617: and decreasing

618: on

619: $D_2\cap\R=(0,d)$.  Thus we have $\epsilon_u=+$ on $D_0\cap\R$ and 

620: $\epsilon_u=-$

621: on $D_0\cap\R$.  This condition continues to hold for $b\ne0$.  Thus we have

622: $\epsilon_u=+$ on $D_0$ and

623: $\epsilon_u=-$ on $D_2$.  This continues to hold for $b\ne0$, so the arrows of

624: $\cG$ emanating from $B_0$ should be labeled

625: $(\cdot,+)$, and the arrows emanating from $B_2$ should be labeled $(\cdot,-)$.

626: In the orientation-preserving case, the only possible labels are $(+,+)$ and

627: $(-,-)$.  In the orientation-reversing case, the only possible labels 

628: are $(+,-)$

629: and $(-,+)$.  Thus we have the labeling shown in the graphs in Figures 5 and 6. \qed

630: 

631: The crossed map $f_{12}$ has degree 2 and is less easy to work with.  The

632: illustrations on the right hand sides of Figures 5 and 6 indicate its

633: combinatorial behavior in the following sense.  The left side of the

634: vertical boundary of $B_{1,r}$ is

635: $\{-c\}\times[-e,e]$, and the right side is $\{c\}\times[-e,e]$.  In 

636: the degenerate

637: case $b=0$, $f_{a,0}$ maps the left boundary to the point 

638: $(a-c^2,-c)$ which is to

639: the left of $B_{2,r}$; and the right boundary goes to $(a-c^2,c)$ 

640: which is directly

641: above

642: $(a-c^2,-c)$.  If $b\ne0$, then the image of the left boundary will 

643: continue to be

644: to the left of $B_{2,r}$ and below the image of the right boundary.  The use we

645: make of this combinatorial/topological information is given in Proposition 3.4,

646: whose proof is a straightforward consequence of the preceding discussion.

647: 

648: \proclaim Proposition 3.4.   Suppose $a,b\in\cA_r$.  Suppose,

649: too, that $A_1$ and $A_2$ are curves that cross $B_{1,r}$ from left 

650: to right and

651: that $A_1$ lies below $A_2$ inside $B_{1,r}$.  If $f$ preserves 

652: orientation, then

653: the curves $C_1=\tilde

654: f_{12}A_1$ and $C_2=\tilde f_{12}A_2$ open to the left, and $C_2$ 

655: lies inside $C_1$

656: as illustrated in Figure 7.  If $f$ reverses orientation, then the relative

657: positions of $C_1$ and

658: $C_2$ are exchanged.

659: 

660: \bigskip

661: \epsfxsize4.5in

662: \centerline{\epsfbox{mapatbend1_1.eps}}

663: \centerline{Figure 7:  Curve $C_2$ lies inside $C_1$: three possibilities.}

664: \bigskip

665: In the sequel, we will work with parameter values in $\cA_r$.  However, for many of the

666: arguments  the essential point is that $(\cB,\cG)$ is a system of crossed mappings with

667: the ``combinatorial'' behavior given in Figures 5 and 6.  Thus we are led to the

668: following condition:

669: $$\eqalign{&(\cB,\cG)\ \hbox{\rm is a family of real, crossed mappings, with the} \cr

670: &{\rm\

671: topological\ configurations\ shown\ in\ Figures\ 5,\ 6,\ and\ 7.}}\eqno(\dag)$$

672: We may summarize the discussion above by the statement: {\sl If $a,b\in\R$ and (\ddag) holds,

673: then (\dag) holds.}   We also introduce the two conditions

674: $$\#(W^s_{02}\cap W^u_{012}\cap B_{2,r})=2{\rm \ if\ }b>0, {\rm\ and\ }

675: \#(W^s_{02}\cap W^u_{12}\cap B_{2,r})=2{\rm\ if\ }b<0.

676: \eqno(*)$$

677: $$\#(W^s_{02}\cap W^u_{01212}\cap B_{2,r})=4{\rm \ if\ }b>0, {\rm\ and\ }

678: \#(W^s_{02}\cap W^u_{12012}\cap B_{2,r})=4 {\rm \ if\ }b<0.\eqno(**)$$

679: \noindent{\bf Remark on notation.}  We have now defined a parameter domain $\cA_r$ as

680: well as three conditions that may or may hold for a given parameter value

681: $(a,b)$.  The condition (\dag) requires

682: the boxes

683: $\cB$ to have specified behavior under $f$ and $f^{-1}$.  The conditions $(*)$ and

684: $(**)$ define dynamical characteristics of $f_{a,b}$.  It will be shown below that

685: (\dag) holds for all parameters in $\cA_r$ and that $(**)$ implies $(*)$.

686: \proclaim Proposition 3.5.  If (\dag) holds, then $(**)\Rightarrow(*)$.

687: 

688: \give Proof.  We will treat only the case $b<0$ since the case $b>0$ 

689: is similar.

690: Let us suppose that $(*)$ fails.  We map $W^u_1$ forward under

691: $\tilde f_{12}$ to $W^u_{12}$.  By  Proposition 3.1, $W^u_{12}\cap 

692: B_{2,r}$ is a

693: nonempty, connected curve, and by Proposition 3.4  it forms a curve 

694: which opens to

695: the left, which by hypothesis does not intersect

696: $W^s_{02}$.  This is pictured in the pair of boxes on the left hand 

697: side of Figure 8.  Next we map

698: $W^u_{12}$ forward under

699: $\tilde f_{20}$. Again by Proposition 3.1, $W^u_{120}\cap B_{0,r}$ is 

700: a nonempty,

701: connected curve.  Since the sign of

702: $f_{20}$ is

703: $(\cdot,-)$, the

704: $x$-direction of the curve is reversed, so $W^u_{120}\cap B_{0,r}$ opens to the

705: right.  By Proposition 3.2,

706: $W^u_{120}$ lies above $W^u_0=W^u_{00}$, which is drawn in gray as a 

707: visual aid to

708: the reader, although it is not necessary for the proof.  (The gray dot is

709: $p_0$, and $W^u_1$ is above $W^u_{01}$ in $B_{0,r}$ by Proposition 3.2.)  Since the

710: sign of $f_{21}$ is $(+,\cdot)$, the vertical orientation is preserved, so

711: $W^u_{121}$ contains $W^u_1$ as well as a curve below it.  Since the sign of

712: $f_{20}$ is $(+,\cdot)$, the vertical orientation is preserved, so 

713: the upper part

714: of $W^u_{120}$ with a single hash mark is identified with $W^u_1$ on the set

715: $B_{0,r}\cap B_{1,r}$.  Since

716: $W^s_{02}\cap W^u_{12}=\emptyset$, $W^u_{120}$ is disjoint from $W^s_0$, so we

717: obtain the picture as in the right hand pair of boxes in Figure 8.

718: 

719: \bigskip

720: \epsfxsize4.5in

721: \centerline{\epsfbox{cimplies_1.eps}}

722: \centerline{Figure 8}

723: \medskip

724: 

725: Next we map $W^u_{120}$ forward under $\tilde f_{01}$.  This is shown 

726: in the left

727: hand picture of Figure 9.  Since $f_{01}$ has signature $(-,\cdot)$, 

728: the vertical

729: orientation is reversed, so $W^u_{1201}$ lies below $W^u_{121}$ and 

730: $W^u_{01}$ in

731: $B_{1,r}$.  Finally, we map forward under

732: $\tilde f_{12}$ and obtain the picture in the right hand box of 

733: Figure 9.  The two

734: arches of $W^u_{12012}$ lie inside $W^u_{12}$ by Proposition 3.4.  Thus

735: $W^u_{12012}$ cannot intersect $W^s_{02}$, so condition $(**)$ does not hold.

736: \qed

737: 

738: \bigskip

739: \epsfxsize2.9in

740: \centerline{\epsfbox{cimpliescont_1.eps}}

741: \centerline{Figure 9}

742: \medskip

743: Figure 10 illustrates conditions $(*)$ and $(**)$ in the case $b>0$.  To understand

744: Figure 10, start in the left hand box with $W^u_0$ and $W^s_0$ 

745: passing through the

746: saddle point $p_0$.  We move $W^u_0$ to box $B_{1,r}$ via the map 

747: $\tilde f_{01}$,

748: and to box $B_{2,r}$ via $\tilde f_{02}$. The map $f_{02}$ has degree 2, and

749: $\tilde f_{02}W^u_{01}=W^u_{012}$ is a curve of degree 2 which opens 

750: to the left by

751: Proposition 3.4.  By condition $(*)$, $W^u_{012}$ crosses $W^s_{02}$. 

752: The crossed

753: map $f_{20}$ has degree 1 and sign $(\cdot,-)$, so the left-opening, degree two

754: curve

755: $W^u_{012}$ produces a degree two curve $W^u_{0120}=\tilde f_{20}W^u_{012}$ in

756: $B_{0,r}$ which opens to the right.  Condition $(*)$ maps forward under 

757: $f_{20}$, so

758: $W^u_{0120}$ intersects $W^u_0$.

759: 

760: The crossed map $f_{21}$ has degree 1, so $\tilde

761: f_{21}W^u_{012}=W^u_{0121}$ has degree 2 and by Proposition 3.2, it  lies above

762: $W^u_{01}$.  Now $(\tilde f_{02}\cup \tilde f_{21})(W^u_{012})$ is a curve in

763: $B_{0}\cup B_1$ of degree 2, and since $W^u_{0120}\cap B_{0,r}$ is 

764: connected, it

765: follows that $W^u_{0121}\cap B_{1,r}$ consists of two curves of degree 1.  By

766: Proposition 3.1, then, it follows that the complex variety 

767: $W^u_{0121}$ consists of

768: two irreducible components.  Now we map $W^u_{0121}\cap B_{1,r}$ under $\tilde

769: f_{12}$, which has degre 2.  By Proposition 3.4, $W^u_{01212}=\tilde

770: f_{12}W^u_{0121}$ lies inside $W^u_{012}$.  By $(**)$, $W^u_{01212}$ intersects

771: $W^s_{02}$.  Note that the arrangement of $W^u_{01212}$ corresponds 

772: to one of the

773: possibilities in Figure 7.  Another possiblity is given in the right 

774: hand of Figure 11.  This picture is mapped forward under $\tilde f_{20}$, to show one

775: possibility for $W^u_{012120}$ inside $B_{0,r}$.

776: 

777: \bigskip

778: \epsfxsize4.5in

779: \centerline{\epsfbox{condition_3.eps}}

780: \centerline{Figure 10: Moving $W^u_0$ forward along the sequence 

781: 01212 (case $b>0$)}

782: 

783: \bigskip

784: \epsfxsize3.2in

785: \centerline{\epsfbox{condition_alt_2.eps}}

786: \centerline{Figure 11: Alternative to Figure 10}

787: 

788: Figure 12 deals with the orientation-reversing case and shows various unstable

789: pieces $W^u_I$ starting with $W^u_1$ through $p_1$ and moving forward along the

790: sequences $I=1200$, 1201, and 12012.  The construction of this picture 

791: was explained in

792: large part in the proof of Proposition 3.5, so we do not repeat it here.

793: 

794: \bigskip

795: \epsfxsize4.5in

796: \centerline{\epsfbox{condition_or_2.eps}}

797: \centerline{Figure 12: Moving $W^u_1$ forward along the sequences 

798: 1200, 1201, and

799: 12012 (case

800: $b<0$)}\medskip

801: When $(**)$ holds, we use Figures 10 and 12 to define $S^\pm$ as the closed

802: subintervals of the left hand component of $\partial_vB_{2,r}$ which meet each

803: component of $\bar W^u_{012}\cup \bar W^u_{01212}$ if $b>0$ (resp.\ 

804: each component

805: of $\bar W^u_{12}\cup\bar W^u_{12012}$ if $b<0$).

806: \proclaim Proposition 3.6.  Suppose that $b>0$ and that 

807: (\dag) and $(**)$

808: hold. Let $I$ be an admissible sequence starting with 0 and ending 

809: with $k$, and

810: let $\Gamma$ be a connected component of $W^u_I$.  Then we have the following:

811: \vskip0pt If $k=0$, then: $\Gamma$ is disjoint from the component of

812: $B_{0,r}-W^u_0$ lying below $W^u_0$.  If $\delta(\Gamma)\ne1$,

813: then $\delta(\Gamma)=2$, and

814: $\Gamma$ intersects $W^s_0\cap B_{0,r}$, and $\bar\Gamma$ intersects 

815: the right hand

816: component of

817: $\partial_vB_{0,r}$ in two points.

818: \vskip0pt If $k=1$, then:  $\delta(\Gamma)=1$, and $\Gamma$ is 

819: disjoint from the

820: topmost and bottommost components of $B_{1,r}-(W^u_{01}\cup W^u_{0121})$.

821: \vskip0pt If $k=2$, then:  $\Gamma$ is disjoint from the innermost 

822: and outermost

823: components of $B_{2,r}-(W^u_{012}\cup W^u_{01212})$.  If

824: $\delta(\Gamma)\ne1$, then $\delta(\Gamma)=2$, and $\bar\Gamma$ intersects both

825: $S^+$ and

826: $S^-$.

827: 

828: \give Proof.  The proof proceeds by induction on the length of the 

829: sequence $I$.

830: First, the case $I=0$ is clear.  Now we suppose that the Proposition holds for

831: $I=I'i$.  We will show that if $(i,j)\in\cG$, then the Proposition holds for

832: $I=I'ij$ by considering five cases.

833: 

834: Case $ij=00$.  Since $f_{00}$ has sign $(+,\cdot)$, $f_{00}$ maps the 

835: component of

836: $B_{0,r}-W^u_0$ above $W^u_0$ to itself.  So $\tilde f_{00}\Gamma$ is 

837: disjoint from

838: the component of $B_{0,r}-W^u_0$ below $W^u_0$.  Now suppose 

839: $\delta(\Gamma)=2$.

840: $f_{00}$ maps $W^s_0$ into itself, and the sign of $f_{00}$ is $(\cdot,+)$, so

841: $\tilde f_{00}\Gamma$ intersects $W^s_0\cap B_{0,r}$, and 

842: $\bar\Gamma$ intersects

843: the right hand component of $\partial_vB_{0,r}$ in two points.

844: 

845: Case $ij=01$.  By Proposition 3.2, $f_{01}(B_{0,r})$ lies below 

846: $W^u_{0121}$.  On

847: the other hand $\Gamma$ is above $W^u_0$ and $f_{01}$ has sign $(+,\cdot)$, so

848: $\tilde f_{01}\Gamma$ is above $W^u_{01}$ in $B_{1,r}$.  It remains 

849: to show that

850: $\tilde f_{01}\Gamma$ consists of two components of degree 1.  For this, we may

851: assume that $\delta(\Gamma)=2$, and $\Gamma\cap W^s_0\cap B_{0,r}\ne\emptyset$.

852: Consider how $\gamma'=(\tilde f_{00}\cup \tilde 

853: f_{01})(B_{0,r}\cap\Gamma)$ maps

854: across

855: $B_{0,r}\cup B_{1,r}$: the left hand side of $\gamma'$ intersects $W^s_0\cap

856: B_{0,r}$ and the right hand side goes across the right hand boundary 

857: of $\partial_v

858: B_{1,r}$.  Thus $\gamma'\cap B_{1,r}$ consists of two curves.  By 

859: Proposition 3.1,

860: $\tilde g_{01}\Gamma\cap B_{0,1}=\gamma'\cap B_1$ consists of two 

861: disks of degree

862: one.

863: 

864: Case $ij=12$.  This is a direct consequence of Proposition 3.4.

865: 

866: Cases $ij=21$ and $ij=20$.  Let $\Gamma$ be as in case $k=2$.  We may 

867: assume that

868: $\delta(\Gamma)-2$.  Since $f_{20}$ and $f_{21}$ have sign 

869: $(\cdot,-)$, it follows

870: that  $\gamma':=(\tilde f_{20}\cup \tilde f_{21})\Gamma\cap 

871: (B_{0,r}\cup B_{1,r})$

872: is a 2-fold curve opening to the right.  Since $\bar\Gamma$ 

873: intersects both $S^+$

874: and $S^-$, we have $\Gamma \cap W^s_{02}\cap B_{2,r}\ne\emptyset$, 

875: and it follows

876: that $\gamma'\cap W^s_0\cap B_{0,r}\ne\emptyset$.  By Proposition 3.2, $\gamma'$

877: lies above $W^u_0$.  This finishes the case $ij=20$.

878: 

879: For the case $ij=21$, we observe that $\gamma'$ is a degree two real 

880: curve which

881: crosses $W^s_0\cap B_{0,r}$ and opens to the right.  Since 

882: $\bar\Gamma$ intersects

883: both $S^+$ and $S^-$, and $f_{21}$ has sign $(\cdot,-)$, it follows that

884: $\bar\gamma'$ intersects the right hand boundary of $\partial_vB_{1,r}$ in two

885: points.  Thus $\gamma'\cap B_{1,r}$ consists of two real curves 

886: crossing $B_{1,r}$

887: horizontally.  It follows from Proposition 3.1 that $\tilde 

888: f_{21}\Gamma$ consists

889: of two components of degree one.

890: \qed

891: 

892: In the following, we let $B^+_{0,r}$ denote the right-hand component of

893: $B_{0,r}-W^s_0$.

894: \proclaim Proposition 3.7.  Suppose that $b<0$, and that 

895: (\dag) and $(**)$

896: hold. Let $I$ be an admissible sequence starting with 1 and ending 

897: with $k$, and

898: let $\Gamma$ be a connected component of $W^u_I$.  Then we have the following:

899: \vskip0pt If $k=0$: $\Gamma$ is disjoint from the topmost and bottommost

900: components of $B^+_{0,r}-(W^u_{120}\cup W^u_{1200})$.  If 

901: $\delta(\Gamma)\ne1$, then

902: $\delta(\Gamma)=2$, and $\Gamma$ intersects $W^s_0\cap B_{0,r}$, and 

903: $\bar\Gamma$

904: intersects the right hand component of

905: $\partial_vB_{0,r}$ in two points.

906: \vskip0pt If $k=1$:  $\delta(\Gamma)=1$, and $\Gamma$ is disjoint from the

907: topmost and bottommost components of $B_{1,r}-(W^u_{1}\cup W^u_{1201})$.

908: \vskip0pt If $k=2$:  $\Gamma$ is disjoint from the innermost and outermost

909: components of $B_{2,r}-(W^u_{12}\cup W^u_{12012})$.  If

910: $\delta(\gamma)\ne1$, then $\delta(\Gamma)=2$, and $\bar\Gamma$ intersects both

911: $S^+$ and

912: $S^-$.

913: 

914: \give Proof.

915: This proof is analogous to the proof of Proposition 3.6; we omit the details.

916: \qed

917: 

918: \proclaim Proposition 3.8.  Suppose that (\dag) and $(**)$

919: hold.  Let $I$ be an admissible sequence of the form $I=0i_1\cdots 

920: i_n2$ if $b>0$

921: or $I=1i_1\cdots i_n2$ if $b<0$.  Then for each component $\Gamma$ of $W^u_I$,

922: $\#(W^s_{02}\cap \Gamma\cap B_{0,r})=\delta(\Gamma)$.  In particular, if the

923: intersection in the definition of $(**)$, is not tangential, then there is no tangency

924: between

925: $W^s_0$ and $W^u_I$.

926: 

927: \give Proof.  This follows from the case $k=2$ in Propositions 3.6 

928: and 3.7.  The

929: only case to consider is $\delta(\Gamma)=2$. Now if $\Gamma$ is not one

930: of the curves $W^u_I$ in condition $(**)$, $\Gamma\cap B_{2,r}$ is 

931: trapped between an

932: inner and an outer curve.  Since its closure intersects both

933: $S^+$ and

934: $S^-$, it must cross $W^s_{02}$ at least twice.  These two 

935: intersections account

936: for the total intersection number, and so these intersections must be simple

937: (nontangential), and there can be no further intersections.

938: \qed

939: 

940: \proclaim Proposition 3.9.  Suppose that (\dag) and $(**)$

941: hold.  Let $I$ be an admissible sequence of the form $I=0i_1\cdots 

942: i_n0$ if $b>0$

943: or $I=1i_1\cdots i_n0$ if $b<0$.  Then for each component $\Gamma$ of $W^u_I$,

944: $\#(W^s_0\cap \Gamma\cap B_{0,r})=\delta(\Gamma)$.  In particular, if the

945: intersection in the definition of $(**)$, is not tangential, then there is no tangency

946: between

947: $W^s_0$ and $W^u_I$.

948: 

949: \give Proof.  This follows by applying the map $\tilde f_{20}$, which 

950: has degree

951: one, to the result of Proposition 3.8.\qed

952: 

953: This allows us to characterize the mappings of maximal entropy.

954: 

955: \proclaim Theorem 3.9.  Suppose that (\dag) holds.  If the real map

956: $f_{a,b}$ has entropy equal to $\log 2$, then $(**)$ holds.    Conversely, if

957: $S\subset\{(a,b)\in\R^2:b\ne0\}$ is a connected set such that $(**)$  holds for all

958: $(a,b)\in S$, and if $f_{a_0,b_0}$ has entropy $\log 2$ for some  $(a_0,b_0)\in S$,

959: then $f_{a,b}$ has entropy $\log2$ for all $(a,b)\in S$.

960: 

961: \give Proof.  The proof will be  based on the following criterion from [BLS]:

962: $f_{a,b}$ has (maximal) entropy $\log2$ if and only if for all saddle 

963: points $p$ and

964: $q$, all (complex) intersection points of $W^s(p)\cap W^u(q)$ belong to $\R^2$.

965: 

966: We

967: suppose first that the entropy of $f_{a,b}$ is $\log2$.  If $b>0$ we take

968: $p=q=p_0$.  By (2.7), $\delta(W^s_{02})=1$ and $\delta(W^u_{01212})=4$.

969: If $b<0$, we take $p=p_0$ and $q=p_1$.  Again by (2.7), we have

970: $\delta(W^u_{12012})=4$. By (2.6) we have

971: $\#(W^s_{02}\cap\Gamma)=\delta(W^s_{02})\delta(\Gamma)=4$ with

972: $\Gamma=W^u_{01212}$ if $b>0$ and $\Gamma=W^u_{12012}$ if $b<0$.  By 

973: the criterion

974: above, all (complex) intersections between

975: $W^s_{02}$ and $\Gamma$ must belong to $\R^2$, so it follows that $(**)$ holds.

976: 

977: Now let us suppose that $(**)$ holds for all $(a,b)\in S$.  Consider the subset

978: $S_0$ of points $(a,b)\in S$ such that the entropy of $f_{a,b}$ is equal to

979: $\log 2$.  Since

980: $(a,b)\mapsto {\rm entropy}(f_{a,b})$ is continuous, it follows $S_0$ 

981: is a closed

982: subset of $S$.  Since

983: $S$ is connected, it suffices to show that $S_0$ is an open subset of $S$.

984: Let us fix a point $(a_0,b_0)\in S_0$.  By Proposition 2.5

985: there is an open set $U_0\subset\C^2$ such that $\bar B_{0,r}\cup\bar

986: B_{1,r}\cup\bar B_{2,r}\subset U_0$, and

987: $$U_0\cap f_{a,b}(B_0\cup B_1\cup B_2)\subset f_{a,b}(B_0\cup B_2)$$

988: $$U_0\cap  f_{a,b}(B_0\cup B_1\cup B_2)\subset f_{a,b}(B_1)$$

989: holds for $(a,b)=(a_0,b_0)$.  Thus it holds for $(a,b)$ in a small 

990: neighborhood of

991: $(a_0,b_0)$.   Thus we also have that 

992: $K_{a_0,b_0}\subset\R^2$ since

993: $f_{a_0,b_0}$ has maximal entropy.  By Proposition 2.1, then,

994: $K_{a_0,b_0}\subset B_{0,r}\cup B_{1,r}\cup B_{2,r}\subset U_0$.  Since

995: $(a,b)\mapsto K_{a,b}$ is upper semicontinuous, it follows that for $(a,b)$

996: sufficiently close to $(a_0,b_0)$ we have $K_{a,b}\subset U_0$, and thus

997: $f_{a,b}$ satisfies the hypotheses of Proposition 2.4.

998: 

999: Now we consider the case $b>0$; the argument for the case $b<0$ is 

1000: similar and is

1001: omitted.  Let

1002: $q\in W^s(p_0)\cap W^u(p_0)$ be any point of intersection.  Replacing $q$ by

1003: $f^{-m}q$ if necessary, we may assume that $q\in B_0$.  Let

1004: $I$ denote the admissible sequence given by Proposition 2.4.  For $n$ 

1005: sufficiently

1006: large, we have

1007: $f^nq\in W^s_{0}$, which is a neighborhood of $p_0$ inside $W^s(p_0)$.  Thus,

1008: writing $I(n):=i_0i_1\cdots i_n$, we have

1009: $f^nq\in W^u_{I(n)}$.  By Proposition 3.8, it

1010: follows that

1011: $W^s_{0}\cap W^u_{I(n)}\subset\R^2$.

1012: Since $f^nq\in W^s_{0}\cap W^u_{I(n)}$, it follows that $q\in\R^2$.  Thus

1013: $W^s(p_0)\cap W^u(p_0)\subset\R^2$, so that $f_{a,b}$ has entropy 

1014: equal to $\log 2$.

1015: \qed

1016: 

1017: \noindent{\bf Remark.}  There is an alternative approach to the ``Conversely'' part of this

1018: Theorem.  Namely, we could use the arguments of this section to show that $W^s_{02}$ and

1019: $W^u_I$ have certain trellis properties, and then we can apply the work of P. Collins [Co] to

1020: conclude that the real map $f$ has entropy $\log 2$.

1021: 

1022: 

1023: \section 4.  The Quadratic Horseshoe Locus

1024: 

1025: In this section we analyze the real, maximal entropy bifurcations in a neighborhood of

1026: (2,0). 

1027: 

1028: \proclaim Lemma 4.1.  Suppose that $(a,b)\in\cA$, and suppose that

1029: $\Delta/3\le\delta\le e^2-4-2\Delta$.  If we define $B_0'$ and $B_2'$ by

1030: $$B_0':=\{|x+2|<\delta,|y|<e\}, \ \ B_2':=\{|x-2|<\delta,|y|<e\}$$ then $f$ induces crossed

1031: mappings from $B_0'$ to itself and from 

1032: $B_0'$ to $B_2'$. In particular, the sets $W^s_0$ and $W^s_{02}$ (as in (2.10)) are given by

1033: $$W^s_0=\bigcap_{n\ge0}f^{-n}B_0', {\rm\ and\ }  W^s_{02}=\bigcap_{n\ge1}B_2'\cap

1034: f^{-n}B_0'.$$

1035: 

1036: \give Proof.  Let us fix $\delta$ such that $\Delta/3\le \delta\le  e^2-4-2\Delta$ and set

1037: $B_0':=\{|x+2|<\delta,|y|<e\}$.  By the upper bound on 

1038: $\delta$, we have

1039: $\{|4-x^2|<\Delta+\delta\}\subset\{|x|<e\}$.  We compute

1040: $$\eqalign{f^{-1}B_0'\cap\{|y|<e\} &= \{|2+\pi_vf(x,y)|<\delta, |\pi_hf(x,y)|<e,|y|<e\}\cr

1041: &\subset \{|2+a-x^2-by|<\delta,|x|<e,|y|<e\}\cr

1042: &\subset\{|4-x^2|<|a-2|+|by|+\delta,|x|<e\}\subset\{|4-x^2|<\Delta+\delta,|x|<e\}

1043: \cr &\subset\{|2-x|<\sqrt{4+\Delta+\delta}-2\}\cup

1044: \{|2+x|<\sqrt{4+\Delta+\delta}-2\}\cr

1045: &\subset\{|2-x|<{\Delta+\delta\over4}\}\cup\{|2+x|<{\Delta+\delta\over4}\}.\cr}$$ In the next

1046: to last line we have removed the condition $|x|<e$ by the  upper bound condition on $\delta$. 

1047: The last line uses the concavity of of the square root. By the lower bound on $\delta$, we

1048: have $(\Delta+\delta)/4<\delta$,  so it follows that $f^{-1}\bar B_0'\cap\partial_v

1049: B'_0=\emptyset$.

1050: 

1051: Next we consider a point $(x',y')\in f^{-1}(\partial_h B'_0)$.  By (?.5),

1052: $$\eqalign{|y'|=|{1\over b}(a-y^2-x)| &> {e\over 

1053: \Delta}(|y|^2-4-|a-2|-|x+2|)\cr &> {e\over\Delta}(e^2-4-\Delta-\delta).\cr}$$ This last

1054: quantity is greater than $e$ by the upper bound on $\delta$, so

1055: $(x',y')\notin\bar B_0'$.  Thus $f$ induces a crossed mapping from

1056: $B_0'$ to itself.  The proof that $f$ induces a crossed mapping from $B_0'$ to

1057: $B_2'$ is the same

1058: \qed

1059: 

1060: \proclaim Corollary 4.2.  If $(a,b)\in\cA$, then $(*)$ holds.

1061: 

1062: 

1063: \proclaim Proposition 4.3.  If $(a,b)\in\cA$, then the horizontal

1064: disk $W^u_1$ has degree one.

1065: 

1066: \give Proof.  Let $\Gamma\in\cD^1_{h,r}(B_1)$ be any real, horizontal  disk.  Then by

1067: Proposition 3.4, $\tilde f_{12}\Gamma\in\cD^2_{h,r}(B_2)$ is a real disk of degree two which

1068: opens to the left. Applying $(\tilde  f_{20}\cup\tilde f_{21})$ to

1069: $\tilde f_{12}\Gamma$, we obtain a disk $\Gamma'$ of degree two, which is horizontal in

1070: $B_0\cup B_1$.  There can be at most one critical point for the projection $\pi_v:\Gamma'\to

1071: B_0\cup B_1$, and if there is a critical point, it must be real, since its conjugate is also a

1072: critical point.

1073: 

1074: Since the sign of $f_{20}\cup f_{21}$ is $(\cdot,-)$, $\Gamma'$ opens to the right.  By

1075: Proposition 3.1, $\tilde f_{20}\tilde f_{12}\Gamma=\Gamma'\cap B_0$ defines a nonempty real

1076: curve in $B_{0,r}$.  Thus, if there is a  critical point, then vertical projection

1077: $\pi_v:\Gamma\cap(B_{0,r}\cup B_{1,r})\to(-d,c)$ has a critical  point.  Since $(*)$

1078: holds, this critical point must belong to $B_{0,r}$, and by (\ddag), this point cannot

1079: belong to $B_{1,r}$.  In particular, it follows that

1080: $\pi_v$ has no critical point  in $\tilde f_{12}\tilde f_{12}\Gamma=\Gamma'\cap B_1$.  Thus

1081: $\Gamma'\cap B_1$ consists of two components.  Since 

1082: $p_1\in B_1\cap B_2$ is a fixed point, one of these components contains $p_1$, and  we denote

1083: this component by $(\tilde f_{21}\tilde f_{12})^\#\Gamma$, which is a disk  of degree one.

1084: 

1085: Now if we choose $\Gamma$ to pass through $p_1$ such that its tangent  at $p_1$ is transverse

1086: to $W^s(p_1)$, then .  It follows that

1087: $(\tilde f_{12}\tilde f_{21})^{\#n}\Gamma$ is a sequence of horizontal disks of degree one,

1088: passing through

1089: $p_1$, which converge to $W^u_1$ as $n\to\infty$.  \qed

1090: 

1091: 

1092: Now let us examine the case $b=0$.  The image of $f_{a,0}$ is the parabola

1093: $$\Gamma:=f_{a,0}(\C^2)=\{x=a-y^2\}=\{(p(t),t):t\in\C\},$$

1094: where $p(z)=a-z^2$.  Throughout our discussion, we assume that 

1095: $|a-2|<\Delta$.  Thus

1096: $a\notin D_0\cup D_1$, and so there are two holomorphic branches of

1097: $p^{-1}(z)=\pm\sqrt{a-z}$ over

1098: $D_0\cup D_1$.  For $j=0,1$, $\Gamma\cap B_j$ consists of two components

1099: $\Gamma'_j$ and

1100: $\Gamma''_j$, each of which is a smooth graph of a branch of $p^{-1}$.  We note

1101: that $f_{a,0}$ is injective on each component $\Gamma'_j$ and $\Gamma''_j$,

1102: $j=0,1$.  On the other hand, $\Gamma\cap B_2$ is connected, and $f_{a,0}$ is

1103: two-to-one on $\Gamma\cap B_2$.

1104: 

1105: Let $p_0=(t_0,t_0)$ denote the fixed point which belongs to $B_0$.  (The following discussion

1106: can be adapted to work with the other fixed point $p_1\in  B_1\cap B_2$, as well.)  Let

1107: $\varphi_a:\C\to\C$ denote the linearizing coordinate such that 

1108: $\varphi_a(0)=t_0$,

1109: $\varphi_a'(0)=1$, and $p(\varphi_a(\zeta))=\varphi_a(\lambda\zeta)$, where

1110: $\lambda:=p'(t_0)$.  If we write $\varphi=\varphi_a$, it follows that

1111: $$\psi_{a,0}(\zeta):=(\varphi(\zeta),\varphi(\lambda^{-1}\zeta))$$ defines a mapping

1112: $\psi_{a,0}:\C\to\Gamma$ which satisfies

1113: $f_{a,0}\circ\psi(\zeta)=\psi(\lambda\zeta)$.

1114: 

1115: We wish to define the sets $W^u_I$ in the case $b=0$.  We let $W^u_0$ be the connected

1116: component of

1117: $\Gamma\cap B_0$ containing $p_0$;  $W^u_{01}$ is the connected component of

1118: $\Gamma\cap B_1$ which intersects $W^u_0$; and $W^u_{012}=\Gamma\cap B_2$. As we  try to

1119: consider longer $I$, we run into the difficulty that the mappings

1120: $\tilde f_{ij}$ are not invertible.  To deal with this, we identify $W^u_I$ in terms of the

1121: parametrization $\psi_{a,0}$ of $\Gamma$.  To do this, let

1122: $\Omega_0\subset\C$ be the connected component of

1123: $\psi^{-1}_{a,0}(W^u_0)=\varphi^{-1}(D_0)$ which contains the origin.  In general, we set

1124: $$\Omega_I:=

1125: \lambda^n\Omega_0\cap\varphi^{-n}D_{i_1}\cap\cdots\cap\varphi^{-n}(D_{i_n})

1126: =\lambda^n\Omega_0\cap\psi_{a,0}^{-n}B_{i_1}\cap\cdots\cap\psi_{a,0}^{-n}(B_{i_n}),

1127: \eqno(4.1)$$ where $I=0i_1\cdots i_n$ is an admissible sequence.  We then identify 

1128: $W^u_I$ in terms of the map $\psi_{a,0}:\Omega_I\to W^u_I$.

1129: 

1130: The usefulness of the case $b=0$ is that it is the limit of the case $b\ne0$. When $b\ne0$, we

1131: let

1132: $\psi_{a,b}:\C\to W^u_{p_0}$ be the uniformization of $W^u(p_0)$,  normalized by the condition

1133: $(\pi_v\circ\psi_{a,b})'(0)=1$.  In this case,

1134: $(a,b,\zeta)\mapsto\psi_{a,b}(\zeta)$ is holomorphic, and we have

1135: $$\lim_{b\to0}\psi_{a,b}=\psi_{a,0},\eqno(4.2)$$ with uniform convergence on compact subsets.

1136: Restricting this to the image of $\Omega_I$, we have:

1137: $$\lim_{b\to0}W^u_I(f_{a,b})=W^u_I(f_{a,0}),\eqno(4.3)$$ where the convergence is in the sense

1138: of the Hausdorff topology.  Taking multiplicities of $W^{s/u}_I(f_{a,0})$ into account, the

1139: convergence  also holds in the sense of currents.

1140: \proclaim Lemma 4.4.  If $|a-2|<\Delta$, then for $I=01212$ and 

1141: $I=12012$,

1142: $\Omega_I$ consists of two connected components with disjoint closures.  If

1143: $b\ne0$ is sufficiently small, then $W^u_I$ consists of two components.

1144: 

1145: \give Proof.  Since $p:D_0\to p(D_0)$ is a conformal equivalence, and

1146: $p^{-1}D_0\subset D_0$, we may define a holomorphic map

1147: $\lim_{n\to\infty}\lambda^np^{-n}:p(D_0)\to\C$  This is the inverse  of $\varphi$, and so

1148: $\varphi:\lambda\Omega_0\to p(D_0)$ is univalent.  Thus 

1149: $\Omega_{01}=\Omega_0\cap

1150: \varphi^{-1}(D_1)$ is connected and relatively compact in $\Omega_0$.  Let $c_{01}$ be the

1151: unique point of $\lambda\Omega_0$ such that 

1152: $\varphi(c_{01})=0$.  It follows that $\varphi'(\lambda c_{01})=(p\circ\varphi(c_{01}))' = 

1153: p'(0)\varphi'(c_0)=0$. Conversely, if $\zeta\in\lambda^2\Omega_0$, and if

1154: $0=\varphi'(\zeta)= p'(\varphi(\lambda^{-1}\zeta)\varphi'(\lambda^{-1}\zeta)\lambda^{-1}$, 

1155: then we must have $p'(\varphi(\lambda^{-1}\zeta))=0$ since $\varphi'\ne0$ on 

1156: $\lambda\Omega_0$. It follows that $\zeta=\lambda c_{01}$, so $\lambda c_{01}$ is the  unique

1157: critical point in $\lambda^2\Omega_0$.  It follows that

1158: $\psi_{a,0}(\zeta)=(\varphi(\zeta),\varphi(\lambda^{-1}\zeta))$ has no critical point on

1159: $\lambda^2\Omega_0$.  Since 

1160: $\psi_{a,0}(\Omega_{012})=\Gamma\cap B_2$ is simply connected, it follows that

1161: $\psi_{a,0}:\Omega_{012}\to\Gamma\cap B_2$ is univalent. By the argument above,

1162: $\lambda^2c_{01}$ is the unique critical point for

1163: $\psi_{a,0}$ in $\lambda^3\Omega_0$.  Now

1164: $\psi_{a,0}(\lambda^2c_{01})=f_{a,0}(a,0)=(a-a^2,a)$, which does not belong to

1165: $B_1$, since $\Re(a-a^2)<-c$.  It follows that $\psi_{a,0}$ is  unbranched on the closure of

1166: $\lambda^3\Omega_0\cap\psi^{-1}_{a,0}(B_1)$.  Recall that

1167: $f_{a,0}:W^u_{012}=\Gamma\cap B_2\to f_{a,0}(W^u_{012})$ is a mapping of degree two.  Thus

1168: $W^u_{0121}$ is the component of $\Gamma\cap B_1$ which is  disjoint from

1169: $W^u_{01}$, and $W^u_{0121}$ has multiplicity two.  It follows that

1170: $\psi_{a,0}:\Omega_{0121}\to W6u_{0121}$ is a covering of degree two.  Since

1171: $\psi_{a,0}$ is unbranched on the closure of

1172: $\Omega_{0121}\subset\lambda^3\cap\psi^{-1}_{a,0}(B_1)$, it follows that

1173: $\Omega_{0121}$ consists of two components with disjoint closures.

1174: 

1175: Let us move forward one more step: since

1176: $f_{a,0}$ is injective on

1177: $W^u_{0121}$, it follows that $\psi_{a,0}$ gives a conformal  equivalence between each

1178: component of $\lambda\Omega_{0121}$ and $f_{a,0}W^u_{0121}$. Intersecting

1179: $\lambda\Omega_{0121}$ with $\psi_{a,0}^{-1}(B_2\cap

1180: f_{a,0}(W^u_{0121}))=\psi_{a,0}^{-1}(B_2\cap\Gamma)=\varphi^{-1}(D_2)$,  then

1181: $\Omega_I$ consists of two components $\Omega_I'$ and

1182: $\Omega_I''$ which have disjoint closures.  If $b\ne0$ is  sufficiently small, then

1183: $\psi_{a,b}^{-1}(W^u_I)$ will be close to $\Omega_I$.  Thus it (as  well as $W^u_I$) has two

1184: components.

1185: \qed

1186: 

1187: Now we pass from unstable manifolds to stable manifolds.  The  vertical complex line through

1188: the fixed point $p_0$ is mapped to $p_0$ under $f_{a,0}$.  If we write

1189: $p_0=(t_0,t_0)$, then

1190: $$W^s_0=\{(x,y):x=t_0,|y|<e\},{\rm\ and\ } W^s_{02}=\{(x,y):x=t_0',|y|<e\},$$ where

1191: $t_0'\in\C$ is the solution to $p(t_0')=t_0$ such that $t_0'\ne t_0$. If $(\ddag)$ holds, then

1192: $$W^s_{02}\cap\Gamma=\{(\zeta,\pm\sqrt{a-\zeta}):\zeta=t'_0\}.\eqno(4.4)$$ This intersection

1193: consists of two distinct points unless $t'_0=a$,  which happens exactly when $a=2$.  We can

1194: work our way backwards, taking successive  preimages, to define $W^s_J(f_{a,0})$ for an

1195: admissible sequence $J$.  As in the  case of unstable manifolds, we have

1196: $$\lim_{b\to0}W^s_J(f_{a,b})=W^s_J(f_{a,0}).\eqno(4.5)$$

1197: \proclaim Proposition 4.5.  Suppose that $(a,b)\in\cA$ and

1198: $|a-2|\ge(e+\Delta)|b|$.  Then for $I=01212$ and $I=12012$,

1199: $W^s_{02}$ intersects $W^u_I$ in four distinct points, and thus the  intersection is not

1200: tangential.

1201: 

1202: \give Proof.   We begin by noting

1203: $$W^u_{I}\subset B_2\cap fB_0\subset

1204: \{|b^{-1}(a-x-y^2)|<\delta,|y|<e\}.$$  If we set $\delta=\Delta/3$,  then by Lemma 3.5 we have

1205: 

1206: $$W^s_{02}\subset B_2'\subset\{|x-2|<{\Delta+\delta\over4},|y|<e\}.$$ Thus

1207: $$\eqalign{W^s_{02}\cap W^u_{I} &\subset\{|x-2|<{\Delta+\delta\over4},|a-x-y^2|<\delta|b|\}\cr

1208: &\subset\{|x-2|<{\Delta+\delta\over4},|a-2-y^2|<|b|\delta+|x-2|\}\cr

1209: &\subset\{|a-2-y^2|<|b|\delta + {\Delta+\delta\over4}\}\cr

1210: &=:U_{a,b}.\cr}$$

1211: The set $U_{a,b}$ is symmetric with respect to $y\mapsto-y$ and is seen to be

1212: disconnected if (and only if) it does not contain $y=0$.  This occurs exactly

1213: when

1214: $|a-2|\ge|b|\delta +

1215: {\Delta+\delta\over4}$.  Now we recall that $\delta=\Delta/3$ and 

1216: substitute the

1217: condition

1218: $(\ddag)$, which gives

1219: $|a-2|\ge |b|\Delta/3 +\Delta/3\ge |b|\Delta/3 +(|a-2|+e|b|)/3$, and this is

1220: equivalent to $|a-2|\ge(\Delta+e)|b|$.

1221: 

1222: Now consider the case $b=0$.  By Lemma 4.4, $\Omega_I$

1223: consists of components

1224: $\Omega_I'$ and

1225: $\Omega_I''$.  Since $a\ne2$, the intersection (4.4) contains two

1226: points, which lie in different components of $U_{a,b}$.  Thus

1227: $\psi_{a,0}(\Omega'_I)$ and $\psi_{a,0}(\Omega_I'')$ each intersect 

1228: $W^s_{02}$ in

1229: two points, which lie in different components of

1230: $U_{a,0}$.  If

1231: $b\ne0$,

1232: $|a-2|\ge(e+\Delta)|b|$, then $W^u_{I}$ consists of two components

1233: ${(W^u_I)}'=\psi_{a,b}(\Omega'_I(a,b))$ and

1234: ${(W^u_I)}''=\psi_{a,b}(\Omega''_I(a,b))$.  Further,  the set

1235: $U_{a,b}$ continues to be disconnected, and by (4.3) the each component of

1236: $U_{a,b}$ will continue to contain a point of

1237: $W^s_{02}\cap {W^u_I}'$.  Since $\delta(W^u_I)'=2$ and  $W^s_{02}\cap 

1238: {(W^u_I)}'$

1239: contains two distinct points, the intersection is not tangential.  A similar

1240: argument for $(W^u_I)''\cap W^s_{02}$ shows that $W^s_{02}\cap W^u_I$ has no

1241: tangency.

1242: \qed

1243: Let us define $$\cD:=\{(a,b)\in\C^2: |a-2| < .237186,  |b|<.08205\}$$

1244: $$T_{I}:=\{(a,b)\in\cD:W^s_{02}{\rm\ intersects\ } W^u_{I}{\rm\ 

1245: tangentially}\}.$$

1246: In the definition of $T_I$, we interpret the case $b=0$ as follows.

1247: By \S1, we know that $T_I\cap\{b\ne0\}$ is a complex subvariety of

1248: $\cD-\{b\ne0\}$.  By (4.5) and (4.2), we have that 

1249: $(T_I\cap\{b\ne0\})\cup(2,0)$ is

1250: the closure of $T_I\cap\{b\ne0\}$ in $\cD$.  With this interpretation,  $T_I$ is a

1251: complex subvariety of $\cD$.

1252: 

1253: \proclaim Proposition 4.6.   For  $I=01212$

1254: and $12012$, $T_I$ is a complex subvariety of $\cD$ with the 

1255: following properties:

1256: \vskip0pt (i) The projection

1257: $\pi_h:T_I\to\{|b|<.08205\}$ is a proper mapping of degree two.

1258: \vskip0pt (ii) $T_I$ is locally reducible at $(2,0)$.

1259: \vskip0pt (iii)  There are real analytic functions

1260: $\kappa^\pm_I:[-.08205,.08205]\to

1261: \R$ with $\kappa_I^-(t)<\kappa_I^+(t)$ for $t>0$ such that

1262: $T_I\cap\R^2$ is the union of the graphs of $\kappa^+_I$ and $\kappa^-_I$.

1263: 

1264: \give Proof.  Note that with our values of $e$ and $\Delta$, (\ddag) holds for

1265: $(a,b)\in\cD$ whenever $b\ne0$.  Further, the condition

1266: $|a-2|\ge(e+\Delta)|b|$ holds for $(a,b)\in\partial_v\cD$.  By Proposition 4.5,

1267: then, $\bar T_I\cap \partial_v\cD=\emptyset$.  Thus $\pi_h$ is a 

1268: proper mapping.

1269: To determine the multiplicity of $\pi_h$, it suffices to determine 

1270: the multiplicity

1271: at

1272: $b=0$.  If $b=0$, then the only tangency occurs at $a=2$.  Now

1273: $W^u_{I}=\Gamma\cap B_2$, with multiplicity two, so in case $a=0$,

1274: $W^s_{02}$ makes a tangential intersection with each component of

1275: $W^u_I$.  It follows that $T_{I}\cap\{b=0\}=\{(2,0)\}$, with multiplicity two.

1276: Thus $\pi_h$ has multiplicity two.

1277: 

1278: For (ii), let $b=0$.  By Lemma 4.4, $\Omega_I$ consists of components 

1279: $\Omega_I'$

1280: and $\Omega_I''$ which have disjoint closures.  Thus, for $b\ne 0$ 

1281: small, there are

1282: domains $\Omega_I'(a,b)$ and $\Omega''_I(a,b)$ which are mapped under 

1283: $\psi_{a,b}$

1284: to the two components of $W^u_I$.  Thus for $|b|<r_0$ small, we may split $T_I$

1285: into $T_I'=\{(a,b)\in\cD:|b|<r_0, W^s_{02}{\rm\ intersects\ }W^u_I(a,b)'{\rm\

1286: tangentially}\}$, and a similar set $T_I''$ for $W^u_I(a,b)''$.

1287: 

1288: Now we consider the projection 

1289: $\pi_h:T_I\cap\R^2\to(-.08205,.08205)$.  This is a

1290: proper mapping of degree two.  Consider a point $(a,b)\in 

1291: T_I\cap\R^2$ with $b<0$

1292: and suppose that $I=12012$.  We may repeat the argument of Proposition 3.5

1293: to conclude that $W^u_{12012}$ consists of two curves in $B_{2,r}$ 

1294: which open to

1295: the left.  By $\gamma'$ and $\gamma''$ we denote the components of 

1296: $W^u_{12012}$

1297: such that

1298: $\gamma'\cap B_{2,r}$ forms the inner curve, and $\gamma''\cap 

1299: B_{2,r}$ forms the

1300: outer curve.

1301: 

1302: Let us note at the outset that $\delta(\gamma')=\delta(\gamma'')=2$, and so

1303: $\#(W^s_{02}\cap\gamma')=\#(W^s_{02}\cap\gamma'')=2$.  If there is a tangency

1304: between $\gamma'$ and

1305: $W^s_{02}$, then the tangency must be real.  For otherwise, if there 

1306: were a point

1307: of tangency

1308: $q\in B_2-B_{2,r}$, the complex conjugate $\bar q$ would also be a point of

1309: tangency, so the total intersection of

1310: $\gamma'$ and $W^s_{02}$ in $B_2$ would be at least four.

1311: 

1312: Now suppose that the outer curve $\gamma''$ is tangential to

1313: $W^s_{02}$.  Then this point of tangency must have order two, and can 

1314: be the only

1315: intersection with

1316: $W^s_{02}$ since the total intersection satisfies $\#(W^s_{02}\cap

1317: W^u_{12012})=2$.  Since $\gamma''\cap B_{2,r}$ opens to the left, it 

1318: follows that

1319: $\gamma''\cap B_{2,r}$ must lie to the left of

1320: $W^s_{02}$.  Thus $\gamma'$ cannot intersect

1321: $W^s_{02}\cap B_{2,r}$.  Thus there can be no tangency between the 

1322: complex disks

1323: $W^s_{02}$ and

1324: $\gamma'$.

1325: 

1326: Thus in the case $b\ne0$, with $a$ and $b$ both real, there cannot be

1327: tangencies (necessarily real) between both components of $W^u_I$ and 

1328: $W^s_{02}$.

1329: In other words, if

1330: $(a,b)\in T'_I\cap\R^2$,

1331: $b\ne0$, then

1332: $(a,b)\notin T''_I\cap\R^2$.  This gives a splitting of $T_I$ into two

1333: components in a neighborhood of $\pi_h^{-1}(-.08205,.08205)$.  Since

1334: $\pi_h$ has degree one on $T'_I\cap\R^2$ and $T''_I\cap\R^2$ these sets

1335: are given as the graphs of real analytic functions.

1336: \qed

1337: Let us set

1338: $$\kappa(t):=\max(\kappa^+_{01212}(t),\kappa^-_{12012}(t)).$$

1339: \proclaim Corollary 4.7.  $\{(a,b)\in\cD\cap\R^2:b\ne0,(**){\rm\ holds}\}

1340: =\{(a,b)\in\cD\cap\R^2: b\ne0,a\ge\kappa(b)\}$.

1341: 

1342: \give Proof.  We consider only the case $b>0$; the other case is similar.  For

1343: $I=01212$, set

1344: $T_I^\pm:=\{a=\kappa^\pm_{01212}(b)\}$.  Thus

1345: $T_I\cap\R^2=T_I^+\cup T^-_I$.    As was noted in the proof of Proposition~7.7,

1346: $T^-_I$ is the set of parameters for which one component of $W^u_I$ 

1347: is tangent to

1348: $W^s_{02}$, and the other component is disjoint from $W^s_{02}$. 

1349: $T^+_I$ is the

1350: set of parameters for which one component of $W^u_I$ is tangential to 

1351: $W^s_{02}$,

1352: and the other component intersects $W^s_{02}$ in two points. \qed

1353: 

1354: 

1355: Let us write

1356: $$\cE:=\{(a,b)\in\R^2: f_{a,b}{\rm\ has\ entropy }<\log2\}$$

1357: $$\cH:=\{(a,b)\in\R^2: f_{a,b}{\rm\ is\ a\ real\ horseshoe}\}$$

1358: 

1359: \proclaim Theorem 4.8. 

1360: $$\cH\cap\cD=\{(a,b)\in\cD\cap\R^2:a>\kappa(b), b\ne0\},$$

1361: $$\cE\cap\cD=\{(a,b)\in\cD\cap\R^2:a<\kappa(b),b\ne0\}.$$

1362: 

1363: \give Proof.  By Theorem 3.9 and Corollary 4.7, the set of parameters

1364: $(a,b)\in\cD\cap\R^2$ for which the entropy is $\log 2$ is exactly the set

1365: $\{a\ge\kappa(b)\}$.  On the other hand, if $a>\kappa(b)$, then by 

1366: Proposition 3.8

1367: there is no tangency.  Since $f$ has maximal entropy, it follows from 

1368: [BS1] that

1369: $f$ is hyperbolic.  Now $\cD\cap\R^2\cap\{a>\kappa(b)\}$ is a connected set of

1370: parameters for which $f_{a,b}$ is hyperbolic.  By Theorem 1.1, this 

1371: set contains

1372: parameters for which $f_{a,b}$ is a real horseshoe.  It follows, then, from the

1373: structural stability of hyperbolic maps that all of these maps are 

1374: horseshoes.\qed

1375: 

1376: \section 5. Generic Unfolding

1377: 

1378: In Theorem 5.2 we establish the ``generic unfolding'' statement in Theorem 2. 

1379: Let us fix $I=01212$ or $I=12012$.  In \S4 we saw that for $(a,b)\in\cD$, $b\ne0$, the set

1380: $W^u_I$ is disconnected and may be split into 

1381: $$W^u_I(a,b)=W^u_I(a,b)'\cup W^u_I(a,b)''.\eqno(5.1)$$

1382: Further we saw that if

1383: $(a_0,b_0)\in\cD\cap\R^2\cap\partial\cH$, then one of these components, say $W^u_I(a_0,b_0)'$,

1384: has a quadratic tangency with $W^s_{02}(a_0,b_0)$.  This splitting may be done for all

1385: $(a,b)\in\cD\cap\R^2$ in such a way that we obtain a continuous family

1386: $$\cD\cap \R^2\ni (a,b)\mapsto W^u_I(a,b)'.$$ 

1387: The horizontal projection $\pi_h(x,y)=y$,

1388: establishes a conformal equivalence

1389: $$\pi_h:W^s_{02}(a,b)\to\{|y|<e\}.$$  For $(a,b)\in\cD$, $b\ne0$, we define the function

1390: $$h(a,b)=\prod_{i\ne j}(\pi_h(p_i)-\pi_h(p_j))$$ 

1391: where the $p_i$ and $p_j$ in the product range over  the four points of

1392: intersection $W^s_{02}(a,b)\cap W^u_I(a,b)$.  Since

1393: $\pi_h|_{W^s_{02}(a,b)}$ is invertible, we see that

1394: $h(a,b)\ne0$ if and only if there are four distinct points of intersection.  Thus $h(a,b)\ne0$

1395: means that the multiplicities of all four intersections are 1, and thus all four

1396: intersections are transverse.  As in \S4 we may extend the definition of $h$ to the case

1397: $b=0$, and we see that

1398: $h$ is analytic in $\cD$.

1399: 

1400: \proclaim Theorem 5.1.  For $(a,b)\in\cD\cap\R^2\cap T_I$ with $b\ne0$, we have ${\partial

1401: h\over\partial a}\ne0$.

1402: 

1403: \give Proof. If $b=0$, then by the discussion in \S4, we see that

1404: $a\mapsto h(a,0)$ has a zero of order 2 at $a=2$, and $h(a,0)\ne0$ for

1405: $\{0<|a-2|<.237186\}$.  

1406: 

1407: By Theorem 4.5, none of the tangencies $T_I$ occur on the vertical boundary of $\cD$.  Thus

1408: $h\ne0$ there.  It follows that for each fixed value $|b_0|\le .08$, the function

1409: $$\{|a-2|<.237186\}\ni

1410: a\mapsto h(a,b_0)$$ is analytic and has exactly two zeros (counted with

1411: multiplicity).  One zero corresponds to a point $(a',b_0)\in T'_I$

1412: and one corresponds to $(a'',b_0)\in T_I''$.  We have seen that

1413: $T'_I\cap\cD\cap\R^2\cap\{b\ne0\}$ is disjoint from $T''_I\cap\cD\cap\R^2\cap\{b\ne0\}$. 

1414: Since the total multiplicity is 2, each of these zeros must be a simple zero.  In particular,

1415: we conclude that

1416: ${\partial h\over \partial z}(a,b)\ne0$ for $(a,b)\in T_I\cap\cD\cap\R^2\cap\{b\ne0\}$. \qed

1417: 

1418: Let us discuss this situation further.  We will consider a sequence of holomorphic

1419: coordinate changes

1420: $(x',y')=(x'(x,y),y'(x,y))$ which in addition depend holomorphically on the parameter

1421: $(a,b)$.  First, we may change coordinates so that

1422: $W^s_{02}(a,b)=\{x=0\}$ since

1423: $W^s_{02}(a,b)$ has degree one in

1424: $B_2$.  Now let

1425: us split

1426: $W^u_I(a,b)$ as in (5.1).  We will show that we may introduce coordinates such that we have

1427: $$W^s_{02}=\{x=0\}{\rm\ \ and\ \  }W^u_I(a,b)=\{x=c_0(a,b)+y^2\}.\eqno(5.2)$$  The {\it generic

1428: unfolding} condition is that ${\partial c_0(a,b)/ \partial a}\ne0$

1429: for $a=a_b$ (see [PT, page 35]).

1430: 

1431: Now let us fix $b_0\in(-.08,.08)$, $b_0\ne0$, and set $a_0=a_{b_0}$.  Thus we have

1432: $$W^u_I(a_0,b_0)'=\{x=\sum_{j=2}^\infty c_j(y-y_0)^j\},$$

1433: where $(0,y_0)$ is the point of tangential intersection, and so $c_0=c_1=0$.  The

1434: coefficient $c_2$ is nonzero because the intersection is quadratic (see [BS1]).  Without

1435: loss of generality we may assume that

1436: $y_0=0$.  Now for

1437: $(a,b)$ near

1438: $(a_0,b_0)$, we have

1439: $$W^u_I(a,b)'=\{x=c_0(a,b) + c_1(a,b)y+c_2(a,b)y^2+\dots\}.$$

1440: Now since $c_2(a,b)\ne0$ and $c_0(a_0,b_0)=c_1(a_0,b_0)=0$, we may solve $\tilde y=\tilde

1441: y(a,b)\sim -c_1/(2c_2)$ such that

1442: $${\partial x\over\partial y} =c_1(a,b)+2c_2(a,b)\tilde y+\dots=0.$$

1443: Replacing $y$ by $y-\tilde y$, we have

1444: $$W^u_I(a,b)'=\{x=\tilde c_0(a,b)+\tilde c_2(a,b)y^2+\dots\}.$$

1445: Finally, since $\tilde c_2\ne0$, we may change coordinates $y'=\sigma(a,b)y$ to obtain (5.2).

1446: 

1447: Now we consider the function $h(a,b)$ in the coordinates $(x,y)$.  We have $W^s_{02}(a,b)\cap

1448: W^u_I(a,b)'=\{(0,\pm\sqrt{-\tilde c_0(a,b)}\}$.  Since $W^u_I(a,b)'\cap

1449: W^u_I(a,b)''=\emptyset$, and $W^u_I(a,b)''$ has no tangency for

1450: $(a,b)$ near $(a_0,b_0)$, we have

1451: $$h(a,b)=-(\sqrt{-\tilde c_0(a,b)}+\sqrt{-\tilde c_0(a,b)})^2\alpha(a,b)=2\tilde

1452: c_0(a,b)\alpha(a,b)$$

1453: where $\alpha$ is a nonvanishing analytic function.  Since $\tilde c_0(a_0,b_0)=0$, we have

1454: $${\partial h\over\partial a}(a,b_0)={\partial \tilde c_0\over\partial

1455: a}(a,b_0)\cdot \alpha(a,b_0)$$ 

1456: for $a=a_0$.  By Theorem 5.1, then, ${\partial\tilde c_0}(a_0,b_0)/\partial a\ne0$.  Thus we

1457: have:

1458: 

1459: \proclaim Theorem 5.2.  $(a,b)\mapsto (W^s_{02}(a,b),W^u_I(a,b))$ is a generic unfolding of a

1460: tangency at the parameter value 

1461: $(a_0,b_0)$.

1462: 

1463: \bigskip

1464: 

1465: \centerline{\bf References}

1466: 

1467: %\item{[A]}  L. Ahlfors, {\sl Conformal Invariants}, McGraw-Hill, New York,

1468: %1973.

1469: %

1470: \item{[BLS]}  E. Bedford, M. Lyubich, and J. Smillie, Polynomial

1471: diffeomorphisms of

1472: $\C^2$.  IV: The measure of maximal entropy and laminar

1473: currents,  Invent.\ Math.\ 112, 77--125 (1993).

1474: 

1475: %\item{[BS1]}  E. Bedford and J. Smillie,  Polynomial diffeomorphisms of

1476: %$\C^2$: Currents, equilibrium measure, and hyperbolicity, Invent. Math. 87,

1477: %69--99 (1990).

1478: 

1479: %\item{[BS6]}  E. Bedford and J. Smillie,  Polynomial diffeomorphisms of

1480: %$\C^2$.  VI: Connectivity of $J$,  Annals of Math.\ 148, 695--735 (1998).

1481: %

1482: %\item{[BS7]}  E. Bedford and J. Smillie,  Polynomial diffeomorphisms of

1483: %$\C^2$.  VII: Hyperbolicity and external rays,  Ann.\ Sci.\ Ecole Norm.\ Sup.,

1484: %32, 455--497 (1999).

1485: 

1486: \item{[BS8]}  E. Bedford and J. Smillie,  Polynomial diffeomorphisms of

1487: $\C^2$.  VIII: Quasi-expansion.  American J. of Math., 124, 221--271, (2002).

1488: 

1489: \item{[BS1]} E. Bedford and J. Smillie, Real polynomial diffeomorphisms with

1490: maximal entropy: Tangencies.  Annals of Math., to appear.

1491: 

1492: \item{[BS3]} E. Bedford and J. Smillie, A new approach to the analysis of analytic

1493: diffeomorphisms, in preparation.

1494: 

1495: %\item{[BiMi]} E. Bierstone and P. Milman, Semianalytic and 

1496: %subanalytic sets, IHES

1497: %Publications Math\'ematiques, 67, 5--42, 1988.

1498: 

1499: %\item{[CJY]}  L. Carleson, P. Jones, and J.-C. Yoccoz, Julia and John,

1500: %Bol.\ Soc.\ Bras.\ Mat., Vol.\ 25, 1--30 (1994).

1501: %

1502: %\item{[Ca]}  J. Carette, Liens entre la g\'eometrie et la dynamique des

1503: %ensembles

1504: %de Julia, Th\`ese, U. de Paris-Sud, Orsay, 1997.

1505: 

1506: %\item{[Ch]} E. Chirka, {\sl Analytic Sets},  Kluwer Academic Publishers, 1989.

1507: 

1508: \item{[Co]} P. Collins, Dynamics forced by surface trellises. {\sl Geometry and topology

1509: in dynamics (Winston-Salem, NC, 1998/San Antonio, TX, 1999)}, 65--86, Contemp. Math.,

1510: 246, Amer. Math. Soc., Providence, RI, 1999. 

1511: 

1512: \item{[DN]} R. Devaney and Z. Nitecki, Shift automorphisms in the 

1513: H\'enon family,

1514: Comm.\ Math.\ Phys.\ 67, 137--48, 1979.

1515: 

1516: %\item{[F]}  W.H.J. Fuchs, {\sl Topics in the Theory of Functions of 

1517: %One Complex

1518: %Variable}, van Nostrand, 1967.

1519: 

1520: \item{[FG1]}  J-E Forn\ae ss and E. Gavosto,  Existence of generic homoclinic tangencies for

1521: H\'enon mappings, J.\ Geom.\ Anal., 2:5 (1992), 429--444.

1522: 

1523: 

1524: \item{[FG2]}  J-E Forn\ae ss and E. Gavosto,    Tangencies for real and complex H\'enon maps:

1525: An analytic method, Experimental Math., Vol. 8 (1999), No. 3, 253--260.

1526: 

1527: 

1528: %\item{[FS]} J-E Forn\ae ss and N. Sibony,  Complex H\'enon mappings in $ C\sp 2$ and

1529: %Fatou-Bieberbach domains. Duke Math. J. 65 (1992), no. 2, 345--380.

1530: 

1531: %\item{[FM]} S. Friedland and J. Milnor, Dynamical properties of plane polynomial

1532: %automorphisms.  Ergodic Theory Dynam. Systems 9 (1989), no. 1, 67--99.

1533: 

1534: \item{[HO]} J.H.\ Hubbard and R.\ Oberste-Vorth, H\'enon mappings in the

1535: complex domain II: Projective and inductive limits of polynomials, in: {\sl

1536: Real and Complex Dynamical Systems}, B. Branner and P. Hjorth, eds. 89--132

1537: (1995).

1538: 

1539: \item{[KKY]} ÊKan, Ittai; Kocak, Huseyin; Yorke, James A. Antimonotonicity: concurrent

1540: creation and annihilation of periodic orbits. Ann. of Math. (2) 136 (1992), no. 2, 219--252. 

1541: 

1542: \item{[MNTU]} S. Morosawa, Y. Nishimura, M. Taniguchi, and T. Ueda, {\sl

1543: Holomorphic Dynamics}, Cambridge U. Press, 2000.

1544: 

1545: \item{[O]} R. Oberste-Vorth, Complex horseshoes and the dynamics of 

1546: mappings of two

1547: complex variables, Ph.D dissertation, Cornell Univ., Ithaca, NY, 1987.

1548: 

1549: \item{[MT]} J. Milnor and W. Thurston, On iterated maps of the interval. {\sl Dynamical

1550: systems (College Park, MD, 1986--87)}, 465--563, Lecture Notes in Math., 1342, Springer,

1551: Berlin, 1988.

1552: 

1553: \item{[PT]}  J. Palis and F. Takens, {\sl Hyperbolicity and sensitive chaotic dynamics

1554: at homoclinic bifurcations. Fractal dimensions and infinitely many attractors.}

1555: Cambridge Studies in Advanced Mathematics, 35. Cambridge University Press, Cambridge,

1556: 1993.

1557: 

1558: \bigskip

1559: \rightline{Eric Bedford}

1560: \rightline{Indiana University}

1561: \rightline{Bloomington, IN 47405}

1562: \bigskip

1563: \rightline{John Smillie}

1564: \rightline{Cornell University}

1565: \rightline{Ithaca, NY 14853}

1566: 

1567: \bye

1568: 

1569: 

1570: 

1571: 

1572: 

1573: 

1574: