1: %Version: A.Degtyarev 31/10/99 19:45:50
2: %Format: amsppt
3: \def\inputfile{r-a}
4: \ifx\rokhloaded\undefined\input rokh.def \fi
5:
6: \let\dd\partial
7: \def\aug{\operator{aug}}
8: \def\Sm{\operator{Sm}}
9:
10: \let\GD\Delta
11:
12: \def\barX{X'}
13: \def\barA{A'}
14: \def\scup{\mathbin{\scriptstyle\cup}}
15: \def\scap{\mathbin{\scriptstyle\cap}}
16:
17: %\nowarning\cite
18:
19: \begin
20:
21: \appendix{Topology of involutions}\label{apA}
22:
23: \subsection{Smith theory}\label{Smith.theory}\label{s7}
24: Most results cited in this section are due to P.~A.~Smith; proofs can be
25: found, e.g., in~\cite{Bredon, Chapter~3}. Throughout the section we
26: consider a topological space~$X$ with involution $c\:X\to X$ and denote
27: $F=\Fix c$ and $\barX=X/c$.
28:
29: \theorem\label{1.1.1}\label{Smith.seq}
30: There are two natural, in respect to equivariant maps, exact sequences,
31: called \rom(homology and cohomology\rom) \emph{Smith sequences}
32: of~$(X,c)$\rom:
33: $$
34: \gather
35: @>>>H_{p+1}(\barX,F)@>\,\,\GD>>H_p(\barX,F)\oplus H_p(F)
36: @>\,\,{\tr^*}+{\inc_*}>>H_p(X)
37: @>\,\,\pr_*>>H_p(\barX,F)@>>>\rlap{\,,}\\
38: @>>>H^p(\barX,F)@>\,\,\pr^*>>H^p(X)@>\,\,{\tr_*}\oplus{\inc^*}>>
39: H^p(\barX,F)\oplus H^p(F)@>\,\,\GD>>H^{p+1}(\barX,F)@>>>\rlap{\,.}
40: \endgather
41: $$
42: The homology and cohomology connecting homomorphisms~$\GD$ are given by
43: $$
44: x\mapsto x\scap\Go\oplus\dd x\quad\text{and}\quad x\oplus f\mapsto
45: x\scup\Go+\Gd f,
46: $$
47: respectively, where $\Go\in H^1(\barX\sminus F)$ is the characteristic
48: class of the double covering $X\sminus F\to\barX\sminus F$. The images
49: of~${\tr^*}+\inc_*$ and~$\pr^*$ consist of invariant classes\rom:
50: $\Im\tr^*\subset\Ker(1+c_*)$ and $\Im\pr^*\subset\Ker(1+c^*)$.
51: \pni
52: \endtheorem
53:
54: \subsubsection{}\label{1.1.1'}
55: The \emph{transfer} homomorphisms $\tr^*$ and~$\tr_*$ in~\ref{1.1.1} can
56: be described as follows. Represent~$X$ as the $S^0$-bundle associated
57: with a relative $D^1$-bundle $T\to\barX$ whose characteristic class
58: is~$\Go$. Namely, take the cylinder of the projection $X\to\barX$ and for
59: each $f\in F$ contract the fiber over $f$. Then $\tr^*$ and $\tr_*$ are
60: the compositions
61: $$
62: \spreadlines{-\jot}
63: \align
64: \tr^*\:\quad&
65: H_p(\barX,F)@>\,\,\th^{-1}>>H_{p+1}(T,X)@>\,\,\dd>>H_p(X),\\
66: \tr_*\:\quad&
67: H^p(X)@>\,\,\Gd>>H^{p+1}(T,X)@>\,\,\th^{-1}>>H^p(\barX,F),
68: \endalign
69: $$
70: where $\th$ stands for the Thom isomorphisms.
71:
72: The following immediate consequences of~\ref{1.1.1}, which we state in
73: the homology setting, has an obvious counterpart for cohomology.
74:
75: \corollary[Corollary \rm(of~\ref{1.1.1})]\label{Sinqty}\label{1.1.2}
76: \roster
77: \item\local1
78: \strut$\dim H_*(F)+2\sum_{p}\dim\Coker({\tr^p}+\inc_p)=\dim H_*(X)$
79: \rom(Smith identity\rom)\rom;
80: \item\local2
81: \strut$\dim H_*(F)\le\dim H_*(X)$ \rom(Smith inequality\rom)\rom;
82: \item\local3
83: \strut$\dim H_*(F)\le\dim H^1(\ZZ;H_*(X))$ \rom(Borel-Swan
84: inequality\rom)\rom;
85: \item\local4
86: \strut$\dim H_*(F)=\dim H_*(X)\bmod2$\rom;
87: \item\local5
88: \strut$\chi(X)=\chi(F)+2\chi(\barX,F)$\rom;
89: \item\local6
90: \strut$\chi(X)=2\chi(\barX)-\chi(F)$
91: \endroster
92: \rom(Recall that $H^1(\ZZ;H_*(X))=\Ker(1+c_*)/\Im(1+c_*)$.\rom)
93: \pni
94: \endcorollary
95:
96: \subsubsection{}\label{1.1.8}
97: If \iref{1.1.2}2 turns into an equality, i.e., $\dim H_*(F)=\dim H_*(X)$,
98: one says that $c$ is an \emph{$M$-involution}. If \iref{1.1.2}3 turns
99: into an equality (which is equivalent to
100: $\Im({\tr^*}+\inc_*)\supset\Ker(1+c_*)$), $c$ is called
101: \emph{\rom($\ZZ$-\rom)Galois maximal}. (This terminology is introduced by
102: V.~A.~Krasnov \cite{Kr:GM}. R.~Thom~\cite{ThomStras} calls a dimension
103: $p\in\N$ \emph{regular} for $(X,c)$ if
104: $\Im({\tr^p}+\inc_p)\supset\Ker(1+c_p)$.)
105:
106: In general, due to~\iref{1.1.2}2 and~\ditto4, one has $\dim H_*(F)=\dim
107: H_*(X)-2d$ for some integer $d\ge0$; in this case $c$ is called an
108: \emph{$(M-d)$-involution}.
109:
110: \subsubsection{Geometrical construction of the Smith sequences}\label{1.1.9}
111: Introduce the \emph{Smith chain complexes}
112: $$
113: \align
114: \Sm_*(X)&=\Ker[(1+c_*)\:S_*(X)\to S_*(X)],\\
115: \Sm_*(X,F)&=\Ker[(1+c_*)\:S_*(X,F)\to S_*(X,F)].
116: \endalign
117: $$
118: and \emph{Smith homology} $H_r(\Sm_*(X))$ and $H_r(\Sm_*(X,F))$. There is
119: a canonical isomorphism $\Sm_*(X,F)=\Im[(1+c_*)\:S_*(X)\to S_*(X)]$ and a
120: canonical splitting $\Sm_*(X)=S_*(F)\oplus\Im(1+c_*)$. Furthermore,
121: $\tr^*\:S_*(\barX,F)\to\Sm_*(X,F)$ is an isomorphism, and in view of the
122: above identifications the Smith sequences are the long homology and
123: cohomology exact sequences associated with the short exact sequence of
124: complexes
125: $$
126: 0@>>>\Sm_*(X)@>\,\,\operator{inclusion}>>S_*(X)
127: @>\,\,1+c_*>>\Sm_*(X,F)@>>>0.
128: $$
129:
130: \theorem[Steenrod squares in the Smith sequences]\label{1.1.10'}
131: For any $x\in H_p(\barX,F)$ and $y\in H^p(X)$ one has
132: $\Sq\tr^*x=\tr^*\Sq(x\scap(1+\Go))$ and $\tr_*\Sq
133: y=(\Sq\tr_*y)\scup(1+\Go)$.
134: \endtheorem
135:
136: \proof
137: The statement follows from~\ref{1.1.1'} and the Thom-Cartan formula for
138: sphere bundles (see~\cite{ThomAEN}): since the total characteristic
139: class of the relative $D^1$-bundle $T\to\barX$ is $1+\Go$, one has
140: $\th^{-1}\Sq(x\scap(1+\Go))=\Sq\th^{-1}x$ and $\th^{-1}\Sq
141: y'=(\Sq\th^{-1}y')\scup(1+\Go)$, where $y'=\Gd y$.
142: \endproof
143:
144: \corollary\label{1.1.10}
145: For any $x\in H_p(\barX,F)$ and $y\in H^p(X)$ one has
146: $\Sq_1\tr^*x=\tr^*\Sq_1x+\inc_*\dd x$ and
147: $\tr_*\Sq^1y=\Sq_1\tr_*y+\Gd\inc^*y$.
148: \endcorollary
149:
150: \proof
151: Since the Smith sequences are exact, $\tr^*(x\scap\Go)=\inc_*\dd x$ and
152: $(\tr_*y)\scup\Go=\Gd\inc^*y$. Hence, the statement follows
153: from~\ref{1.1.10'}.
154: \endproof
155:
156: \subsubsection{Relative version of the Smith sequences}\label{1.1.11}
157: If $A\subset X$ is a $G$-invariant subspace, there is a natural exact
158: sequence
159: $$
160: \align
161: @>>>H_{p+1}(\barX,F\cup\barA)&@>\,\,\GD>>
162: H_p(\barX,F\cup\barA)\oplus H_p(F,F\cap\barA)
163: @>\,\,{\tr^*}+{\inc_*}>>\\
164: &\qquad\qquad@>\,\,{\tr^*}+{\inc_*}>>H_p(X,A)
165: @>\,\,\pr_*>>H_p(\barX,F\cup\barA)@>>>
166: \endalign
167: $$
168: (and a similar sequence in cohomology), where $\barA=A/c$. Its properties
169: and consequences are similar to those of the absolute Smith sequence. In
170: particular, the Smith inequality~\iref{1.1.2}2 turns into
171: $$
172: \dim H_*(F,F\cap A)\le\dim H_*(X,A).\eqtag\label{1.1.12}
173: $$
174:
175: \subsection{Kalinin's spectral sequence}\label{C1.7}
176: In this section we describe a relatively new powerful tool, so called
177: \emph{Kalinin's spectral sequence}. Originally it was introduced by
178: I.~Kalinin~\cite{Kalinin} as a stabilized version of the Borel-Serre
179: spectral sequence of the fibration $X\times_{\ZZ}S^\infty\to\Rp\infty$,
180: where $X$~is a topological space with involution~$c$ and $S^\infty$ is
181: considered with the standard antipodal involution. The sequence starts at
182: the homology $H_*(X)$ and converges to the total homology $H_*(\Fix c)$.
183: The resulting filtration $\CF^*$ on $H_*(\Fix c)$ and the isomorphisms
184: $\vr_p$ between the limit term of the spectral sequence and
185: $\Grad_{\CF}H_*(\Fix c)$ were discovered by O.~Viro geometrically before
186: Kalinin's work and were primarily related to the Smith exact sequence.
187: The relation between the two construction is established
188: in~\cite{Degt:Stiefel}. As is shown in~\cite{DegtKh:2} (preprint version)
189: or~\cite{DIK}, Kalinin's spectral sequence can be derived from the Smith
190: exact sequence as well (see~\ref{C1.7.D}).
191:
192: %Related constrictions
193: A construction similar to Kalinin's spectral sequence was studied by
194: V.~A.~Krasnov in a series of papers; see, e.g.,~\cite{Kr:94}.
195:
196: Below we give an alternative, geometrical, description of the homology
197: version of Kalinin's spectral sequence and Viro homomorphisms~$\vr_*$ and
198: state their main properties. (As usual, the cohomology version is obtained
199: by raising all indices and reversing all arrows; we mention it when
200: speaking about multiplications and Poincar\'e duality.) Proofs of these
201: results and/or further references can be found in~\cite{DegtKh:2} and
202: \cite{DIK}. The differentials~$\rd_p$ of the spectral sequence and Viro
203: homomorphisms~$\vr_p$ are often regarded as additive relations (i.e.,
204: partial multivalued homomorphisms) $H_p(X)\dasharrow H_{p+r-1}(X)$ and
205: $\bv_p\: H_*(\Fix c)\dasharrow H_p(X)$, respectively. As our approach is
206: geometrical, we have to use chains; depending on the nature of~$X$ one
207: may regard them as singular, simplicial, smooth, etc. To assure
208: convergence of the spectral sequence, $X$ must satisfy certain
209: conditions, which depend on the homology theory chosen (e.g., sheaf
210: theories and locally compact finite dimensional spaces). Certainly, these
211: conditions are always fulfilled for compact smooth manifolds and finite
212: $CW$-complexes.
213:
214: Thus, let us fix a good (see above) topological space~$X$ with
215: involution~$c$. Consider the partial homomorphisms $\bv_p\:H_*(\Fix
216: c)\dasharrow H_p(X)$ and the $\Z$-graded spectral sequence
217: $(\rH_*(X),\rd_*)$ defined as follows:
218: \roster
219: \item
220: $\bv_0$ is zero on $H_{\ge1}(\Fix c)$ and its restriction to $H_0(\Fix
221: c)$ coincides with the inclusion homomorphism\rom;
222: \item
223: $\vr_p$ is defined on a\/ \rom(nonhomogeneous\/\rom) element~$x\in
224: H_*(\Fix c)$ represented by a cycle $\sum x_i$ (where $x_i$ is the
225: $i$-dimensional component of~$x$) if and only if there exist some chains
226: $y_i$ in~$X$, $1\le i\le p$, so that $\partial y_1=x_0$ and $\partial
227: y_{i+1}=x_i+(1+c_*)y_i$ for $i\ge1$. In this case $\vr_p x$ is
228: represented by the class of $x_p+(1+c_*)y_p$ in $H_p(X)$\rom;
229: \item
230: $\RH1_*=H_*(X)$ and $\Rd1_*=1+c_*$\rom;
231: \item
232: $\rd_p$ is defined on a cycle $x_p$ in~$X$ if and only if there are chains
233: $y_p=x_p$, $y_{p+1}$,\dots, $y_{p+r-1}$ so that $\partial
234: y_{i+1}=(1+c_*)y_i$. In this case $\rd_px_p=(1+c_*)y_{p+r-1}$.
235: \endroster
236:
237: \claim[Spectral sequence of involution]\label{C1.7.A}
238: The homomorphisms~$\bv_*$ and spectral sequence $(\rH_*,\rd_*)$ are
239: natural with respect to equivariant maps. Furthermore, $\rH_*$ and $\rd_*$
240: do form a spectral sequence \rom(i.e., $\rd_p$ are well defined
241: homomorphisms $\rH_p\to\rH_{p+r-1}$ and\/
242: $\RH{r+1}_p=\Ker\rd_p/\Im\rd_{p-r+1}$\rom), and this sequence converges to
243: $H_*(\Fix c)$ via~$\bv_*$, i.e., $\bv_p$ induces an \rom(honest\rom)
244: isomorphism $\CF^p/\CF^{p+1}\to\iH_p$, where
245: $\CF^p=\operatorname{Domain}\vr_p=\Ker\vr_{p-1}$.
246: \qed
247: \endclaim
248:
249: \remark{Remark}
250: A detailed analysis of the construction above shows that $\bv_p$ can as
251: well be regarded as a true homomorphism
252: $$
253: \bv_p\:\Ker[\bv_{p-1}\:H_*(\Fix c)\dasharrow\RH{r-1}_{p-1}]\to\rH_p,
254: \quad r>1
255: $$
256: (where $\bv_p\:H_*(\Fix c)\dasharrow\RH1_p=H_p(X)$ is defined as the
257: inclusion homomorphism on $H_p(\Fix c)$ and zero on $H_{>p}(\Fix c)$\,).
258: These homomorphisms are, in fact, related to the original, nonstabilized,
259: Borel-Serre spectral sequence (see~\cite{Degt:Stiefel} for details).
260: \endremark
261:
262: The cohomology version $\rH^*\Rightarrow H^*(\Fix c)$ is constructed
263: similarly and is dual to the homology one. The cup- and cap-products
264: in~$X$ naturally descend to, respectively, a $\ZZ$-algebra structure
265: in~$\rH^*$ and $\rH^*$-module structure in~$\rH_*$. If $X$ is a connected
266: closed $N$-manifold and $\Fix c\ne\varnothing$, the fundamental class
267: $[X]$ survives to~$\iH_N$ and the multiplication
268: $\capprod[X]\:\rH^p\to\rH_{N-p}$ is an isomorphism (Poincar\'e duality).
269: This, in the usual way, defines a pairing
270: $\circ\:\rH_p\otimes\rH_q\to\rH_{p+q-N}$, called \emph{Kalinin's
271: intersection pairing}.
272:
273: \claim[Kalinin's intersection pairing]\label{C1.7.B}
274: Let $X$ be a smooth closed connected $N$-man\-i\-fold with a smooth
275: involution $c\:X\to X$. Denote by~$w(\nu)$ the total Stiefel-Whitney
276: class of the normal bundle~$\nu$ of~$\Fix c$ in~$X$. Then for
277: $a\in\CF^p$ and $b\in\CF^q$ one has $w(\nu)\capprod(a\circ b)\in
278: \CF^{p+q-N}$ and
279: $$
280: \bv_p a\circ\bv_q b=\bv_{p+q-N}[w(\nu)\capprod(a\circ b)].\rlap\qed
281: $$
282: \endclaim
283:
284: \remark{Remark}
285: The above identity contains, as a special case, the
286: Borel-Haefliger~\cite{BH} version of the Bezout theorem over the reals
287: (with the $\ZZ$-valued intersection): given two real algebraic cycles
288: $A$, $B$ of complementary dimension in a real algebraic manifold~$X$,
289: their intersection number in~$X$ is congruent $\bmod\,2$ to the
290: intersection number of~$\RA$ and~$\R B$ in~$\RX$. To see this, it is
291: sufficient to notice that $[\RA]$ is the highest degree term of
292: $\bv^{-1}[A]$.
293: \endremark
294:
295: The homology Steenrod operations $\Sq_t\: H_*(X)\to H_{*-t}(X)$ also
296: descend to $\rH_*$. In order to describe their relation to the ordinary
297: Steenrod operations in $H_*(\Fix c)$, introduce the \emph{weighed Steenrod
298: operations}
299: $$
300: \hSq_t\: H_p(X)\to H_{\le p}(X),\qquad
301: x\mapsto\sum_{0\le j\le t}{\tsize\binom{P-p}{t-j}}\Sq_jx,
302: $$
303: where $P>p+t$ is a power of~2. (The binomial coefficients do not depend
304: on~$P$, see, e.g., \cite{SteenrodEpstein, Lemma~I.2.6}.) Then one has:
305:
306: \claim[Steenrod squares in Kalinin's spectral sequence]\label{C1.7.C}
307: If $x\in\CF^p$ and $t\ge0$, then \smash{$\hSq_tx\in\CF^{p-t}$} and
308: $$
309: \Sq_t\bv_px=\bv_{p-t}\hSq_tx.\rlap\qed
310: $$
311: \endclaim
312:
313: Kalinin's spectral sequence can alternatively be derived from the Smith
314: exact sequence~\ref{Smith.seq}.
315: %$$
316: %\multline
317: %\dots\to H_{r+1}(X',\Fix c)@>\GD_{r+1}>> H_r(X',\Fix c)\oplus
318: %H_r(\Fix c)@>\Ga_r>>\\ @>\Ga_r>>H_r(X)@>\Gb_r>>H_r(X',\Fix c)\to\dots
319: %\endmultline
320: %$$
321: %(Recall that $X'=X/c$ is the orbit space of~$X$.)
322: Below we give the corresponding description of the differentials~$\rd_*$
323: and Viro homomorphisms~$\vr_*$; the construction of the spectral sequence
324: and filtration~$\CF^*$ can be found in~\cite{DIK}.)
325:
326: \claim[Kalinin's spectral sequence and Smith exact sequence]\label{C1.7.D}
327: The differentials~$\rd_p$ and Viro homomorphisms~$\vr_p$ are given by
328: $$
329: \gather
330: \rd_p=\tr^{p+r-1}\circ\GD_{p+r-1}^{-1}\circ\ldots\circ
331: \GD_{p+1}^{-1}\circ\pr_p\:H_p(X)\dasharrow H_{p+r-1}(X),\\
332: \vspace{1\jot}
333: \bv_p({\textstyle\sum_{r\le p}x_i})=\inc_p x_p+\tr^p y'_p
334: \:H_*(\Fix c)\dasharrow H_p(X),
335: \endgather
336: $$
337: where $\inc_*\: H_*(\Fix c)\to H_*(X)$ is the inclusion
338: homomorphism, $\tr^*$, $\pr_*$, and $\GD_*$ are as in
339: the Smith exact sequence~\ref{Smith.seq}, and, in the latter equation,
340: $x_r\in H_r(\Fix c)$ and $y'_p\in H_p(X',\Fix c)$ is defined recursively
341: via $y'_0=0$, \ $y'_{r+1}=\GD_{r+1}^{-1}(x_r\oplus y'_r)\in
342: H_{r+1}(X',\Fix c)$.
343: \qed
344: \endclaim
345:
346: \subsubsection{Lifting involutions}\label{apdx.lifts}
347: Let $(X,c)$ be a topological space with involution, $Y\to X$ a double
348: covering, and $\Go\in H^1(X)$ its characteristic class. Then $c$ lifts to
349: an involution on~$Y$ if and only if $\Go$ survives to $\iH^1(X)$. If this
350: is the case, there obviously are two lifts $c_1,c_2\:Y\to Y$, which
351: commute with each other and with the deck translation~$\tau$ of the
352: covering and satisfy the identity $c_1\circ c_2=\tau$.
353:
354: An obvious necessary condition for $\Go$ to survive to $\iH^1$ is that it
355: must be $c^*$-invariant (i.e., survive to $\RH2^1$). Note that if $X$ is
356: connected and $\Fix c\ne\varnothing$, this condition is also sufficient
357: as all the differentials landing in $\rH^0$ are trivial.
358:
359: Everything above applies to branched coverings as well, with~$X$ replaced
360: with $X\sminus A$, where $A$ is the branch set.
361:
362: \subsubsection{Complex surfaces}\label{Kalinin.surface}
363: Kalinin's spectral sequence, as well as the related intersection pairing
364: and Steenrod operations have a transparent geometrical meaning in the
365: case when $X$ is a complex surface and $c=\conj$ a real structure. Let
366: $C_i$, $i=1,\dots,k$, be the components of~$\RX$, $\<C_i>\in H_0(\RX)$
367: the corresponding classes, and $C_I=\sum_{i\in I}\<C_i>$ for
368: $I\subset\{1,\dots,k\}$. The values $\bv_0\<C_i>$, $\bv_1\Ga$ (with
369: $\Ga\in H_1(\RX)$), and $\bv_2[C_i]$ are always well defined and coincide
370: with the images of the classes under the inclusion homomorphisms. The
371: value $\bv_1\<C_I>$ is well defined whenever the number $\ls|I|$ of
372: summands is even; it is represented by the collection $(1+\conj_*)y_1$ of
373: invariant circles, where $y_1$ is a collection of arcs in~$X$ with one
374: boundary point in each of the summands~$C_i$, $i\in I$ (and no other
375: boundary points). If, under an appropriate choice of~$y_1$, the curve
376: $(1+\conj_*)y_1$ is null homologous, $(1+\conj_*)y_1=\dd y_2$, then
377: $\bv_1\<C_I>=0$ and $(1+\conj_*)y_2$ represents~$\bv_2\<C_I>$. Similarly,
378: if $\bv_1\Ga=0$, i.e., $\Ga$ is homologous to $(1+\conj_*)y_1$ for a
379: %cycle
380: chain~$y_1$\mnote{\Dg: edited after Zv} in~$X$, there is a chain~$y_2$
381: in~$X$ such that $\dd y_2=\Ga+(1+\conj_*)y_1$, and $(1+\conj_*)y_2$
382: represents~$\bv_2\Ga$. Finally, if $\bv_1\Ga=\bv_1\<C_I>$ for some
383: $0$-class $C_I$ and $1$-class $\Ga\in H_1(\RX)$, then $\bv_2(\Ga+\<C_I>)$
384: is defined and is represented by $(1+\conj_*)y_2$, where $\dd
385: y_2=\Ga+(1+\conj_*)y_1$ and $y_1$ is an appropriately chosen collection
386: of arcs with one boundary point in each of~$C_i$, $i\in I$, and no other
387: boundary points.
388:
389: Elements of the form $\bv_2[C_i]$, $\bv_2\<C_I>$, $\bv_2\Ga$, and
390: $\bv_2(\Ga+\<C_I>)$ span~$\iH_2(X)$; with an abuse of the language we let
391: $\sum\bv_2x_j=\bv_2\sum x_j$, provided that the latter is well defined,
392: even if the summands are not. According to~\ref{C1.7.B}, Kalinin's
393: intersection form on~$\iH_2(X)$ is the one given by Table~\ref{tabKif},
394: where the intersection $\Ga\ii\Gb$ is regarded as an element of
395: $H_0(\RX)$, and $\aug(\Ga\ii\Gb)$ and $\aug_i(\Ga\ii\Gb)$ are,
396: respectively, the total intersection number and its portion falling
397: into~$C_i$. The notation $\Gd_{ij}$~stands for the Kronecker symbol:
398: $\Gd_{ii}=1$ and $\Gd_{ij}=0$ if $i\ne j$. The intersection form extends
399: linearly to the classes of the form $\bv_2(\Ga+\<C_I>)$, as if $\bv_2\Ga$
400: and $\bv_2\<C_I>$ were well defined.
401:
402: \topinsert
403: \table\label{tabKif}
404: Kalinin's intersection form
405: \endtable
406: \offinterlineskip
407: \def\sp{\quad}
408: \let\a\Ga
409: \let\b\Gb
410: \def\d[#1]{\Gd_{#1}}
411: \def\\{\cr\noalign{\hrule}\omit\cr}
412: \let\c\ii
413: \line{\hfill\vbox{
414: \ialign{\vrule height11pt depth5pt\hfil\sp$#$\sp\hfil\vrule&&
415: \hfil\sp$#$\sp\hfil\vrule\cr
416: \noalign{\hrule}\omit\cr
417: &\bv_2\<C_I> &\bv_2\a &\bv_2[C_i]\\
418: \bv_2\<C_K> &0 &0 &\sum_{k\in K}\d[ik]\\
419: \bv_2\b &0 &\aug(\a\c\b) &\aug_i(\b\c\b)\\
420: \bv_2[C_k] &\sum_{i\in I}\d[ik]&\aug_k(\a\c\a) &\d[ik]\chi(C_i)\\
421: }}\hfill}
422: \endinsert
423:
424: The Bockstein homomorphism $\Sq_1\:\iH_2(X)\to\iH_1(X)$ is given by
425: \ref{C1.7.C}: one has $\Sq_1\bv_2[C_i]=\bv_1w_1(C_i)$ and
426: $\Sq_1\bv_2(\Ga+\<C_I>)=\bv_1\Ga$.
427:
428: We conclude this section with a few examples of
429: calculation of the filtration~$\CF^*$ and Viro homomorphisms~$\vr_*$. The
430: first example may serve as a model for regular complete intersections
431: in~$\Cp{n}$ (see \cite{Kalinin}), the second one, as a local model for
432: involutions on smooth manifolds.
433:
434: \claim[Filtrations in $\Rp{n}$ \ROM(see~\cite{Kalinin})]\label{C1.7.E}
435: Let $X=\Cp{n}$ and $c=\conj$. Then the filtrations~$\CF^*$ on
436: $H_*(\Rp{n})$ and $\CF_*$ on $H^*(\Rp{n})$ are given by
437: $$
438: \CF^r=\Sq H_{\ge{r/2}}(\Rp{n}),\qquad \CF_r=\Sq H^{\le{r/2}}(\Rp{n}),
439: $$
440: where $\Sq$ stands for both homology and cohomology total Steenrod
441: squares.
442: \qed
443: \endclaim
444:
445: \claim[Filtration in Thom spaces\/ \ROM(see~\cite{DegtKh:2})]\label{C1.7.F}
446: Let $\nu$ be an $m$-dimensional vector bundle over a finite cell
447: complex~$F$, and let~$T$ and~$\partial T$ be the associated disk and
448: sphere bundles, respectively, supplied with the antipodal involution.
449: Then the filtrations~$\CF^*$ on $H_*(F)$ and~$\CF_*$ on $H^*(F)$
450: associated with Kalinin's spectral sequence of pair~$(T,\partial T)$ are
451: given by
452: $$
453: \CF^{p}=w(\nu)^{-1}\capprod H_{\ge p-m}(F),\qquad
454: \CF^{p}=w(\nu)\cupprod H^{\le p-m}(F),
455: $$
456: where $w(\nu)=1+w_1(\nu)+w_2(\nu)+\dots$ is the total Stiefel-Whitney
457: class of~$\nu$. The Viro homomorphisms are
458: $$
459: \gather
460: \vr_p\:\CF^p\to H_p(T,\partial T),\quad
461: x\mapsto\th^{-1}[w(\nu)\capprod x]_{p-m},\\
462: \vr^p\: H^p(T,\partial T)\to H^*(F)/\CF_{p-1},\quad
463: x\mapsto w(\nu)\cupprod\th^{-1}x,
464: \endgather
465: $$
466: where $\th$ is the Thom isomorphism and $[x]_r$ denotes the
467: $r$-dimensional component of a nonhomogeneous homology class~$x$.
468: \qed
469: \endclaim
470:
471: Next statement is a generalization of~\ref{C1.7.E};
472: it can be deduced from~\ref{C1.7.E}
473: using the
474: splitting
475: principle.
476:
477: \claim[Filtrations in Grassmann manifolds]\label{gras}
478: Let $X=G_k(\C^n)$ be the Grassmann manifold of $k$-planes in~$\C^n$
479: and $c=\conj$ the standard conjugation. Then the
480: filtrations~$\CF^*$ on $H_*(G_k(\R^n))$ and $\CF_*$ on $H^*(G_k(\R^n))$
481: are given by
482: $$
483: \CF^r=\Sq H_{\ge{r/2}}(G_k(\R^n)),\qquad \CF_r=\Sq H^{\le{r/2}}(G_k(\R^n)),
484: $$
485: where $\Sq$ stands for both homology and cohomology total Steenrod
486: squares.
487: \qed
488: \endclaim
489:
490: The following consequence of~\ref{gras} is, in fact, equivalent to it.
491: A related identity is found in~\cite{Kr:94}.
492:
493: \claim[Corollary]\label{ChtoWh}
494: Let $X$ be a topological space with involution $c\:X\to X$ and $E$ a real
495: vector bundle over~$X$ in the sense of Atiyah \cite{At:Kreal} \rom(i.e.,
496: $E$ is a $\C$-vector bundle supplied with an anti-linear involutive
497: lift~$c_E$ of~$c$\rom). Then for all~$i$ one has $\bv^{2i}c_i(E)=\Sq
498: w^i(\R E)$, where $\R E=\Fix c_E$ is an $\R$-vector bundle over~$\Fix c$.
499: \qed
500: \endclaim
501:
502:
503: \subsection{Involutions on manifolds}\label{involutions}
504: Let $X$ be a smooth closed oriented $2k$-manifold and $c$ a smooth
505: involution, preserving the orientation if $k$ is even and reversing it
506: otherwise. Define the \emph{twisted intersection form}
507: $$
508: B^c\:(H_{k}(X;\Z)/\Tors)\otimes(H_{k}(X;\Z)/\Tors)\to\Z,\quad
509: B^c(x,y)=x\ii c_*y.
510: $$
511: This form is symmetric; its signature~$\Gs(c)$ is called the
512: \emph{signature of the involution}. If $k$ is odd, the signature is $0$,
513: otherwise $\Gs(c)=\Gs_+-\Gs_-$, where $\Gs_\pm$ are the signatures of the
514: restrictions of the intersection form to the eigenspaces $H^{\pm1}_{k}$
515: of~$c_*$.
516:
517: \theorem\label{2.1.1}
518: If all the components of~$F=\Fix c$ are of dimension~$k$, one has
519: $\Gs(c)=F\ii F$, where $F\ii F$ is the normal Euler number of~$F$ in~$X$.
520: \pni%[AS]
521: \endtheorem
522:
523: Since the real part $\RX$ of a complex manifold $X$ with real structure
524: is totally real, all the components of~$\RX$ are of dimension
525: $n=\dim_{\C}X$ and $\RX\ii\RX=(-1)^{n/2}\chi(\RX)$. If $X$ is supplied
526: with an invariant K\"ahler metric, the real part is Lagrangian.
527:
528: \claim[Corollary]\label{Hst}
529: If $X$ is a $2k$-dimensional compact complex manifold with real structure
530: $\conj$, one has $\Gs(\conj)=(-1)^k\chi(\RX)$.
531: \qed
532: \endclaim
533:
534: If $X$ is a closed compact (not necessarily orientable) manifold and
535: $\dim X=2k$ is even, one can define a $\ZZ$-version $B^c\:H_k(X)\otimes
536: H_k(X)\to\ZZ$ of the twisted intersection form. It is also symmetric.
537:
538: \claim[Arnol$'$d lemma]\label{Arnold.lemma}
539: Let $X$ be a closed smooth manifold with involution~$c$ and $\dim X=2k$
540: even. If $\dim\Fix c\le k$, then the fundamental class~$[\Fix c]_k$ of
541: the union of the $k$-dimensional components of~$\Fix c$ realizes
542: in~$H_k(X)$ the characteristic class of $B^c\:H_k(X)\otimes H_k(X)\to\ZZ$.
543: \pni%[Arn]
544: \endclaim
545:
546: \remark{Remark}
547: In general the characteristic class of~$B^c$ is the $k$-dimensional
548: component of $\inj_!(v(\tau)\,v^{-1}(\nu))$, where $\inj\:\Fix c\to X$ is
549: the inclusion, $\tau$ and~$\nu$ are, respectively, the tangent and normal
550: bundles of~$\Fix c$, and $v$ stands for the total Wu class. See~\cite{CM}
551: for details.
552: \endremark
553:
554:
555: \end
556:
557:
558: %%% Local Variables:
559: %%% mode: latex
560: %%% TeX-master: t
561: %%% End:
562: