1: \documentstyle[12pt,leqno,amssymb,graphics]{amsart}

2: 

3: 

4: %Immersions of non-orientable surfaces

5: 

6: \setlength{\topmargin}{-10mm}

7: \setlength{\oddsidemargin}{-2mm}

8: \setlength{\evensidemargin}{-2mm}

9: \setlength{\textwidth}{6.5in}

10: \setlength{\textheight}{9.2in}

11: \renewcommand{\baselinestretch}{1.5}

12: 

13: \newtheorem{thm}{Theorem}[section]

14: \newtheorem{lemma}[thm]{Lemma}

15: \newtheorem{prop}[thm]{Proposition}

16: \newtheorem{cor}[thm]{Corollary}

17: 

18: \theoremstyle{definition}

19: \newtheorem{dfn}[thm]{Definition}

20: \theoremstyle{remark}

21: \newtheorem{remark}[thm]{Remark}

22: 

23: 

24: 

25: \begin{document}

26: 

27: \newcommand{\ct}{\cite}

28: \newcommand{\pr}{\protect\ref}

29: \newcommand{\su}{\subseteq}

30: \newcommand{\pa}{{\partial}}

31: \newcommand{\im}{{Imm(F,\E)}}

32: \newcommand{\lm}{{\lambda}}

33: \newcommand{\tc}{{\mathfrak{T}}}

34: \newcommand{\Q}{{\Bbb Q}}

35: \newcommand{\hf}{{1 \over 2}}

36: \newcommand{\CC}{{\mathcal C}}

37: 

38: \newcommand{\R}{{\Bbb R}}

39: \newcommand{\Z}{{\Bbb Z}}

40: \newcommand{\E}{{{\Bbb R}^3}}

41: \newcommand{\C}{{{\Bbb Z}/2}}

42: \newcommand{\B}{{\Bbb B}}

43: 

44: \newcommand{\FF}{{\mathcal F}}

45: 

46: \newcommand{\A}{{\mathcal{A}}}

47: \newcommand{\Ai}{{{\mathrm{Arf}}(g^i)}}

48: \newcommand{\Ar}{{\mathrm{Arf}}}

49: \newcommand{\r}{{\mathrm{rank}}}

50: \newcommand{\I}{{\mathrm{Id}}}

51: \newcommand{\spp}{{\mathrm{supp}}}

52: 

53: \newcommand{\F}{{\mathbf{F}}}

54: \newcommand{\G}{{\Bbb G}}

55: \newcommand{\cd}{{\mathrm{codim}}}

56: \newcommand{\p}{{\psi}}

57: \newcommand{\hp}{{\Psi}}

58: \newcommand{\hhp}{\widehat{\Psi}}

59: \newcommand{\pp}{{\widehat{\psi}}}

60: 

61: \newcommand{\4}{{\mathcal{H}}}

62: \newcommand{\N}{{\mathcal{N}}}

63: \newcommand{\ng}{{{\mathcal{N}}_g}}

64: 

65: \newcommand{\hM}{{\widehat{\mathcal{M}}}}

66: \newcommand{\m}{{\mathcal{M}}_g}

67: \newcommand{\hm}{{\hM_g}}

68: \newcommand{\hmi}{{\hM_{g^i}}}

69: \newcommand{\mi}{{\M_{g^i}}}

70: 

71: \newcommand{\f}{{\overline{f}}}

72: 

73: \newcommand{\hc}{{H_1(F;\C)}}

74: \newcommand{\ohg}{{O(\hc,g)}}

75: 

76: \newcommand{\ohi}{{O(\hc,g^i)}}

77: \newcommand{\Tc}{{{\mathcal{T}}_c}}

78: \newcommand{\SP}{{{\mathcal{S}}_P}}

79: 

80: \newcommand{\Y}{{\mathcal{Y}}}

81: \newcommand{\T}{{\mathcal{T}}}

82: \newcommand{\U}{{\mathcal{U}}}

83: \newcommand{\tF}{{\widetilde{F}}}

84: 

85: 

86: \newcommand{\one}{{\widetilde{F_1}}}

87: \newcommand{\two}{{\widetilde{F_2}}}

88: \newcommand{\tfk}{{\widetilde{F_k}}}

89: \newcommand{\th}{{\widetilde{h}}}

90: \newcommand{\ep}{{\epsilon}}

91: \newcommand{\wb}{{W^\bot}}

92: \newcommand{\w}{{\bigcup_i w_i}}

93: \newcommand{\eep}{{\widehat{\epsilon}}}

94: 

95: 

96: \newcounter{numb}

97: 

98: 

99: \title{Immersions of Non-orientable Surfaces}

100: \author{Tahl Nowik}

101: \address{Department of Mathematics, Bar-Ilan University, 

102: Ramat-Gan 52900, Israel.}

103: \email{tahl@@math.biu.ac.il}

104: \date{October 1, 2003}

105: \thanks{Partially supported by the Minerva Foundation}

106: 

107: 

108: \begin{abstract}

109: Let $F$ be a closed non-orientable surface.

110: We classify all finite order invariants of immersions of $F$

111: into $\E$, with values in any Abelian group. We show they are all

112: functions of the universal order 1 invariant that we construct 

113: as $T \oplus P \oplus Q$ where $T$ is a $\Z$ valued invariant reflecting the number of triple

114: points of the immersion, and $P,Q$ are $\C$ valued invariants characterized by the property

115: that for any regularly homotopic 

116: immersions $i,j:F\to \E$, $P(i)-P(j) \in \C$ 

117: (respectively $Q(i)-Q(j) \in \C$)

118: is the number mod 2 of tangency points (respectively quadruple points)

119: occurring in any generic regular homotopy between $i$ and $j$. 

120: 

121: For immersion $i:F\to\E$ and diffeomorphism $h:F\to F$ such that $i$ and $i \circ h$

122: are regularly homotopic we show:

123: $$P(i\circ h)-P(i)=Q(i\circ h)-Q(i) =\bigg(\r(h_*-\I)+\ep(\det h_{**})\bigg)\bmod{2}$$

124: where $h_*$ is the map induced by $h$ on $\hc$, $h_{**}$ is

125: the map induced by $h$ on $H_1(F;\Q)$, and for $0 \neq q \in \Q$, 

126: $\ep(q) \in\C$ is 0 or 1

127: according to whether $q$ is positive or negative, respectively.

128: \end{abstract}

129: 

130: 

131: 

132: 

133: \maketitle

134: 

135: \section{Introduction}

136: 

137: Finite order invariants of immersions of a closed surface into $\E$ have been defined in \ct{o},

138: and the case of orientable surfaces has been studied in \ct{a},\ct{o},\ct{h},\ct{f}.

139: In the present work we establish all analogous results for the non-orientable case.

140: We classify all finite order invariants. For each $n$ we construct a

141: universal order $n$ invariant, and for any $n>1$ it

142: is constructed as an explicit function of the universal order 1 invariant.

143: The universal order 1 invariant is given by 

144: $T \oplus P \oplus Q$ where $T$ is a $\Z$ 

145: valued invariant reflecting the number of triple

146: points of the immersion, and $P,Q$ are $\C$ valued invariants characterized by the property

147: that for any regularly homotopic 

148: immersions $i,j:F\to \E$, $P(i)-P(j) \in \C$ 

149: (respectively $Q(i)-Q(j) \in \C$)

150: is the number mod 2 of tangency points (respectively quadruple points)

151: occurring in any generic regular homotopy between $i$ and $j$. 

152: We give an explicit formula for $P(i \circ h)-P(i)$ and $Q(i \circ h)-Q(i)$ 

153: for any diffeomorphism $h:F \to F$ such that $i$ and $i \circ h$ are regularly homotopic.

154: This requires the study of the quadratic form induced on $\hc$ by an immersion $i:F\to\E$,

155: where in the case of non-orientable surfaces we name it an $\4$-form, rather than quadratic form,

156: to emphasize its distinct character.  In order to obtain the stated formulae for $P,Q$, we study the 

157: group of diffeomorphisms of $F$ which preserves the induced $\4$-form.

158: 

159: The extension of the classification of finite order invariants from the orientable to the 

160: non-orientable case (Section \pr{fin}) is rather straight forward, whereas

161: the study of the group of diffeomorphisms preserving an $\4$-form (Section \pr{og}), and the formula 

162: for the invariants $P,Q$ (Section \pr{fmq}), will require substantially new analysis.

163: 

164: 

165: \section{Finite order invariants}\label{fin}

166: 

167: Let $F$ be a closed non-orientable surface. $\im$ denotes the 

168: space of all immersions of $F$ into $\E$, with the $C^1$ topology.

169: A CE point of an immersion $i:F\to \E$ is a point of self intersection

170: of $i$ for which the local stratum in $\im$ corresponding to the 

171: self intersection, has codimension one. 

172: For $F$ non-orientable we distinguish four types of CEs which we name

173: $E, H, T, Q$. 

174: In the notation of \ct{hk} they are respectively

175: $A_0^2|A_1^+$, $A_0^2|A_1^-$, $A_0^3|A_1$, $A_0^4$.

176: The four types may be demonstrated by the following local models, where letting $\lm$ vary, 

177: we obtain a 1-parameter family of immersions which is transverse to the 

178: given codim 1 stratum, intersecting it at $\lm=0$.

179: 

180: $E$: \ \ $z=0$, \ \ $z=x^2+y^2+\lm$. See Figure \pr{fet}, ignoring the vertical plane.

181: 

182: $H$: \ \ $z=0$, \ \ $z=x^2-y^2+\lm$. See Figure \pr{fht}, ignoring the vertical plane.

183: 

184: 

185: $T$: \ \ $z=0$, \ \ $y=0$, \ \ $z=y+x^2+\lm$. See Figure \pr{ftq}, ignoring the vertical 

186: plane $x=0$.

187: 

188: $Q$: \ \ $z=0$, \ \ $y=0$, \ \ $x=0$, \ \ $z=x+y+\lm$. This is simply four planes passing 

189: through one point, any three of which are in general position.

190: 

191: A choice of one of the two sides of the local codim 1 stratum at a given point of the 

192: stratum, is represented by the choice of $\lm<0$ or $\lm>0$ in the formulae above. We will 

193: refer to such a choice as a 

194: \emph{co-orientation} for the configuration of the self intersection. 

195: For types $E$ and $T$, the configuration of the self intersection at the two sides 

196: of the stratum is distinct, namely, for $\lm<0$ there is an additional 

197: 2-sphere in the image of the immersion, and we permanently

198: choose this side ($\lm<0$) as our positive side for the co-orientation. 

199: For types $H$ and $Q$, the configuration of the self intersection on the two 

200: sides of the strata are 

201: indistinguishable, and in fact we will see that the strata in this case are globally one-sided,

202: and so no coherent choice of co-orientation is possible.

203: 

204: We fix a closed non-orientable surface $F$ and a regular homotopy class $\A$ of 

205: immersions of $F$ into $\E$.

206: We denote by $I_n\su \A$ ($n\geq 0$) the space of all immersions in $\A$ which have precisely 

207: $n$ CE points (the self intersection being elsewhere stable).

208: In particular, $I_0$ is the space of all stable immersions in $\A$. 

209: 

210: For an immersion $i:F\to\E$ having a CE located at $p\in\E$, we will now define

211: the degree $d_p(i)$. This will differ from the definition

212: for orientable surfaces, given in \ct{o}, in two ways. 

213: First, for nonorientable surface, 

214: the degree of a map $F \to \E - \{ p\}$ is defined only mod 2. Second,

215: we do not have an orientation which determines into what side of $p$ we must 

216: push each sheet participating in the CE,

217: in order for the degree to be computed. 

218: So for non-orientable surface we define $d_p(i) \in \{ + , - \}$, as follows: 

219: For CE of type $E,T$ we move the immersion to the positive side determined by its

220: permanent co-orientation, obtaining an immersion $i'$ where a new 

221: 2-sphere appears in the image. We define $d_p(i)\in \{ + , - \}$ to be the degree of the 

222: map $i':F \to \E- \{ p' \}$ where $p'$ is any point in the open 3-cell bounded by the new 2-sphere,

223: and where $+ =$ even and $- =$ odd. 

224: For CEs of type $H,Q$,

225: let $V$ denote the region between the sheets of the surface which appears once we move 

226: away from the stratum (i.e. once $\lm\neq 0$). 

227: For type $Q$ this is a 3-simplex defined by our four sheets. For type $H$ this region is not 

228: bounded by the local configuration,

229: but may still be defined, e.g. for $\lm>0$ in the formula for $H$, 

230: $V$ will be a region consisting of points close to the origin and 

231: satisfying $0\leq z\leq x^2-y^2+\lm$. We define 

232: $d_p(i)\in \{ + , - \}$ to be the degree of the 

233: map $i':F \to \E- \{p' \}$ where $p'$ is any point in $V$, and notice that since there is an even

234: number of sheets involved in a CE of type $H$ or $Q$, the definition is independent of the side of the 

235: stratum we choose to move into, (i.e. whether we choose $\lm>0$ or $\lm<0$).

236: 

237: We define

238: $C_p(i)$ to be the expression $R_e$ where $R$

239: is the symbol describing the configuration of the CE of $i$ at $p$ ($E$, $H$, $T$, or $Q$) and $e=d_p(i)$. 

240: We define $\CC_n$ to be the set of all

241: \emph{un-ordered} $n$-tuples of expressions $R_e$ with $R$ one of the four symbols

242: and $e\in\{+,-\}$.

243: So $\CC_n$ is the set of un-ordered $n$-tuples of elements of 

244: $\CC_1 = \{E_+ , E_- , H_+ , H_- , T_+ , T_- , Q_+ , Q_- \}$.

245: We define $C:I_n \to \CC_n$ by

246: $C(i)=[C_{p_1}(i),\dots, C_{p_n}(i)] \in \CC_n$ where $p_1,\dots,p_n$ are the $n$

247: CE points of $i$. The map 

248: $C:I_n \to \CC_n$ is easily seen to be surjective.

249: 

250: A regular homotopy between two immersions in $I_n$ is called an AB equivalence if 

251: it is alternatingly of type A and B, where $J_t:F\to\E$ ($0\leq t \leq 1$)

252: is of type A if it is of the form

253: $J_t = U_t \circ i \circ V_t$ where 

254: $i:F\to \E$ is an immersion and

255: $U_t:\E\to\E$, 

256: $V_t:F\to F$ are isotopies, and $J_t:F\to\E$ ($0\leq t \leq 1$)

257: is of type B if $J_0 \in I_n$ and there are  

258: little balls $B_1,\dots,B_n\su\E$ centered at the $n$ CE points of $J_0$

259: such that $J_t$ fixes $U=(J_0)^{-1}(\bigcup_k B_k)$ 

260: and moves $F-U$ within $\E - \bigcup_k B_k$.

261: 

262: \begin{prop}\label{AB}

263: Let  $i,j\in I_n$, then $i$ and $j$ are AB equivalent iff $C(i)=C(j)$.

264: \end{prop}

265: 

266: \begin{pf}

267: The proof proceeds as in \ct{o} Proposition 3.4 except for one step that must be added 

268: in the present case, of non-orientable surfaces. 

269: In the final stage of the proof of \ct{o} Proposition 3.4, 

270: we have that $i'''|_D$ and $j|_D$ 

271: are homotopic in $\E-\bigcup_k B_k$ relative $\pa D$, (since 

272: $\widehat{d}_{p_k}(i''')=\widehat{d}_{p_k}(j)$ for all $k=1,\dots,n$,

273: where $\widehat{d}_p$ denotes the $\Z$ valued degrees defined in \ct{o} for the orientable case).

274: In the present case this will not necessarily be true. If $f:S^2 \to \E-\bigcup_k B_k$ 

275: is the map determined by the pair of maps $i'''|_D,j|_D$, 

276: then instead of $f$ being null-homotopic 

277: as in the orientable case, it may only be 

278: of even degree with respect to each $p_1,\dots,p_n$.

279: This can be remedied as follows: Let $h$ be one of the 1-handles having both

280: ends attached to $D_1$, so

281: $D_1 \cup h$ is a Mobius band and the 2-handle $D$ is

282: glued to it twice in the same direction.

283: A homotopy of $h$ which traces a sphere in $\E$ enclosing just one $p_i$ and not the others, will

284: change the degree of $f$ with respect to $p_i$ by 2, and leave the degree with respect

285: to all other $p_k$ unchanged. We realize such homotopies by regular homotopies

286: until the new map $f$ has degree 0 with respect to each $p_1,\dots,p_n$. 

287: 

288: 

289: \end{pf}

290: 

291: Given an immersion $i\in I_n$, a \emph{temporary co-orientation} for $i$ is a choice 

292: of co-orientation at each of the $n$ CE points $p_1, \dots , p_n$ of $i$.

293: Given a temporary co-orientation $\tc$ for $i$ and a subset $A\su \{p_1,\dots,p_n\}$,

294: we define $i_{\tc,A} \in I_0$ to be the immersion obtained from $i$ by resolving all CEs

295: of $i$ at points of $A$ into the 

296: positive side with respect to $\tc$,

297: and all CEs not in $A$ into the negative side.

298: Now let $\G$ be any Abelian group and let $f:I_0\to\G$ be an invariant, i.e. a function which is 

299: constant on each connected component of $I_0$.

300: Given $i\in I_n$ and a temporary co-orientation $\tc$ for $i$,

301: $f^\tc(i)$ is defined as follows:

302: $$f^\tc(i)=\sum_{ A \su \{p_1,\dots,p_n\} } (-1)^{n-|A|} f(i_{\tc,A})$$

303: where $|A|$ is the number of elements in $A$.

304: The statement $f^\tc(i)=0$ is independent of the temporary co-orientation $\tc$

305: so we

306: simply write $f(i)=0$.

307: An invariant $f:I_0\to\G$ is called \emph{of finite order} if 

308: there is an $n$ such that $f(i)=0$ for all $i\in I_{n+1}$.

309: The minimal such $n$ is called the \emph{order} of $f$.

310: The group of all invariants on $I_0$ of order at most $n$ is denoted 

311: $V_n = V_n(\G)$.

312: 

313: Let $f \in V_n$. If $i\in I_n$ has at least one CE

314: of type $H$ or $Q$ and $\tc$ is a temporary co-orientation for $i$,

315: then $2f^\tc(i)=0$, the proof being the same as in \ct{o} Proposition 3.5.

316: So in this case $f^\tc(i)$ is independent of $\tc$. 

317: We use this fact to extend any $f\in V_n$ to $I_n$

318: by setting for any $i\in I_n$, $f(i) = f^\tc(i)$,

319: where if $i$ includes at least one CE of type $H$ or $Q$ then

320: $\tc$ is arbitrary, and if all CEs of $i$ are of type $E$ and $T$

321: then the permanent co-orientation is used for all CEs of $i$.

322: We will always assume without mention that any $f \in V_n$ is extended to $I_n$ 

323: in this way.

324: (If $f\in V_n$ then we are not extending $f$ to $I_k$ for

325: $0<k<n$). 

326: 

327: We remark at this point that the same argument as explained in \ct{o} Remark 3.7, showing that

328: for orientable surfaces, the strata corresponding to configurations $H^1$ and $Q^2$

329: may not be globally co-oriented, will show that

330: the same is true for non-orientable surfaces for configurations $H$ and $Q$.

331: 

332: For $f\in V_n$ and $i,j\in I_n$, 

333: if $C(i)=C(j)$ then $f(i)=f(j)$, the proof being the same as in \ct{o} Proposition 3.8,

334: so any $f\in V_n$ induces a well defined

335: function $u(f):\CC_n\to\G$. 

336: The map $f\mapsto u(f)$ induces an injection $u:V_n / V_{n-1} \to \CC_n^*$ 

337: where $\CC_n^*$ is the group of all functions from $\CC_n$ to $\G$.

338: We will find the image of $u$ for all $n$, by this we classify all finite

339: order invariants.

340: 

341: Let $i\in \A$ be an immersion with a self intersection of local 

342: codim 2 at $p$ and 

343: $n-1$ additional self-intersections of local codim 1 (i.e. CEs) at $p_1,\dots,p_{n-1}$.

344: We look at a 2-parameter family of immersions which moves $F$ only in a neighborhood 

345: of $p$, such that the immersion $i$ corresponds to 

346: parameters $(0,0)$ and such that this 2-parameter family is transverse to the local 

347: codim 2 stratum at $i$.

348: In this 2-parameter family of immersions we look at a loop which 

349: encircles the point of 

350: intersection with the codim 2 strata,

351: i.e. a circle around the origin in the parameter plane.

352: This circle crosses the local codim 1 strata some $r$ times.

353: Between each two intersections we have an immersion in $I_{n-1}$ 

354: with the same $n-1$ CEs, at $p_1,\dots,p_{n-1}$.

355: At each intersection with the local codim 1 strata, an $n$th CE is added, 

356: obtaining an immersion in $I_n$. Let $i_1,\dots,i_r$ be the $r$ immersions in $I_n$ so 

357: obtained and let $\ep_k$, $k=1,\dots,r$ be $1$ or $-1$ according to whether we are passing 

358: the $n$th CE of $i_k$ in the direction of its 

359: permanent co-orientation, if it has one, and if the CE is of type $H$ or $Q$ then $\ep_k$ is 

360: arbitrarily chosen. 

361: For $f \in V_n$, $f(i_k)$ is defined and it is easy to see that 

362: $\sum_{k=1}^r \ep_k f(i_k) = 0$. 

363: Looking at $u:V_n / V_{n-1} \to \CC_n^*$  we thus obtain relations that must be satisfied by 

364: a function in $\CC_n^*$ in order for it to lie in the image of $u$.

365: We will now find all relations on $\CC_n^*$ obtained in this way.

366: The relations on a $g\in\CC_n^*$ will be written as relations on the symbols $R_e$, e.g.

367: $0 = T_e - T_{-e}$ will stand for the set of all relations of the form

368: $0 = g([T_e , {R_2}_{e_2},\dots,{R_n}_{e_n}]) - 

369: g([T_{-e}, {R_2}_{e_2},\dots,{R_n}_{e_n}])$ with 

370: arbitrary ${R_2}_{e_2},\dots,{R_n}_{e_n}$. 

371: We already know that the following two relations hold:

372: $0=2H_e$, $0=2Q_e$ (for both $e$).

373: 

374: 

375: We look at local 2-parameter 

376: families of immersions which are transverse to 

377: the various local codim 2 strata.

378: These may be divided into six types which we name after the types of CEs 

379: appearing in the 2-parameter family:

380: $EH$, $TT$, $ET$, $HT$, $TQ$, $QQ$. In the notation of \ct{hk} they are 

381: respectively: $A_0^2|A_2$, $A_0^3|A_2$, $(A_0^2|A_1^+)(A_0)$, 

382: $(A_0^2|A_1^-)(A_0)$, $(A_0^3|A_1)(A_0)$, $A_0^5$.

383: Formula and sketch for local model for such strata,

384: the bifurcation diagrams, and the relations obtained, are as follows.

385: A sign $\pm$ appears wherever the element is known to be of order 2 (which is

386: wherever there is no co-orientation).

387: 

388: \begin{figure}[h]

389: \scalebox{0.6}{\includegraphics{NEH.eps}}

390: \caption{$EH$ configuration}\label{feh}

391: \end{figure}

392: 

393: $EH$: \ \ $z=0$, \ \ $z=y^2 + x^3+\lm_1 x + \lm_2$.

394: \begin{equation}\label{eeh}

395: 0 = E_e \pm H_e

396: \end{equation}

397: 

398: 

399: 

400: 

401: \begin{figure}[h]

402: \scalebox{0.6}{\includegraphics{NTT.eps}}

403: \caption{$TT$ configuration}\label{ftt}

404: \end{figure}

405: 

406: 

407: $TT$: \ \ $z=0$, \ \ $y=0$, \ \ $z=y+x^3+\lm_1 x  + \lm_2$.

408: \begin{equation}\label{ett}

409: 0 = T_e - T_{-e}

410: \end{equation}

411: 

412: 

413: 

414: \begin{figure}[h]

415: \scalebox{0.6}{\includegraphics{NET.eps}}

416: \caption{$ET$ configuration}\label{fet}

417: \end{figure}

418: 

419: $ET$: \ \ $z=0$, \ \ $x=0$, \ \ $z=(x-\lm_1)^2 + y^2 + \lm_2$.

420: \begin{equation}\label{eet}

421: 0 = T_{-e} - T_e - E_{-e} + E_e

422: \end{equation}

423: 

424: 

425: 

426: 

427: 

428: 

429: \begin{figure}[h]

430: \scalebox{0.6}{\includegraphics{NHT.eps}}

431: \caption{$HT$ configuration}\label{fht}

432: \end{figure}

433: 

434: 

435: $HT$: \ \ $z=0$, \ \ $x=0$, \ \ $z=(x-\lm_1)^2 - y^2 + \lm_2$.

436: \begin{equation}\label{eht}

437: 0 = -T_e + T_{-e} \pm H_{-e} \pm H_e

438: \end{equation}

439: 

440: 

441: 

442: 

443: \begin{figure}[h]

444: \scalebox{0.6}{\includegraphics{NTQ.eps}}

445: \caption{$TQ$ configuration}\label{ftq}

446: \end{figure}

447: 

448: 

449: $TQ$: \ \ $z=0$, \ \ $y=0$, \ \ $x=0$, \ \ $z=y+(x-\lm_1)^2 + \lm_2$.

450: \begin{equation}\label{etq}

451: 0 =  \pm Q_e \pm Q_{-e} - T_e + T_{-e}

452: \end{equation}

453: 

454: 

455: 

456: \begin{figure}[h]

457: \scalebox{0.6}{\includegraphics{NQQ.eps}}

458: \caption{$QQ$ configuration}\label{fqq}

459: \end{figure}

460: 

461: 

462: $QQ$: \ \ Five planes meeting at a point.

463: \begin{equation}\label{eqq}

464: 0 = 5(\pm Q_e) + 5(\pm Q_{-e})

465: \end{equation}

466: 

467: 

468: For the first five types,

469: the bifurcation diagram and degrees are obtained from the sketch and formula in a straight forward 

470: manner. 

471: The diagram for $QQ$ is obtained as explained in \ct{o}.

472: The degrees appearing in the present $QQ$ diagram require explanation, but we may avoid it since 

473: from the first five relations it already follows that $Q_+=Q_-$ and so since $2Q_e=0$,

474: the last relation does not add to the first five relations, whatever the degrees may be. 

475: Letting $\B = \{ x \in \G : 2x=0 \}$ the above relations may be summed up as follows:

476: 

477: \begin{itemize}

478: \item $T_+ = T_-$

479: \item $E_+ = E_- = H_+ = H_- \in \B$

480: \item $Q_+ = Q_- \in \B$.

481: \end{itemize}

482: We denote by $\Delta_n = \Delta_n(\G)$ the subgroup of $\CC_n^*$ of all functions 

483: satisfying these relations, so the image of $u:V_n \to \CC_n^*$ is contained in 

484: $\Delta_n$.

485: 

486: We define the universal Abelian group $\G_U$ by the Abelian group presentation

487: $\G_U = \left< t , p , q \ | \ 2p = 2q = 0 \right>$. 

488: We define the universal element $g_1^U\in\Delta_1(\G_U)$ by $g_1^U(T_e) = t$,

489: $g_1^U(E_e)=g_1^U(H_e)=p$, $g_1^U(Q_e)=q$, 

490: so indeed $g_1^U\in\Delta_1(\G_U)$. 

491: Then for arbitrary Abelian group $\G$ we have 

492: $\Delta_1(\G) \cong Hom(\G_U , \G)$ where the isomorphism 

493: maps a homomorphism $\phi\in Hom(\G_U , \G)$ 

494: to the function $\phi \circ g_1^U \in \Delta_1(\G)$.

495: We will show that there is an order 1 

496: invariant $f_1^U:I_0\to\G_U$ with $u(f_1^U)=g_1^U$. It will follow that for any 

497: group $\G$, $u:V_1/V_0\to \Delta_1(\G)$ is surjective, since if $g\in\Delta_1(\G)$ 

498: and $g=\phi \circ g_1^U$ where $\phi\in Hom(\G_U , \G)$ 

499: then $u(\phi\circ f_1^U) = g$.

500: 

501: We define $f_1^U:I_0\to\G_U$ as follows:

502: Choose a base immersion $i_0 \in I_0$ once and for all.

503: Then for each $i \in I_0$ take a generic regular homotopy

504: $H_t:F \to \E$ ($0 \leq t \leq 1$)

505: from $i_0$ to $i$ and define $f_1^U(i)=n_1 t + n_2 p + n_3 q \in \G_U$ where

506: $n_1 \in \Z$ is the number of CEs of type $T$ occurring along $H_t$, each counted as $\pm 1$ according

507: to the permanent coorientation of $T$, and $n_2 \in \C$ (respectively $n_3 \in \C$) is the number mod 2 of

508: tangencies (respectively quadruple points), occurring in $H_t$.

509: If $f_1^U$ is well defined, i.e. independent of the choice of $H_t$, then clearly $u(f_1^U)=g_1^U$.

510: So it remains to show that $f_1^U$ is independent of $H_t$, which is equivalent to showing that

511: the value is 0 for any \emph{closed} regular homotopy, i.e. one that begins and ends with the 

512: same immersion.

513: 

514: Now $\pi_1(\A) \cong \C \oplus \C$. 

515: Though this is different from the fundamental group in the orientable case ($\C \oplus \Z$),

516: the two generators of $\pi_1(\A)$ are obtained in the same way, namely, the first generator is given by

517: one full rigid rotation of the surface in $\E$, and the second generator is obtained by

518: only moving a disc in $F$ using the infinite cyclic generator in $\pi_1(Imm(S^2,\E))$

519: as shown in \ct{q}. So the proof that $f_1^U$ is well defined may proceed exactly as in the orientable case,

520: as appears in \ct{o}, once we know that the same invariant is well defined for $S^2$.

521: But for $S^2$, \ct{o} gives a full classification, and if hats will denote the corresponding constructions

522: for orientable surfaces, then this particular invariant for $S^2$ is obtained from the following

523: $g \in \widehat{\Delta}_1(\G_U)$ (note we are taking orientable $\Delta_1$ of non-orientable $\G_U$):

524: $g(T^a_m)=t$, $g(E^a_m)=g(H^a_m)=p$, $g(Q^a_m)=q$, for all $a,m$. (See \ct{o} for definition of the symbols

525: $R^a_m$).

526: This completes the proof that $f_1^U$ is well defined.

527: The invariant $f_1^U$ is clearly a \emph{universal} order 1 invariant, 

528: where universal order $n$ invariant is defined as follows:

529: 

530: \begin{dfn}\label{uni}

531: A pair $(\G,f)$ where $\G$ is an Abelian group and 

532: $f:I_0 \to \G$ is an order $n$ invariant,

533: will be called a \emph{universal order $n$ invariant}

534: if for any Abelian group $\G'$ and any order $n$ invariant 

535: $f':I_0 \to \G'$ there exists a unique homomorphism $\varphi:\G \to \G'$ 

536: such that $f' - \varphi \circ f$ is an invariant of order 

537: at most $n-1$. 

538: \end{dfn}

539: 

540: 

541: We have $\G_U  = \Z t \oplus (\C)p \oplus (\C)q$ and let $T,P,Q$

542: be the projection of $f_1^U$ into these three factors of $G_U$. 

543: We may write an explicit formula for $T$, namely, $T(i) = {N-c \over 2}$ 

544: where $N$ is the number of triple points in $i$

545: and $c=0$ if $\chi(F)$ is even, $c=1$ if $\chi(F)$ is odd.

546: Indeed, passing a CE of type $T$ in the positive direction

547: with respect to the permanent coorientation increases the number of triple point by 2, and the other CE

548: types leave the number of triple points unchanged.

549: (This formula may however differ by a constant from the $T$ defined using the given $i_0$ as base immersion).

550: Note that indeed $N-c \over 2$ is an integer, as is shown in 

551: \ct{b}, and may also be deduced using our present considerations.

552: For $P,Q$ we do not have an explicit formula for all immersions, however in Section \pr{fmq}

553: we will give an explicit formula for a certain situation.

554: 

555: 

556: We now classify all higher order invariants. We will define $E_n \su \Delta_n$ by two additional

557: restrictions on the functions $g\in \Delta_n$.

558: Let $Y = \{ T_+,H_+,Q_+ \} \su \CC_1$, then any $g \in \Delta_n$ is determined by its values on

559: un-ordered $n$-tuples of elements of $Y$ and so we may state these relations in terms of

560: such $n$-tuples. 

561: Given an un-ordered $n$-tuple $z$ of elements of $Y$, 

562: we define $m_{H_+}(z)$ and $m_{Q_+}(z)$ 

563: as the number of times that

564: $H_+$ and $Q_+$ appear in $z$ respectively.

565: We define $r(z)$, (the \emph{repetition} of $H_+$ and $Q_+$ in $z$), as 

566: $$r(z)= \max(0, m_{H_+}(z)-1) + \max(0, m_{Q_+}(z)-1).$$

567: 

568: 

569: \begin{dfn}\label{En}

570: Given an Abelian group $\G$,

571: $E_n = E_n(\G) \su \Delta_n(\G)$ is the subgroup consisting

572: of all $g\in \Delta_n(\G)$

573: satisfying the following two additional restrictions:

574: 

575: \begin{enumerate}

576: 

577: \item When $n \geq 3$, $g$ must satisfy the relation $H_+ H_+ Q_+ = H_+ Q_+ Q_+$.

578: By this we mean that 

579: $g([H_+, H_+, Q_+, {R_4}_{e_4},\dots,{R_n}_{e_n}]) =

580: g([H_+, Q_+, Q_+, {R_4}_{e_4},\dots,{R_n}_{e_n}])$ for

581: arbitrary ${R_4}_{e_4},\dots,{R_n}_{e_n} \in Y$.

582: \label{e1}

583: 

584: \item For any un-ordered $n$-tuple $z$ of elements of $Y$,

585: $g(z)\in 2^{r(z)}\G$, i.e. there exists an element $a\in \G$ such

586: that $g(z)=2^{r(z)} a$.

587: \label{e2}

588: 

589: \end{enumerate}

590: \end{dfn}

591: 

592: 

593: 

594: 

595: We define algebraic structures $K \su L \su M$, where $L$ is a commutative 

596: ring, $K$ is a subring of $L$, and $M$ is a module over $K$.

597: $L$ is defined as the ring of formal power series with integer coefficients

598: and variables $t,p,q$ and with relations 

599: \begin{itemize}

600: \item $2p=2q=0$.

601: \item $p^2 q = p q^2$. 

602: \end{itemize}

603: 

604: Given a monomial $f$, 

605: we define $m_p(f)$ and $m_q(f)$ as the multiplicity of $p$ and $q$ in $f$ respectively. We 

606: define $r(f)$, (the repetition of $p$ and $q$ in $f$), as 

607: $$r(f)= \max(0, m_p(f)-1) + \max(0, m_q(f)-1).$$ 

608: $r(f)$ is preserved under the 

609: relations in $L$ and so is well defined on 

610: equivalence classes of monomials.

611: The equivalence class of a monomial $f$ 

612: will be denoted $\f$. 

613: Now $K \su L$ is defined to be the subring of power series

614: including only the variable $t$.

615: On the other hand 

616: we extend $L$ to a larger structure $M$ which will 

617: be a module over the subring $K$, 

618: as follows:

619: For each $\f \in L$ for which $f$ is a monomial with coefficient 1,

620: we adjoin a new element $\zeta_\f$ satisfying the

621: relation $2^{r(\f)} \zeta_\f = \f$.  

622: Now we let $K$ act on $M$ by the natural extension of the action 

623: $t^m \cdot \zeta_{\f} = \zeta_{\overline{t^mf}}$.

624: We note that the whole of $L$ cannot act on $M$ in this way, since we would get

625: contradictions such as 

626: $0 = 2p \cdot \zeta_p = 2 \zeta_{p^2} = 2^{r(p^2)} \zeta_{p^2} = p^2 \neq 0$.

627: In particular, we do not have a ring structure on $M$.

628: For each $n\geq 0$ we denote by $K_n \su L_n \su M_n$ the additive subgroups

629: of $K\su L \su M$ respectively generated by the monomials of degree $n$

630: (where $\zeta_\f \in M$ is considered a monomial of the same degree as $\f$).

631: We note $L_1=M_1=\G_U$.

632: We have $L_1=K_1\oplus S$ where $S \su L_1$

633: is the four element subgroup generated by $p,q$.

634: 

635: We now define a function

636: $\FF:L_1\to M$ as follows: We first define $\FF:K_1 \to K$ as the group 

637: homomorphism

638: from the additive group $K_1 = \{ mt : m \in \Z \} \cong \Z$ to the multiplicative group of 

639: invertible elements in $K$, given on the generator $t$ of $K_1$ 

640: by $$\FF(t)=\sum_{n=0}^\infty t^n.$$ This is indeed an invertible element, giving

641: $\FF(-t)=\bigg(\sum_{n=0}^\infty t^n\bigg)^{-1} = 1-t$.

642: 

643: We then define $\FF:S\to M$ explicitly on the four elements of $S$ as follows: 

644: \begin{enumerate}

645: \item $\FF(0)=1$.

646: \item $\FF(p)=\sum_{n=0}^\infty \zeta_{p^n}$.

647: \item $\FF(q)=\sum_{n=0}^\infty \zeta_{q^n}$.

648: \item $\FF(p+q)=1+p+q+\sum_{n=2}^\infty (\zeta_{p^n} + \zeta_{q^n} + \zeta_{\overline{pq^{n-1}}})$.

649: \end{enumerate}

650: 

651: Finally, $\FF:L_1 \to M$ is defined as follows: Any element in $L_1$ is uniquely written as

652: $k+s$ with $k\in K_1, s\in S$, and we define $\FF(k+s)=\FF(k)\FF(s)$

653: where the product on the right is the action of $K$ on $M$. 

654: Let $\FF_n:L_1 \to M_n$ be the projection of $\FF$ into $M_n$.

655: We now state our classification theorem for finite order invariants.

656: The proof follows exactly as for the corresponding

657: claim for orientable surfaces appearing in \ct{h}. 

658: 

659: \begin{thm}\label{high}

660: For any closed non-orientable surface $F$, regular homotopy class $\A$ of immersions of $F$ 

661: into $\E$ and Abelian group $\G$, the 

662: image of the injection $u:V_n / V_{n-1} \to \Delta_n$ is $E_n$,

663: and the invariant $\FF_n \circ f_1^U : I_0 \to M_n$ is a universal order $n$ invariant.

664: \end{thm}

665: 

666: 

667: \section{$\4$-forms and automorphisms of $F$}\label{og}

668: 

669: Denote $\4=(\hf\Z)/(2\Z)$, which is a cyclic group of order 4 

670: (and $\4$ stands for \emph{half} integers).

671: The group $\C=\Z / 2\Z$ is contained in $\4$ as a subgroup.

672: 

673: \begin{dfn}\label{d1}

674: Let $E$ be a finite dimensional vector space over $\C$.

675: An $\4$-form is a map $g:E \to \4$ satisfying the following two conditions:

676: 

677: 

678: 

679: \begin{enumerate}

680: \item  For all $x,y \in E$: $$g(x+y) = g(x) + g(y) + C(x,y)$$

681: where $C:E \times E \to \C$  ($\su\4)$ is a non-degenerate symmetric bilinear form.

682: \item There is at least one $x \in E$ with $g(x) \not\in \C$

683: i.e. $g(x)=\pm\hf$.

684: \end{enumerate}

685: \end{dfn}

686: 

687: It follows that $g(0)=0$ and $2g(x) = C(x,x)$ showing $C(x,x) =1$ iff

688: $g(x) = \pm\hf$ and so there is at least one $x \in E$ with $C(x,x)=1$.

689: One can then show

690: there exists an ``orthonormal'' basis for $E$, i.e. a basis $e_1,\dots,e_n$ satisfying

691: $C(e_i,e_j)=\delta_{ij}$.

692: For such basis if $x=\sum_{j=1}^k e_{i_j}$ ($1\leq i_1 < \cdots < i_k \leq n$) 

693: then $g(x)=\sum_{j=1}^k d_{i_j}$ where $d_i=g(e_i)=\pm\hf$.

694: 

695: For $\4$-form $g$ on $E$ we define $O(E,g)$ to be the group of all

696: linear maps $T:E \to E$ satisfying $g(Tx)=g(x)$ for all $x \in E$.

697: It follows that $C(Tx , Ty) = C(x, y)$ for all $x,y \in E$ and that 

698: $T$ is invertible.

699: For $a \in E$ we define $T_a:E \to E$ to be the map $T_a(x)=x + C(x,a) a$.

700: Then $T_a \in O(E,g)$ iff $g(a)=1$ or $a=0$. 

701: For $a,b \in E$ we define $S_{a,b}:E \to E$ by:

702: $$S_{a,b}(x) =   T_a \circ T_b \circ T_{a+b} .$$  

703: One verifies directly that

704: if $a,b \in E$ satisfy $g(a)=g(b)=g(a+b)=0$ (which is equivalent to

705: $g(a)=g(b)=C(a,b)=0$), then $S_{a,b} \in O(E,g)$. 

706: In this case $S_{a,b}$ may also be written as: $S_{a,b}(x)=x + C(x,b)a + C(x,a)b$.

707: 

708: \begin{thm}\label{orth}

709: Let $g$ be an $\4$-form on $E$, then $O(E,g)$ is generated by

710: the set of elements of the following two forms:

711: \begin{enumerate}

712: \item $T_a$ for $a\in E$ with $g(a)=1$.

713: \item $S_{a,b}$ for $a,b \in E$ with $g(a)=g(b)=g(a+b) = 0$.

714: \end{enumerate}

715: Furthermore, if $\dim E\geq 9$

716: then the elements of the first form alone generate $O(E,g)$.

717: 

718: \end{thm}

719: 

720: \begin{pf}

721: Let $T\in O(E,g)$ and let $e_1,\dots,e_n$

722: be an orthonormal basis for $E$. Assuming inductively, with decreasing $k$,

723: that $T$ fixes $e_{k+1},\dots,e_n$, we will compose 

724: $T$ with elements of the above two forms to obtain a map which additionally

725: fixes $e_k$, eventually fixing all $e_i$ which will prove the statement.

726: For $x \in E$ define $\spp(x)$ to be the

727: set $\{ e_{i_1},\dots,e_{i_m} \}$ 

728: ($i_1 < \cdots < i_m$) such that $x=\sum_{j=1}^m e_{i_j}$. 

729: For $i \leq k$ and $j \geq k+1$, $C(Te_i,e_j)=C(e_i,e_j)=0$, 

730: and so for $i \leq k$, $\spp(Te_i) \su \{e_1,\dots,e_k \}$.

731: Denote $v=Te_k$, then $g(v)=g(e_k)=\pm\hf$.

732: 

733: 

734: \emph{Case A}: $C(v, e_k) =0$.

735: Letting $a=e_k+v$ we have $g(a)=g(e_k)+g(v)+C(e_k, v) =1$, 

736: $T_a \circ T(e_k)=T_a(v)=e_k$, and for $i \geq k+1$, $T_a \circ T (e_i)=T_a(e_i)=e_i$, so we are done.

737: 

738: \emph{Case B}: $C(v, e_k) =1$.

739: Since $\spp(v) \su \{e_1,\dots,e_k \}$, if $k=1$ then $v=e_k$ and we are done, and 

740: so we assume $k \geq 2$. 

741: We first show that $\spp(v) \neq \{e_1,\dots,e_k \}$.

742: Indeed if $v=\sum_{i=1}^k e_i$ then 

743: $C(Te_{k-1} ,\sum_{i=1}^k e_i) = C(Te_{k-1} , Te_k) =C(e_{k-1} , e_k)=0$.

744: Since $\spp(Te_{k-1}) \su \{ e_1,\dots,e_k \}$ this shows

745: that the number of elements in $\spp(Te_{k-1})$

746: is even, which implies $C(Te_{k-1} , Te_{k-1})=0$,

747: contradicting $C(e_{k-1} , e_{k-1}) =1$. 

748: Since $C(v , e_k) =1$ we know $e_k \in \spp(v)$, so say $e_1 \not\in \spp(v)$. 

749: If $g(e_1)=g(e_k)$ then 

750: since $C(e_1, v) =0$ and $C(e_1 , e_k) =0$,

751: by the argument of Case A, we can map $v$ to $e_1$ and then $e_1$ to $e_k$ and we are done. 

752: Otherwise, say $g(v)=\hf$ and $g(e_1)=-\hf$. 

753: Assuming $v \neq e_k$ (otherwise we are done), $e_k$ is not the only element in

754: $\spp(v)$.

755: The condition $g(v)=g(e_k)$ implies one of three possibilities (after relabeling indices): 

756: \begin{enumerate}

757: \item $k\geq 4$, $e_2,e_3 \in \spp(v)$, $g(e_2)=\hf$, $g(e_3)=-\hf$.

758: \item $k\geq 6$, $e_2,e_3,e_4,e_5  \in \spp(v)$ and the value of $g$ on

759: $e_2,e_3,e_4,e_5$ is $\hf$.  

760: \item $k\geq 6$, $e_2,e_3,e_4,e_5  \in \spp(v)$ and the value of $g$ on

761: $e_2,e_3,e_4,e_5$ is $-\hf$.

762: \end{enumerate}

763: In case 1 define $a=e_1+e_2, b=e_3+e_k$.

764: In case 2 define $a=e_1+e_2, b=e_3+e_4+e_5+e_k$.

765: In case 3 define $a=e_1+e_2+e_3+e_4, b=e_5+e_k$.

766: In all three cases $g(a)=g(b)=g(a+b)=0$ and so $S_{a,b}$ belongs to the

767: set of proposed generators.

768: In all three cases $C(v , a) =1$ and $C(v, b) =0$ and so 

769: $S_{a,b}(v)=v+b$. Furthermore, $S_{a,b}$ fixes all $e_j$ for $j \geq k+1$, 

770: and $C(v+b , e_k) = C(v, e_k) + C(b, e_k) = 1 + 1 =0$.

771: So we can use $S_{a,b}$ to map $v$ to $v+b$, 

772: and then by Case A we can map $v+b$ to $e_k$, and we are done.

773: 

774: We conclude by showing that if $n \geq 9$ then the elements $T_a$ alone generate

775: $O(E,g)$. Indeed in the proof above we used only maps $S_{a,b}$ where  

776: $\spp(a) \cup \spp(b)$

777: includes at most six elements. If $n \geq 9$ then for

778: each such pair $a,b$ there are at

779: least three basis elements not in $\spp(a) \cup \spp(b)$. 

780: In the span of such three basis elements there is an element $s$ with $g(s)=1$. 

781: One verifies directly, by checking on basis elements, 

782: that $S_{a,b} = T_s \circ T_{s+a} \circ T_{s+b} \circ T_{s+a+b}$.

783: Since $g(s)=g(s+a)=g(s+b)=g(s+a+b)=1$ we have indeed expressed the given element

784: $S_{a,b}$ as a product of four generators of the first form.

785: \end{pf}

786: 

787: \begin{dfn}\label{d2}

788: For $T \in O(E,g)$ let $\psi(T) = \psi_E(T)= \r(T - \I) \bmod{2} \in \C$

789: \end{dfn}

790: 

791: \begin{prop}\label{hom}

792: The map $\psi : O(E,g) \to \C$ is a homomorphism. 

793: \end{prop}

794: 

795: \begin{pf}

796: 

797: 

798: Assume first that $\dim E \geq 9$. In this case we know $O(E,g)$ is generated by elements 

799: of the form $T_a$ with $g(a)=1$. 

800: Let $\F(T) = \{x \in E : Tx=x\} = \ker(T-\I)$ then $\r(T-\I) = \cd \F(T)$. 

801: We may now use \ct{a} Lemma 3.1, which is stated in a slightly different setting, 

802: (of $\C$ valued quadratic forms,) but whose proof applies word by word to our setting.

803: The statement of \ct{a} Lemma 3.1 is as follows: If $g(a)=1$ then for any $T\in O(E,g)$,

804: $\cd \F(T \circ T_a) = \cd \F(T) \pm 1$. Since $\cd \F(T_a) =1$ and the

805: elements $T_a$ generate $O(E,g)$, $\psi$ is a homomorphism.

806: 

807: Assume now $\dim E <9$ and take some $E'$ with $\4$-form $g'$ such that

808: $\dim E + \dim E' \geq 9$. The function

809: $g\oplus g' : E \oplus E' \to \4$ defined by 

810: $g \oplus g' (x,x') = g(x)+g'(x')$, is an $\4$-form on $E \oplus E'$.

811: Let $u:O(E,g) \to O(E\oplus E' , g \oplus g')$ be the embedding given by

812: $u(T)= T \oplus \I_{E'}$. Since $\r(T-\I_E) = \r(T \oplus \I_{E'} - \I_{E \oplus E'})$

813: we have $\psi_E = \psi_{E \oplus E'} \circ u$.

814: Since $\dim(E \oplus E') \geq 9$,

815: $\psi_{E \oplus E'}$ is a homomorphism and so $\psi_E$ is a homomorphism.

816: \end{pf}

817: 

818: 

819: Returning to surfaces,

820: let $F$ be a closed non-orientable surface, and let $g:\hc \to \4$ be an $\4$-form

821: whose associated bilinear form $C(x,y)$ is the algebraic intersection form $x \cdot y$

822: on $\hc$. Let $\N = \N(F)$ be the group of all diffeomorphisms $h:F \to F$ up to isotopy.

823: For $h:F \to F$ let $h_*$ denote the map it induces on $\hc$. The subgroup

824: $\ng = \N(F)_g \su \N$ is defined by

825: $\ng = \{ h \in \N : h_* \in O(\hc , g) \}$.

826: 

827: A simple closed curve will be called a \emph{circle}. If $c$ is a circle in $F$, 

828: the homology class of $c$ in $\hc$ will be denoted by $[c]$. A circle $c$ in $F$ 

829: has an annulus neighborhood if $[c] \cdot [c] =0$ and a Mobius band neighborhood 

830: if $[c]\cdot[c] =1$. Such circles will be called $A$-circles and $M$-circles, respectively.

831: Given an $A$-circle $c$ in $F$, a Dehn twist along $c$ will be denoted $\Tc$.

832: The map induced on $\hc$ by $\Tc$ is $T_{[c]}$,

833: and so $\Tc \in \ng$ iff $g([c])=1$ or $[c]=0$.

834: Also, since $(T_{[c]})^2=\I$ whenever $[c]\cdot [c] =0$, $(\Tc)^2 \in\ng$ for any $A$-circle $c$.

835: 

836: Let $P \su F$ be a disc with two holes. 

837: Let $c,d,e$ be the three boundary circle of $P$, then $[c]+[d]=[e]$.

838: If $g([c])=g([d])=0$ (and so $g([e])=0$) then define $\SP = \Tc \circ \T_d \circ \T_e$.

839: The map induced by $\SP$ on $\hc$ is $S_{[c],[d]}$.

840: Finally, a $Y$-map $h:F\to F$ as defined in \ct{l} induces the identity on

841: $\hc$ and so $h \in \ng$.

842: 

843: \begin{dfn}\label{d3}

844: Let $F$ be a closed non-orientable surface and $g$ an $\4$ form on $\hc$. 

845: A map $h:F \to F$ will be called \emph{good} if it is of one of the following five

846: forms:

847: \begin{enumerate}

848: \item $h = (\Tc)^2$ for any $A$-circle $c$. 

849: \item $h=\Tc$ for an $A$-circle $c$ with $g([c])=1$

850: \item $h=\Tc$ for an $A$-circle $c$ with $[c]=0$

851: \item $h = \SP$ for some disc with two holes $P \su F$ with boundary circles

852: $c,d,e$ satisfying $g([c])=g([d])=0$ ($=g([e])$.

853: \item $h$ is a $Y$-map (defined in \ct{l}).

854: \end{enumerate}

855: A good map will be called of type 1 - 5 accordingly.

856: 

857: \end{dfn}

858: 

859: Whenever we consider two circles in $F$, we will assume they intersect transversally,

860: $|c \cap d|$ will then denote the number of intersection points between circles $c,d$.

861: (And so the algebraic intersection $[c]\cdot [d]$ in $\hc$ is the reduction mod 2 of 

862: $|c \cap d|$.)

863: 

864: \begin{thm}\label{K}

865: Let $F$ be a closed non-orientable surface and let $K_F = \{ h \in \N : h_* = \I \}$.

866: Then $K_F$ is generated by the good maps of type 1,3,5.

867: \end{thm}

868: 

869: 

870: \begin{pf}

871: Let $c_1,\dots,c_n$ be a family of disjoint $M$-circles in $F$ such that cutting

872: $F$ along $c_1,\dots,c_n$ produces a disc with $n$ holes. Then $[c_1],\dots,[c_n]$

873: is an orthonormal basis for $\hc$.

874: Now let $h \in K_F$.

875: We will compose $h$ with maps of type 1,3,5 until we obtain the identity.

876: Assume inductively that $h(c_i)=c_i$ for all $i \leq k-1$ (not necessarily respecting

877: the orientation of $c_i$). Denote $a = h(c_k)$, then $a$ is disjoint from $c_1,\dots,c_{k-1}$

878: and since $[a] \cdot [c_j] = [c_k] \cdot [c_j] = 0$ for $j \geq k+1$,

879: $|a \cap c_j|$ is even for $j \geq k+1$. We will now perform good maps of type 1,

880: which fix $c_1,\dots,c_{k-1}$, to map $a$ to a circle disjoint from all $c_j$, $j \geq k+1$.

881: Assume for some $j \geq k+1$, $|a \cap c_j|>0$, then since it is even and since $c_j$ is 

882: an $M$-circle, there must be two adjacent intersection

883: points along $c_j$ at which $a$ crosses $c_j$ in the same direction with respect to an

884: orientation on a neighborhood of the subinterval $b \su c_j$ satisfying $b \cap a = \pa b$.

885: Now $\pa b$ divides $a$ into two intervals $a',a''$ and since $a$ is an $M$ circle,

886: one of the circles $a' \cup b$ or $a'' \cup b$ is an $A$-circle. Say $c = a' \cup b$ is 

887: an $A$-circle. Then as shown in \ct{a} Figure 2, 

888: $|(\Tc)^2(a) \cap c_j|=|a \cap c_j|-2$. 

889: This map fixes 

890: $c_1,\dots,c_{k-1}$ and we may repeat this procedure until the image of $a$ (which we again

891: name $a$), is disjoint from all $c_j$, $j \neq k$.

892: 

893: We will now use good maps of type 5 to map $a$ onto $c_k$. 

894: Let $M_1,\dots,M_{k-1},M_{k+1},\dots,M_n$ be disjoint Mobius band neighborhoods 

895: of $c_1,\dots,c_{k-1},c_{k+1},\dots,c_n$ which are also disjoint from $c_k$ and $a$. 

896: Let $\tF$ be the projective plane obtained from $F$ by collapsing each $M_i$ to a point

897: $p_i$ ($i=1,\dots,k-1,k+1,\dots,n$). In $\tF$ one can isotope $a$ to coincide with $c_k$.

898: We lift this isotopy to $F$ as follows: Every time $a$ is about to pass one of the points $p_i$

899: in $\tF$, we realize this passage in $F$ by performing a good map of type 5, which drags

900: $M_i$ along a circle $a'$ which is close to $a$ and intersects $a$ once, returning $M_i$ 

901: to place (while reversing the orientation of $c_i$). This pushes $a$ to the other side 

902: of $\pa M_i$ in $F$ and so to the other side of $p_i$ in $\tF$.

903: In this way we bring $a$ to coincide with $c_k$.

904: 

905: We continue by induction until $h(c_i)=c_i$ for all $1\leq i \leq n$, but when performing the 

906: procedure involving $Y$-maps with the last circle $c_n$, we choose the isotopy in $\tF$

907: to bring $h(c_n)$ \emph{orientation preservingly} onto $c_n$. Since cutting $F$ along

908: $c_1,\dots,c_n$ produces an orientable surface (a sphere with $n$ holes), then if $c_n$ 

909: is mapped onto itself orientation preservingly, then the same must be true for

910: all $c_i$. After some isotopy we obtain that $h$ is the identity on all $c_i$.

911: 

912: Let $S$ be the sphere with $n$ holes obtained from $F$ by cutting it along 

913: $c_1,\dots,c_n$, then $h$ induces a diffeomorphism from 

914: $S$ to $S$ which is the identity on $\pa S$.

915: It is known that any such map is a composition of Dehn twists. Since any circle 

916: in $S$ separates $S$, the corresponding circle in $F$ will separate $F$ and so these

917: Dehn twists on $F$ are good maps of type 3.

918: 

919: \end{pf}

920: 

921: 

922: \begin{thm}\label{ng}

923: Let $F$ be a closed non-orientable surface and let $g$ be an $\4$-form on $\hc$.

924: Then $\ng$ is generated by the good maps. 

925: Furthermore, if $\dim \hc \geq 9$ then good maps of type 4 are not needed.

926: \end{thm}

927: 

928: \begin{pf}

929: For $h \in \ng$ we first compose $h$ with good maps of type 2,4 to obtain a map

930: in $K_F$, as follows. Any element $a \in \hc$ with $g(a)=1$ may be realized by

931: a circle in $F$ and so any generator $T_a$ of Theorem \pr{orth} may be realized 

932: by a good map of type 2.

933: Any $a,b \in \hc$ with $g(a)=g(b)=a \cdot b =0$ may be realized by a pair of disjoint circles

934: $c,d$ in $F$. Let $l$ be an arc connecting $c$ to $d$ (its interior being disjoint from $c,d$)

935: then a regular neighborhood $P$ of $c \cup l \cup d$ is a disc with two holes with $c,d$

936: isotopic to two of its boundary components. So any generator $S_{a,b}$ of Theorem \pr{orth}

937: may be realized by a good map of type 4. So by Theorem \pr{orth} we may compose $h$ with 

938: good maps

939: of type 2,4 to obtain a map in $K_F$, and if $\dim \hc \geq 9$ then maps of type 4 

940: are not needed. We complete the proof using Theorem \pr{K}.

941: \end{pf}

942: 

943: 

944: \section{Tangencies and quadruple points of regular homotopies}\label{fmq}

945: 

946: 

947: For regularly homotopic immersions $i,j:F \to \E$, let $P(i,j)=P(i)-P(j)$ and $Q(i,j)=Q(i)-Q(j)$.

948: So $P(i,j) \in \C$ (respectively $Q(i,j) \in\C$)

949: is the number mod 2 of tangency points (respectively quadruple points)

950: occurring in any generic regular homotopy between $i$ and $j$. 

951: In this section we prove the following:

952: 

953: \begin{thm}\label{qm}

954: Let $F$ be a closed non-orientable surface. Let $i:F\to \E$ be a stable immersion

955: and let $h:F \to F$ be a diffeomorphism such that $i$ and $i \circ h$ are regularly

956: homotopic.

957: Then 

958: $$P(i,i\circ h)=Q(i,i\circ h) =\bigg(\r(h_*-\I)+\ep(\det h_{**})\bigg)\bmod{2}$$

959: where $h_*$ is the map induced by $h$ on $\hc$, $h_{**}$ is

960: the map induced by $h$ on $H_1(F;\Q)$ and for $0 \neq q \in \Q$, 

961: $\ep(q) \in \C$ is 0 or 1

962: according to whether $q$ is positive or negative, respectively.

963: \end{thm}

964: 

965: For $h \in \ng$ define 

966: $\Omega(h)= ( \r(h_*-\I)+\ep(\det h_{**}) )\bmod{2} =\psi(h_*) + \ep(\det h_{**})  \in \C$.

967: So we must show $P(i,i\circ h)=Q(i,i\circ h)=\Omega(h)$.

968: By \ct{f} Proposition 7.1 we know $P(i,i\circ h)=Q(i,i\circ h)$. 

969: Indeed the proof of this case appearing there, 

970: does not use any assumption on orientability of the surface.

971: So it is enough to prove $Q(i,i\circ h)=\Omega(h)$.

972: 

973: Let $F$ be any closed surface and let $c \su F$ be a separating circle. Denote by

974: $F_1,F_2$ the two subsurfaces into which $c$ divides $F$. Denote by $\one$ the closed surface

975: obtained from $F$ by collapsing $F_2$ to a point, and similarly denote by $\two$

976: the closed surface obtained from $F$ by collapsing $F_1$ to a point. 

977: If $h:F \to F$ is a diffeomorphism such that $h(F_1) = F_1$ and $h(F_2) = F_2$

978: then $h$ induces maps $h_k:\tfk \to \tfk$ ($k=1,2$). 

979: As appears in \ct{a} Section 5.3, 

980: if $i:F \to \E$ is an immersion then it determines 

981: regular homotopy classes of immersions  $i_k:\tfk \to \E$, 

982: and if $h$ as above satisfies that $i$ and $i \circ h$ are regularly homotopic

983: then also $i_k$ and $i_k \circ h_k$ are regularly homotopic ($k=1,2$).

984: It is shown in \ct{a} Section 5.3 for orientable $F$, that for such $i$ and $h$,

985: if $h$ is orientation preserving then 

986: $Q(i,i\circ h)=Q(i_1,i_1\circ h_1) + Q(i_2,i_2\circ h_2)$ ($\in \C$)

987: whereas if $h$ is orientation reversing then

988: $Q(i,i\circ h)=Q(i_1,i_1\circ h_1) + Q(i_2,i_2\circ h_2) +1$ ($\in \C$).

989: But in fact the proof does not use the property that $h$ itself is orientation preserving

990: or reversing, but rather the corresponding property that $h|_c:c \to c$ is orientation

991: preserving or reversing. This latter property is meaningful also for non-orientable

992: surfaces, and indeed the proof works just the same for non-orientable surfaces. And so we have:

993: 

994: \begin{lemma}\label{add}

995: Let $F$ be any closed surface. In the above setting, if

996: $h|_c:c\to c$ is orientation preserving then 

997: $$Q(i,i\circ h)=Q(i_1,i_1\circ h_1) + Q(i_2,i_2\circ h_2)$$

998: and if $h|_c:c\to c$ is orientation reversing then

999: $$Q(i,i\circ h)=Q(i_1,i_1\circ h_1) + Q(i_2,i_2\circ h_2) +1.$$

1000: \end{lemma}

1001:  

1002: \begin{cor}\label{cadd}

1003: In the above setting if $h$ satisfies that $h(x)=x$ for all $x \in F_2$

1004: then $$Q(i,i\circ h)=Q(i_1,i_1\circ h_1).$$ 

1005: \end{cor}

1006: 

1007: \begin{lemma}\label{tf}

1008: In the above setting if $h$ satisfies that $h(x)=x$ for all $x \in F_2$,

1009: then $$\Omega(h)=\Omega(h_1).$$ 

1010: \end{lemma}

1011: 

1012: \begin{pf}

1013: 

1014: We first show $\psi(h_*)=\psi({h_1}_*)$. We have 

1015: $\hc \cong H_1(F_1;\C) \oplus H_1(F_2;\C) \cong H_1(\one;\C) \oplus H_1(\two;\C)$ and

1016: under this isomorphism, $h_*$ corresponds to ${h_1}_* \oplus \I$. It follows

1017: that $\psi(h_*)=\psi({h_1}_*)$. 

1018: 

1019: Next we show $\det h_{**} = \det {h_1}_{**}$. Since

1020: $\widetilde{H}_0(F_2;\C)=0$, we have the 

1021: exact sequence 

1022: $$H_1(F_2;\Q) \to H_1(F;\Q) \to H_1(F,F_2;\Q) \to 0.$$

1023: Let $r$ denote the map that $h$ induces on $H_1(F,F_2;\Q)$, then since

1024: $h$ induces the identity on $H_1(F_2;\Q)$, we have $\det h_{**} = \det r$.

1025: But $r$ corresponds to ${h_1}_{**}$ under the natural isomorphism

1026: $H_1(F,F_2;\Q) \to H_1(\one,p;\Q) = H_1(\one;\Q)$.

1027: 

1028: \end{pf}

1029: 

1030: Now for non-orientable $F$

1031: let $i:F \to \E$ be an immersion and let $g^i$ be the $\4$-form on $\hc$ 

1032: determined by $i$, as defined in \ct{p}. 

1033: Note that the notation in \ct{p} differs from ours in

1034: that $\4$ is taken there to be $\Z / 4\Z$ rather than $(\hf\Z) / 2\Z$. And so

1035: the numerical value of $g^i$ appearing there is twice the value here, and

1036: our relation $g(x+y)=g(x)+g(y)+x \cdot y$ is replaced there by 

1037: $g(x+y)=g(x)+g(y)+ 2(x \cdot y)$.

1038: We have as in \ct{a} Proposition 5.2, for any diffeomorphism 

1039: $h:F \to F$, $i$ and $i \circ h$ are regularly homotopic iff $h \in \N_{g^i}$.

1040: 

1041: Letting $n=\dim \hc$, we

1042: will prove that $Q(i,i \circ h) = \Omega(h)$ for all $n$ in the following order:

1043: $n=1$, $n=2$, $n \geq 9$, and finally $3 \leq n \leq 8$. 

1044: Denote $g=g^i$.

1045: Since by Theorem \pr{hom}, $\psi$ is a homomorphism 

1046: on $O(\hc,g)$

1047: and so $\Omega$ is a homomorphism on $\ng$, and since by \ct{a} Lemma 5.5

1048: (whose proof applies to non-orientable $F$) $h \mapsto Q(i,i \circ h)$ is a homomorphism,

1049: it is enough to check the equality $Q(i,i \circ h) = \Omega(h)$ for generators

1050: of $\ng$. 

1051: 

1052: For $n=1$, $F=\R P^2$ the projective plane. Since $\N(\R P^2)$ is trivial,

1053: the equality follows trivially.

1054: 

1055: For $n=2$, $F=Kl$ the Klein bottle. It is shown in \ct{l} that

1056: $\N(Kl) \cong \C \oplus \C$. There is a unique circle $c$ in $Kl$ up to isotopy which

1057: separates $Kl$ into two punctured projective planes $F_1,F_2$. 

1058: We have $\dim H_1(Kl;\C) =2$ with unique orthonormal basis $e_1,e_2$ 

1059: where $e_k$ ($k=1,2$) corresponds to the unique generator of $H_1(F_k,\C)$. 

1060: With respect to this basis 

1061: $h_* \in \{ 

1062: \left(\begin{smallmatrix} 1 & 0 \\ 0 & 1 \end{smallmatrix}\right) ,

1063: \left(\begin{smallmatrix} 0 & 1 \\ 1 & 0 \end{smallmatrix}\right) \}$

1064: for all $h \in \N(Kl)$.

1065: On the other hand $\dim H_1(Kl;\Q) =1$ with generator given by the circle $c$, and

1066: we have $h_{**} \in \{ \I , -\I \}$ for all $h \in \N(Kl)$. 

1067: The map $h \mapsto (h_*,h_{**})$ indeed realizes the 

1068: isomorphism $\N(Kl) \cong \C \oplus \C$. 

1069: For the unique $Y$-map $u:Kl \to Kl$, we have $u_* = \I$ and so 

1070: $u \in \ng$ and $\psi(u_*)=0$. On the other hand $u_{**}=-\I$ so 

1071: $\det u_{**} =-1$ so $\ep(\det u_{**}) =1$ giving finally $\Omega(u)=1$.

1072: Now we may isotope $u$ such that $u(F_1)=F_1$ and $u(F_2)=F_2$ and

1073: in this case $u|_c : c \to c$ is orientation reversing. 

1074: Since for $k=1,2$, $u_k \in \N(\R P^2)$ which is trivial, 

1075: we have by Lemma \pr{add} that $Q(i,i\circ u)=1$. 

1076: So we have shown $Q(i,i\circ u)=1=\Omega(u)$.

1077: 

1078: If $g(e_1) \neq g(e_2)$ then $u$ is the only non-trivial map in $\N_g$ and so we are done.

1079: Otherwise $g(e_1)=g(e_2)$ and then $\ng=\N$. By \ct{a} Lemma 5.8

1080: (whose proof applies to non-orientable $F$) we may replace $i$ by any other immersion in

1081: its regular homotopy class. Indeed we construct $i$ as follows: $F_1$ will be immersed

1082: in the positive side of the $yz$ plane, with its boundary $c$ embedded in the $yz$ plane

1083: symmetrically with respect to the $z$ axis. The immersion of $F_1$ is furthermore chosen to give

1084: the correct value for $g(e_1)$. Now the image of $F_2$ will be the punctured projective 

1085: plane obtained from the image of $F_1$ by a $\pi$ rotation around the $z$ axis, and

1086: so $g(e_2)=g(e_1)$ also has the correct value and so the new immersion $i$ is indeed

1087: in the correct regular homotopy class.

1088: The symmetry of $i(F)$ implies that there is

1089: a diffeomorphism $v:F \to F$ such that $\pi$ rotation of $i(F)$ around the $z$ axis is a 

1090: regular homotopy between $i$ and $i \circ v$. 

1091: Now $v_* = \left(\begin{smallmatrix} 0 & 1 \\ 1 & 0 \end{smallmatrix}\right)$

1092: and so $v$ together with the $Y$-map $u$ generate $\N$, and $\psi(v_*)=1$.

1093: Since $v$ reverses the orientation of $c$ we have $v_{**}=-\I$ and so

1094: $\ep(\det v_{**})=1$.

1095: Together, $\Omega(v)=1+1=0$. Now the rigid rotation of $i(F)$ 

1096: which is the regular homotopy between $i$ and $i \circ v$, has no quadruple points at all, and 

1097: so we get $Q(i,i\circ v)=0=\Omega(v)$, which completes the case $n=2$.

1098: 

1099: For $n\geq 9$ we have $\ng$ generated by good maps of type 1,2,3,5.

1100: If $h \in \ng$ is of type 3, then $h=\Tc$ for a separating circle $c \su F$.

1101: Let $c'$ be a circle parallel to $c$, dividing $F$ into $F_1,F_2$ with $c \su F_1$.

1102: Then $h(x)=x$ for all $x \in F_2$ and $h_1$ is isotopic to the identity

1103: on $\one$. And so by Corollary \pr{cadd} and Lemma \pr{tf} we have

1104: $Q(i,i\circ h) = Q(i_1,i_1\circ h_1) = 0 = \Omega(h_1) = \Omega(h)$.

1105: If $h$ is of type 1,2, then $h$ is $\Tc$ or $(\Tc)^2$ where (since

1106: we have already discussed good maps of type 3), we may assume $c$

1107: is a non-separating $A$-circle. So there is an $M$-circle $d \su F$ such that

1108: $|c \cap d|=1$ and so a regular neighborhood $F_1$ of $c \cup d$ is a punctured

1109: Klein bottle. Since we have already established the theorem for $n=2$ we know

1110: $Q(i_1,i_1\circ h_1)= \Omega(h_1)$. Since $h(x)=x$ for all $x \in F_2 = F-F_1$,

1111: we have by Corollary \pr{cadd} and Lemma \pr{tf}, 

1112: $Q(i,i\circ h) = Q(i_1,i_1\circ h_1) = \Omega(h_1) = \Omega(h)$.

1113: Finally if $h$ is of type 5, i.e. a $Y$-map, then again there is a punctured

1114: Klein bottle $F_1 \su F$ with $h(x)=x$ for $x \in F-F_1$ and so we are done as in the

1115: previous case, which completes the proof for $n\geq 9$.

1116: 

1117: For $3 \leq n \leq 8$, since $F$ is non-orientable, $h$ is isotopic to a map

1118: which fixes a disc $D \su F$ pointwise, so assume $h$ satisfies this property.

1119: Take any closed surface $F'$ with 

1120: $\dim \hc + \dim H_1(F';\C) \geq 9$ and construct $G = F \# F'$ where the connect

1121: sum operation is performed by deleting the above mentioned disc $D$ from $F$, and some

1122: disc $D'$ from $F'$. There clearly exists an immersion $j:G \to \E$ such that 

1123: $j|_{F-D}=i|_{F-D}$. 

1124: By our assumption on $D$ we can extend $h|_{F-D}$ to a diffeomorphism

1125: $u:G \to G$ by defining $u(x)=x$ for all $x \in F' -D'$. 

1126: Denote $G_1=F-D$ and $G_2 = F'-D'$ then $\widetilde{G_1}$ is naturally identified with

1127: $F$ and under this identification $j_1$ corresponds to $i$ and $u_1$ corresponds to $h$.

1128: Since $\dim H_1(G;\C) \geq 9$ our theorem is already proved for $G$ and so by Corollary

1129: \pr{cadd} and Lemma \pr{tf} we get

1130: $Q(i,i\circ h) = Q(j,j\circ u) = \Omega(u) = \Omega(h)$.

1131: This completes the proof of Theorem \pr{qm}.

1132: 

1133: 

1134: 

1135: 

1136: \begin{thebibliography}{StoPC}

1137: 

1138: \bibitem[B]{b}

1139: T. F. Banchoff:

1140: ``Triple points and surgery of immersed surfaces.''

1141: \emph{Proc. Amer. Math. Soc.} 46 (1974) 407--413.

1142: 

1143: \bibitem[HK]{hk}

1144: C.A. Hobbs, N.P. Kirk: ``On the Classification and Bifurcation of Multigerms of Maps 

1145: from Surfaces to 3-Space.'' - \emph{Mathematica Scandinavica}

1146: 89 (2001) 57--96. 

1147: 

1148: \bibitem[L]{l}

1149: W.B.R. Lickorish: 

1150: ``Homeomorphisms of non-orientable two-manifolds.'' 

1151: \emph{Proceedings of the Cambridge Philosophical Society} 59 (1963) 307--317.

1152: 

1153: 

1154: \bibitem[N1]{q}

1155: T. Nowik:

1156: ``Quadruple points of regular homotopies of surfaces in 3-manifolds.''

1157: \emph{Topology} 39 (2000) 1069-1077.

1158: 

1159: \bibitem[N2]{a}

1160: T. Nowik:

1161: ``Automorphisms and embeddings of surfaces and quadruple points of regular homotopies.'' 

1162: \emph{Journal of Differential Geometry}  57 (2001) 421-455.

1163: 

1164: \bibitem[N3]{o}

1165: T. Nowik:

1166: ``Order one invariants of immersions of surfaces into 3-space.'' 

1167: \emph{Mathematische Annalen} - to appear

1168: (may be viewed at http://www.math.biu.ac.il/$\sim$tahl).

1169: 

1170: \bibitem[N4]{h}

1171: T. Nowik:

1172: ``Higher order invariants of immersions of surfaces into 3-space.'' 

1173: - Preprint - (http://www.math.biu.ac.il/$\sim$tahl).

1174: 

1175: \bibitem[N5]{f}

1176: T. Nowik

1177: ``Formulae for Order One Invariants of Immersions and Embeddings of Surfaces.''

1178: - Preprint - (http://www.math.biu.ac.il/$\sim$tahl).

1179: 

1180: \bibitem[P]{p}

1181: U. Pinkall: 

1182: ``Regular homotopy classes of immersed surfaces.''

1183: \emph{Topology} 24 (1985) No.4, 421--434.

1184: 

1185: 

1186: 

1187: 

1188: \end{thebibliography}

1189: 

1190: 

1191: 

1192: %%%%%%%%%%%%%%

1193: \end{document}

1194: %%%%%%%%%%%%%%

1195: 

1196: 

1197: