1: \documentclass[12pt,reqno,draft]{amsart}
2:
3: \textwidth=37pc %+6
4: \hoffset=-4pc %-3
5:
6: %\usepackage{euscript}
7: %\usepackage{eufrak}
8: %\usepackage{verbatim}
9: \usepackage{amscd,amsmath}
10: \usepackage{amssymb}
11: \usepackage{amsthm}
12: \usepackage{enumerate}
13:
14: \theoremstyle{plain}
15: \newtheorem{theorem}{Theorem}[section]
16: \newtheorem{lemma}[theorem]{Lemma}
17: \newtheorem{proposition}[theorem]{Proposition}
18: \newtheorem{corollary}[theorem]{Corollary}
19: \theoremstyle{definition}
20: \newtheorem{remark}[theorem]{Remark}
21: \newtheorem*{remarks}{Remarks}
22: \newtheorem*{conjecture}{Conjecture}
23: \newtheorem{definition}[theorem]{Definition}
24:
25: \def\zb{\overline{Z}_i}
26: \def\db{\overline{\partial}}
27: \def\S{S\mathfrak g^*}
28: \def\C{\mathbb C}
29: \def\R{\mathbb{R}}
30: \def\N{\mathbb{N}}
31: \def\cD{\EuScript{D}}
32: \def\cE{\EuScript{E}}
33: \def\cF{\EuScript{F}}
34: \def\cL{\EuScript{L}}
35: \def\cP{\EuScript{P}}
36: \def\fU{\EuFrak{U}}
37: \def\fP{\EuFrak{P}}
38: \def\bu{\bullet}
39: \def\limsup{\operatornamewithlimits{\overline{\text{\rm{lim}}}}}
40: \DeclareMathOperator{\interior}{int}
41: \DeclareMathOperator{\diam}{diam}
42: \renewcommand{\epsilon}{\varepsilon}
43: \renewcommand{\phi}{\varphi}
44: \DeclareMathSymbol{\varnothing}{\mathord}{AMSb}{"3F}
45: \renewcommand{\emptyset}{\varnothing}
46: %\renewcommand{\baselinestretch}{1.5}
47:
48: %\date{\today}
49: %\date{July 20, 2001}
50:
51: \title{Bar complexes and formality of pull-backs}
52:
53: \author{Steven Lillywhite}
54:
55: %\dedicatory{To whomever, with admiration.}
56:
57: \address{Department of Mathematics, University of Toronto,
58: 100 St. George St., Toronto, Ontario,
59: M5S 3G3} \email{sml@math.toronto.edu}
60:
61:
62: %\thanks
63:
64: \keywords{Bar complexes, differential graded algebras, rational
65: homotopy theory, formality}
66: %multifractal analysis, multifractal rigidity, pointwise dimension}
67: \subjclass{Primary: 57T30, 55P62. Secondary: 55T20}
68:
69:
70: \begin{document}
71:
72: \begin{abstract}
73: We prove a result concerning formality of the
74: pull-back of a fibration. Our approach is to use bar complexes
75: in the category of commutative differential graded algebras.
76: \end{abstract}
77:
78: \maketitle
79:
80: \section{Introduction}
81: In this note, we show that the pull-back of a fibration by a formal
82: map is formal. The fibration is required to be totally non-homologous
83: to zero, and to be a formal map as well. We are here referring to
84: the notion of formality in the setting of rational homotopy theory.
85: This result extends a theorem of Vigu\'e-Poirrier, \cite{VP},
86: where it is proved that the fibre of such a fibration is
87: a formal space. Our proof makes use of bar complexes,
88: which, when
89: we use a normalization due to Chen, become commutative differential
90: graded algebras useful for rational homotopy theory.
91: We conclude with an example which generalizes a result of Baum and Smith,
92: \cite{BS}.
93:
94: \section{Review of rational homotopy theory}
95: In this section we briefly recall some notions from rational
96: homotopy theory. References for this material are numerous and
97: we mention \cite{AP}, \cite{BG}, \cite{Tanre}.
98:
99: We introduce the category of commutative differential graded algebras
100: over a field $k$ of characteristic zero. We assume that all algebras
101: are concentrated in non-negative degrees, have a differential which
102: raises degree by one, and are augmented. Furthermore, we shall assume
103: that $H^0(A)\approx k$ for all algebras $A$. We denote this category
104: by $k\mathcal{CDGA}$ and refer to objects in it as $k$CDGA's. A morphism
105: of $k$CDGA's which induces an isomorphism on cohomology is called a
106: {\it quasi-isomorphism}. There is a notion of homotopy between
107: maps of $k$CDGA's which becomes an equivalence relation on the
108: set of maps from $A_1$ to $A_2$, when the source, $A_1$, is of a special
109: type called a {\it KS complex}, which is basically a free algebra whose
110: differential respects an ordering on the generators.
111:
112: Among the KS complexes in $k\mathcal{CDGA}$
113: is an important class called {\it minimal} algebras, which are
114: essentially characterized by being free with decomposable differential. For
115: every $k$CDGA $A$, there exists a minimal $k$CDGA $\mathcal M$, and
116: a quasi-isomorphism $\mathcal M\to A$. Such a minimal algebra is called
117: a {\it minimal model} of $A$. It is unique up to isomorphism, and furthermore,
118: a map of algebras $f: A_1\to A_2$ determines a map $f:\mathcal M(A_1)
119: \to \mathcal M(A_2)$ which is unique up to homotopy.
120: An algebra $A$ is called {\it formal} if there are quasi-isomorphisms of
121: $k$CDGA's $A\leftarrow \mathcal M(A) \to H(A)$.
122: This is equivalent to demanding that there be a sequence of $k$CDGA
123: quasi-isomorphisms $A\leftarrow A_1\to \dots \leftarrow A_n \to H(A)$.
124:
125:
126:
127:
128:
129:
130: If $X$ is a path-connected topological space, then there is a functor
131: which associates to $X$ a $k$CDGA $A(X)$, known as the Sullivan-de
132: Rham algebra. For a large class of spaces, including simply-connected
133: spaces of finite $\mathbb Q$-type (see definition below),
134: the minimal model of $A(X)$
135: determines the rational homotopy of $X$, $\pi^\bu(X)\otimes\mathbb Q$. We
136: say that a space $X$ is {\it formal} if $A(X)$ is formal. Then for
137: these formal spaces, their rational
138: homotopy is determined by their cohomology algebras.
139:
140: \section{Bar complexes and Eilenberg-Moore theory}
141: In this section we shall discuss the theory of Eilenberg and Moore
142: concerning pull-backs of fibrations.
143: For references, see \cite{Mc}, \cite{Smith}, or \cite{EMo}.
144: For the rest of the paper, we shall assume that all spaces are
145: path-connected and of finite $k$-type (meaning that $H^n(X; k)$
146: is finite-dimensional for all $n\ge 1$). We shall also assume
147: that all fibrations are Serre fibrations.
148:
149: Let us suppose that we have a fibration $F\to E\overset{p}
150: {\to} B$ and a map $f: X\to
151: B$, so that we obtain a pull-back diagram:
152:
153: \begin{equation}\label{diagram}
154: \begin{CD}
155: E_f @>{\tilde f}>> E \\
156: @V{\tilde p}VV @V{p}VV \\
157: X @>{f}>> B
158: \end{CD}
159: \end{equation}
160:
161: \noindent
162: Then the maps $f^*$ and $p^*$ make $A^\bu(X)$ and $A^\bu(E)$
163: (differential graded) modules over $A^\bu(B)$. Let us assume that $B$
164: is simply-connected. Then a theorem of Eilenberg and Moore asserts
165: that there is an isomorphism
166: \begin{equation}
167: \theta: Tor_{A^\bu(B)}(A^\bu(X), A^\bu(E))\to H^\bu(E_f).
168: \end{equation}
169:
170: \noindent
171: We may use the bar resolution to obtain a resolution of, say, $A^\bu(X)$
172: by $A^\bu(B)$-modules. Since we are considering $A^\bu(-)$
173: to be the Sullivan-de Rham complex, we will be using Chen's
174: normalized bar resolution, see \cite{Ch} or \cite{GJP}.
175:
176: More specifically, the bar complex is
177:
178: \begin{equation}
179: B(A^\bu(X), A^\bu(B), A^\bu(E))=\bigoplus_{i=0}^{\infty}
180: A^\bu(X)\otimes_k (sA^\bu(B))^{\otimes i}\otimes_k A^\bu(E)
181: \end{equation}
182:
183: \noindent
184: where the tensor products are over the ground field $k$, and $s$ denotes
185: the suspension functor on graded vector spaces which lowers degree by one.
186: Hence the degree of an element $(\alpha,\omega_1,\dots,\omega_k,\beta)$
187: is: $deg(\alpha)+\sum_{i=1}^k (deg(\omega_i)-1)+deg(\beta)$,
188: where $\alpha\in A^\bu(X)$, $\omega_i\in A^\bu(B)$, and $\beta\in
189: A^\bu(E)$. Actually,
190: the bar complex is bigraded. We introduce the {\it bar degree}, denoted
191: $B(A^\bu(X), A^\bu(B), A^\bu(E))_{\bu}$. The bar degree of an element
192: $(\alpha,\omega_1,\dots,\omega_k,\beta)$ is defined to be $-k$. The
193: other grading is the normal tensor product grading, the degree of an
194: element $(\alpha,\omega_1,\dots,\omega_k,\beta)$ being
195: $deg(\alpha)+\sum_{i=1}^k deg(\omega_i)+deg(\beta)$.
196:
197:
198: There are two differentials of total degree +1:
199:
200: \begin{align}
201: d(\alpha,\omega_1,\dots,\omega_k,\beta)&=(d\alpha, \omega_1,\dots,\omega_k,
202: \beta)\\\notag
203: &+\sum_{i=1}^k(-1)^{\epsilon_{i-1}+1}
204: (\alpha,\omega_1,\dots,\omega_{i-1},d\omega_i,\omega_{i+1},\dots,\omega_k,\beta
205: )\\\notag
206: &+(-1)^{\epsilon_k}(\alpha,\omega_1,\dots,\omega_k,d\beta)\\
207: -\delta(\alpha,\omega_1,\dots,\omega_k,n)&=(-1)^{\epsilon_0}
208: (\alpha\omega_1,\omega_2,\dots,\omega_k,\beta)\\\notag
209: &+\sum_{i=1}^{k-1}(-1)^{\epsilon_i}(\alpha,\omega_1,\dots,
210: \omega_{i-1},\omega_i\omega_{i+1},\omega_{i+2},\dots,\omega_k,\beta)\\\notag
211: &+(-1)^{\epsilon_{k-1}+1}(\alpha,\omega_1,\dots,\omega_{k-1},\omega_k\beta)
212: \end{align}
213:
214: \noindent
215: where
216: $\epsilon_i=deg\alpha+
217: deg\omega_1+\dots +deg\omega_i-i$. The differential $\delta$ has
218: degree $+1$ with respect to the bar grading, while the differential
219: $d$ has degree $+1$ with respect to the tensor product grading.
220: One may verify that
221: $d\delta+\delta d=0$, and we put $D\overset{def}{=}
222: d+\delta$ to be the total differential.
223: With the given bigrading,
224: we get a double complex with the two differentials $d$ and $\delta$.
225: If we filter the bar complex so that we take $d$-cohomology first
226: in the associated spectral sequence, then we obtain the spectral
227: sequence of Eilenberg and Moore.
228:
229: Chen's normalized version of this bar complex is the following.
230: If $f\in A^0(B)$, let $S_i(f)$ be the operator on
231: $B(A^\bu(X),A^{\bu}(B),A^\bu(E))$ defined by
232: \begin{equation}
233: S_i(f)(\alpha,\omega_1,\dots,
234: \omega_k,\beta)
235: =(\alpha,\omega_1,\dots,\omega_{i-1},f,\omega_i,\dots,\omega_k,\beta)
236: \end{equation}
237: for
238: $1\le i\le k+1$. Let $W$ be the subspace of $B(A^\bu(X),A^{\bu}(B),A^\bu(E))$
239: generated by the images of $S_i(f)$ and $DS_i(f)-S_i(f)D$. Then define
240:
241: \begin{equation}
242: \bar B(A^\bu(X),A^{\bu}(B),A^\bu(E))\overset{def}{=}
243: B(A^\bu(X),A^{\bu}(B),A^\bu(E))/W.
244: \end{equation}
245:
246: \noindent
247: Then $W$ is closed under $D$ and when $H^0(B)=k$ (B is connected), then
248: $W$ is acyclic so that $\bar B(A^\bu(X), A^\bu(B), A^\bu(E))$
249: is quasi-isomorphic
250: to $B(A^\bu(X), A^\bu(B), A^\bu(E))$. Notice that in the normalized
251: bar complex, there are no elements of negative degree, and
252: with our assumption that $B$ is simply-connected, we are
253: assured convergence of the associated Eilenberg-Moore spectral sequence.
254: The map $\theta$
255: mentioned above is
256: induced by the map
257: \begin{equation}
258: \theta: B(A^\bu(X), A^\bu(B), A^\bu(E))\to A^\bu(E_f)
259: \end{equation}
260: which sends all tensor products to zero except for
261: $A^\bu(X)\otimes_k A^\bu(E)$, where the map is: $\alpha\otimes\beta\mapsto
262: \tilde p^*\alpha\wedge \tilde f^*\beta$. Note that $\theta(W)=0$, so that
263: we get an induced map
264: \begin{equation}
265: \theta: \bar B(A^\bu(X), A^\bu(B), A^\bu(E))\to
266: A^\bu(E_f).
267: \end{equation}
268: The bar complex computes Tor, and the theorem of Eilenberg and Moore
269: states that this map $\theta$ is a quasi-isomorphism.
270:
271: The normalized bar complex may also be augmented. The augmentation,
272: $\epsilon$, maps all elements of positive total degree to zero. The
273: elements of degree zero have the form $(f,g)$, where $f\in A^0(X)$
274: and $g\in A^0(E)$. Then we define $\epsilon(f,g)=\epsilon_X(f)\epsilon_E(g)
275: =f(x_0)g(e_0)$,
276: where $x_0$ and $e_0$ are chosen base-points in $X$ and $E$, respectively,
277: and $\epsilon_X, \epsilon_E$ are the augmentations of
278: $A^\bu(X), A^\bu(E)$, respectively.
279: If we choose base-points so that the pull-back diagram above preserves
280: all base-points, then $\theta$ is an augmentation preserving map.
281:
282: The bar complex has a natural coalgebra structure. Since we are
283: inputting $k$CDGA's to the bar complex, we also obtain a commutative
284: differential graded algebra structure
285: on the bar complex via the shuffle product.
286:
287: More specifically, if $(a_1, \dots, a_p)$
288: and $(b_1, \dots, b_q)$ are two ordered sets, then a {\it shuffle}
289: $\sigma$ of $(a_1, \dots, a_p)$ with $(b_1, \dots, b_q)$ is a permutation
290: of the ordered set $(a_1, \dots, a_p, b_1, \dots, b_q)$ which preserves
291: the order of the $a_i$'s as well as the order of the $b_j$'s. That is,
292: we demand that if $i<j$, then $\sigma(a_i)<\sigma(a_j)$ and
293: $\sigma(b_i)<\sigma(b_j)$.
294:
295: We obtain a product on $\bar B(A^\bu(X), A^\bu(B), A^\bu(E))$
296: by first taking the normal tensor product on the $A^\bu(X)\otimes A^\bu(E)$
297: factors, then taking the tensor product of this product with the
298: shuffle product on the $A^\bu(B)^{\otimes i}$ factors. As usual,
299: we introduce a sign $(-1)^{|\alpha||\beta|}$ whenever $\alpha$
300: is moved past $\beta$. The actual formula for the product is the
301: following. Let $a, x\in A^\bu(X)$, $b_i, y_i\in A^\bu(B)$, and
302: $c, z\in A^\bu(E)$.
303:
304: \begin{equation}
305: (a, b_1,\dots, b_k, c)\bu(x, y_1,\dots, y_l, z)=
306: \sum_{\sigma}(-1)^{\eta +n_\sigma}(ax, \sigma(b_1,\dots,b_k;y_1,\dots,
307: y_l), cz)
308: \end{equation}
309: where $\eta=|c||x|+|x|\{|b_1|+\dots +|b_k|-k\}+|c|\{|y_1|+\dots +|y_l|-l\}$,
310: the sum is over all shuffles $\sigma$ of the set
311: $(b_1,\dots, b_k)$ with the set $(y_1, \dots, y_l)$,
312: and
313: $$n_\sigma=\sum_{(i,j)}(|b_i|-1)(|y_j|-1)$$
314: where the sum is over all pairs
315: $(i,j)$ such that $b_i$ is moved past $y_j$ in the shuffle $\sigma$.
316: One may check that this product is associative, graded commutative, and
317: that the differential $D$ is a derivation with respect to this product.
318:
319: \begin{lemma}
320: The product defined above induces a product on Chen's normalized bar
321: complex.
322: \end{lemma}
323:
324: \begin{proof}
325: We have that $\bar B(A^\bu(X), A^\bu(B), A^\bu(E))=
326: B(A^\bu(X), A^\bu(B), A^\bu(E))/W$.
327: We will show that $W$ is an ideal. Now $W$ is generated by
328: the images of $S_i(f)$ and $DS_i(f)-S_i(f)D$. It is obvious that
329: $\alpha\bu S_i(f)\beta\in W$ for any $\alpha, \beta$. Moreover,
330: $\alpha\bu (DS_i(f)\beta-S_i(f)D\beta)=\alpha\bu DS_i(f)\beta
331: -\alpha\bu S_i(f)D\beta$. Now $\alpha\bu S_i(f)D\beta\in W$ as we
332: just noted. Moreover, since $D$ is a derivation, we have
333: \begin{equation}\label{der}
334: D(\alpha\bu S_i(f)\beta)= D\alpha\bu S_i(f)\beta+ (-1)^{|\alpha|}
335: \alpha\bu DS_i(f)\beta
336: \end{equation}
337: Now, $\alpha\bu S_i(f)\beta\in W$, and $W$ is closed under $D$,
338: so the left-hand side of \ref{der} is in $W$. Also,
339: $D\alpha\bu S_i(f)\beta\in W$. Hence, $\alpha\bu DS_i(f)\beta\in W$
340: as well.
341:
342: \end{proof}
343:
344:
345:
346: We have arrived at the following lemma.
347:
348: \begin{lemma}\label{applemma1}
349: Assume that we have the pull-back diagram \ref{diagram},
350: where $p$ is a fibration and $B$ is simply connected. Then the normalized
351: bar complex
352: $$
353: \bar B(A^\bu(X), A^\bu(B), A^\bu(E))
354: $$
355: is a $k$CDGA.
356: Moreover,
357: $$\theta : \bar B(A^\bu(X), A^\bu(B), A^\bu(E))\to A^\bu(E_f)$$
358: is a quasi-isomorphism of $k$CDGA's.
359: \end{lemma}
360:
361:
362: \begin{remark}
363: We note that Chen's normalization is functorial. That is, if we have
364: a commutative diagram of $k$CDGA's
365: \begin{equation}
366: \begin{CD}
367: A_2 @<<< B_2 @>>> C_2 \\
368: @AAA @AAA @AAA \\
369: A_1 @<<< B_1 @>>> C_1
370: \end{CD}
371: \end{equation}
372: then we get a map of $k$CDGA's $\bar B(A_1, B_1, C_1)\to
373: \bar B(A_2, B_2, C_2)$.
374: \end{remark}
375:
376: The next lemma concerns quasi-isomorphisms of bar complexes.
377: The main idea of the proof may be found in \cite{VP}, Lemme 4.3.3, and
378: so we omit the proof here.
379:
380: \begin{lemma}\label{barquasi}
381: Suppose that $A_1\leftarrow B_1\to C_1$ and $A_2\leftarrow B_2\to C_2$
382: are two sequences of maps of $k$CDGA's with $H^1(B_1)=0=H^1(B_2)$.
383: Suppose further
384: that $B_1$ is a KS-complex, and that we have a homotopy commutative diagram
385: of the form
386:
387: \begin{equation}
388: \begin{CD}
389: A_2 @<<< B_2 @>>> C_2 \\
390: @AAA @AAA @AAA \\
391: A_1 @<<< B_1 @>>> C_1
392: \end{CD}
393: \end{equation}
394:
395: \noindent
396: where the vertical arrows are all $k$CDGA quasi-isomorphisms.
397: Then $\bar B(A_1, B_1, C_1)$ is quasi-isomorphic to
398: $\bar B(A_2, B_2, C_2)$ in $k$CDGA (via a sequence of $k$CDGA
399: quasi-isomorphisms).
400:
401:
402: \end{lemma}
403:
404:
405:
406:
407:
408: \section{Formality of pull-backs}
409: We can use the normalized bar complex to extend a result of Vigu\'e
410: -Poirrier concerning formality of the fiber of a fibration,
411: \cite{VP}, Th\'eor\`eme 4.4.4.
412: Our proof is also shorter and more direct than in \cite{VP},
413: which deals with more general considerations.
414:
415:
416:
417:
418: \begin{definition} Suppose that $A\overset{f}{\leftarrow}B
419: \overset{g}{\to} C$ are morphisms of $k$CDGA's. Then we shall
420: say that $f$ and $g$ are {\it compatibly formal} if there exists a
421: homotopy commutative diagram
422:
423: \begin{equation}
424: \begin{CD}
425: A @<{f}<< B @>{g}>> C\\
426: @AAA @AAA @AAA \\
427: \mathcal M(A) @<<< \mathcal M(B) @>>> \mathcal M(C) \\
428: @VVV @VVV @VVV \\
429: H(A) @<{f}<< H(B) @>{g}>> H(C)
430: \end{CD}
431: \end{equation}
432: where the middle row are minimal models for $A, B$, and $C$, and the
433: vertical arrows are quasi-isomorphisms. We shall say that maps,
434: $f, g$, of spaces $X\overset{f}{\to} Y \overset{g}{\leftarrow} Z$
435: are {\it compatibly formal} if the the corresponding maps
436: $A(X)\overset{f}{\leftarrow} A(Y) \overset{g}{\to} A(Z)$ are
437: compatibly formal.
438: \end{definition}
439:
440: Consider again the pull-back diagram \ref{diagram}
441: where $p$ is a fibration with fiber $F$, and $B$ is simply-connected.
442:
443: \begin{theorem}\label{appprop}
444: Assume that $p$ and $f$ are compatibly
445: formal maps,
446: and suppose that the Serre spectral
447: sequence for the fibration $p$ degenerates at the $E_2$ term. Then
448: the pull-back $E_f$ is formal.
449: \end{theorem}
450:
451: \begin{proof}
452: By \ref{applemma1} we have a quasi-isomorphism of $k$CDGA's
453: \begin{equation}
454: \theta : \bar B(A^\bu(X), A^\bu(B), A^\bu(E))\to A^\bu(E_f)
455: \end{equation}
456:
457: By the assumption of compatible formality, we have
458: a homotopy commutative diagram whose vertical arrows are quasi-isomorphisms
459:
460: \begin{equation}\label{good}
461: \begin{CD}
462: A^\bu(X) @<<< A^\bu(B) @>>> A^\bu(E)\\
463: @AAA @AAA @AAA\\
464: \mathcal M(X) @<<< \mathcal M(B) @>>> \mathcal M(E) \\
465: @VVV @VVV @VVV\\
466: H^\bu(X) @<<< H^\bu(B) @>>> H^\bu(E)
467: \end{CD}
468: \end{equation}
469:
470: \noindent
471: Then we obtain a sequence of $k$CDGA quasi-isomorphisms
472: \begin{equation}
473: \bar B(A^\bu(X), A^\bu(B), A^\bu(E))\leftarrow \dots
474: \to \bar B(H^\bu(X), H^\bu(B), H^\bu(E))
475: \end{equation}
476: by \ref{barquasi}. Now the bar complex
477: $$\bar B(H^\bu(X), H^\bu(B), H^\bu(E))$$ \noindent
478: has only a single differential,
479: $\delta$. Since we have assumed the Serre spectral sequence for the
480: fibration $p$ to degenerate at the $E_2$ term, it follows that
481: $H^\bu(E)$ is a free $H^\bu(B)$-module. Thus if we grade according
482: to the bar degree for $\delta$, we find that
483:
484: \begin{enumerate}
485: \item $H_+(\bar B(H^\bu(X), H^\bu(B), H^\bu(E))_\bu)=0$
486: \item $H_0(\bar B(H^\bu(X), H^\bu(B), H^\bu(E))_\bu)\approx
487: H_\bu(\bar B(H^\bu(X), H^\bu(B), H^\bu(E))_\bu$
488: \end{enumerate}
489:
490: \noindent
491: Hence the projection to cohomology
492: \begin{align}
493: \bar B(H^\bu(X), H^\bu(B), H^\bu(E))_{\bu}&\to
494: \bar B(H^\bu(X), H^\bu(B), H^\bu(E))_0\notag\\
495: &\to H_\bu(\bar B(H^\bu(X), H^\bu(B), H^\bu(E))_\bu\notag\\
496: &\approx H^\bu(E_f)\notag
497: \end{align}
498: is a $k$CDGA quasi-isomorphism and $E_f$ is consequently formal.
499:
500: \end{proof}
501:
502: \section{An Example}
503: Suppose that $B$ is a simply-connected space with the property that
504: the cohomology of $B$ is a free $k$CDGA. Then $B$ is a formal
505: space, and the cohomology, $H^\bu(B)$, is a minimal model for $A^\bu(B)$.
506: Let $E\overset{p}{\to} B$ be a fibration with $E$ a formal space,
507: and let $X$ be a formal space with a map to $B$, $f: X\to B$.
508: Then it is easy to see that $f$ and $p$ are compatibly formal.
509:
510: Let $E_f$ be the pull-back of the fibration $p$ by the map $f$, as
511: in diagram \ref{diagram}.
512: Then \ref{applemma1} and \ref{barquasi} imply that $H^\bu(E_f)$ and
513: $Tor_{H^\bu(B)}(H^\bu(X), H^\bu(E))$ are isomorphic as algebras.
514: This extends a result of Baum and Smith, \cite{BS}, where
515: this was proven by other means
516: for $X$ a compact globally symmetric space, and
517: $E=BH$, $B=BG$, for $G$ a compact, connected Lie group, and
518: $H\subset G$ a closed, connected subgroup.
519: Moreover, if the Serre spectral sequence for the fibration
520: $E\overset{p}{\to} B$ degenerates at the $E_2$ term, then by
521: \ref{appprop}, we have that $E$ is a formal space.
522:
523:
524:
525:
526:
527:
528: \bibliographystyle{amsplain}
529: \bibliography{master}
530:
531:
532: \end{document}
533:
534: %% END
535:
536:
537:
538:
539:
540:
541:
542:
543:
544:
545:
546:
547:
548:
549:
550:
551:
552:
553:
554:
555:
556:
557:
558:
559:
560:
561:
562:
563:
564:
565:
566:
567: