1: \newcommand{\bull}{\vrule height .9ex width .8ex depth -.1ex}
2: % square bullet
3: \newcommand{\ppp}{\hfill $\bull$ }
4: \documentclass[11pt]{article}
5: \usepackage{epsfig}
6: \author{ M.-L. Labbi\thanks{
7: Address: Department of Mathematics, College
8: of Science, University of Bahrain, Isa Town 32038 Bahrain.
9: E-mail: labbi@sci.uob.bh }}
10: \title{Manifolds with positive second H. Weyl curvature invariant}
11: \date{}
12: \usepackage{amsmath}
13: \usepackage{amscd}
14: \newtheorem{theorem}{Theorem}[section]
15: \newtheorem{corollary}[theorem]{Corollary}
16: \newtheorem{lemma}[theorem]{Lemma}
17: \newtheorem{example}[theorem]{Example}
18: \newtheorem{remark}[theorem]{Remark}
19: \newtheorem{proposition}[theorem]{Proposition}
20: \newtheorem{definition}{Definition}
21: \begin{document}
22: \maketitle
23: \begin{abstract} The second H. Weyl curvature invariant of a Riemannian
24: manifold, denoted $h_4$, is the
25: second curvature invariant which appears in the well known tube formula
26: of H. Weyl. It coincides with the Gauss-Bonnet integrand in dimension 4.
27: A crucial property of $h_4$ is that it is nonnegative
28: for Einstein manifolds, hence it provides a geometric obstruction
29: to the existence of Einstein metrics in dimensions $\geq 4$, independently from the
30: sign of the Einstein constant.
31: This motivates our study of the positivity of this invariant.
32: Here in this paper we prove many constructions of metrics with positive
33: second H. Weyl curvature invariant, generalizing similar well known results
34: for
35: the scalar curvature.
36: \end{abstract}
37: \par\bigskip\noindent
38: {\bf Mathematics Subject Classification (2000).} 53C21, 53B20.
39: \par\medskip\noindent
40: {\bf Keywords.} H. Weyl curvature invariants, Einstein manifold, surgery.
41: \par
42: \section{Introduction and statement of the results}
43: Let $(M,g)$ be a smooth Riemannian manifold of dimension $n\geq 4$. Let
44: $R$, $cR$ and $c^2R$ denote respectively the Riemann curvature tensor,
45: Ricci tensor and the scalar curvature of $(M,g)$. The {\bf {\sl second
46: Hermann Weyl curvature invariant}}, which throughout this paper
47: shall be written in abridged form as {\bf {shwci}} and denoted by $h_4$,
48: can be defined by
49: \begin{equation*}
50: h_4=\|R\|^2-\|cR\|^2+{1\over 4}\|c^2R\|^2
51: \end{equation*}
52: \par
53: A crucial property of $h_4$ is that it is nonnegative for Einstein manifolds
54: (see section 3 below), and so it provides a new geometric obstruction to the
55: existence
56: of Einstein metrics independently from the sign of the Einstein constant.
57: In particular, the manifolds which do not admit any
58: metric with positive {\sl shwci} cannot admit any Einstein metric.\par
59: Recall that in dimensions greater than 4, we do not know any topological
60: restriction for a manifold to be Einstein. If one requires that the Einstein
61: constant to be positive, then one has two geometric obstructions
62: $cR>0$ and $c^2R>0$. \par\medskip\noindent
63: It would be then with a great benefit to have a classification of
64: manifolds with positive {\sl shwci}.\par\medskip
65: Here in this paper, we shall inaugurate the study of the positivity
66: properties
67: of this important invariant.\par
68: In section 2, we introduce and study in general the Hermann Weyl curvature
69: invariants which
70: appear in the tube formula. Many examples are included.\par
71: In section 3, we study separately the case of the second invariant.
72: We prove that
73: it is nonnegative for Einstein manifolds and nonpositive for conformally
74: flat manifolds with zero scalar curvature. The limit cases are also discussed.
75: \par Then we prove theorem A which states
76: that positive (resp. nonnegative) $p$-curvature implies positive (resp.
77: nonnegative) {\sl shwci}
78: where $p=[(n+1)/2]$. In particular, positive (resp. nonnegative) sectional
79: curvature implies
80: positive (resp. nonnegative) {\sl shwci}. Also if $n\geq 8$, positive
81: (resp. nonnegative) isotropic curvature implies
82: positive (resp. nonnegative) {\sl shwci}.\par
83: Then one can apply our previous constructions in the class of manifolds with positive
84: $p$-curvature (see \cite{Lab2,Lab3,Lab4}) to get many examples of metrics with positive shwci.\par
85: In section 4 we prove the following useful theorem. It generalizes a similar result for
86: the scalar curvature.:\par\medskip\noindent
87: {\bf Theorem B.} {\sl Suppose that the total space $M$ of a Riemannian submersion
88: is compact
89: and the fibers ( with the induced metric)
90: are with positive {\sl shwci} then the manifold M admits a
91: Riemannian metric with positive {\sl shwci}.}
92: \par\medskip\noindent
93: The section is then ended with two applications of this theorem.\par\medskip
94: In section 5, we prove the following stability theorem in the class of compact manifolds with positive
95: {\sl shwci}.:
96: \par\medskip\noindent
97: {\bf Theorem C.} {\sl If a manifold $M$ is obtained from a compact manifold $X$ by surgery in codimension
98: $\geq 5$, and $X$ admits a metric of positive {\sl shwci}, then so does $M$.\par
99: In particular, the connected sum of two compact manifolds of dimensions $\geq 5$ and each one is with positive
100: {\sl shwci} admits a metric with positive {\sl shwci}. }
101: \par\medskip\noindent
102: Theorem C generalizes a celebrated theorem of Gromov-Lawson and Schoen-Yau for the scalar curvature.\par\smallskip\noindent
103: As a consequence of the previous theorem we prove that there are no restrictions on the fundamental group of a compact manifold of dimension $\geq 6$ to carry a metric with positive
104: {\sl shwci}.\par\medskip
105: Finally, let us mention that it would be interesting to prove, like in the case of the scalar curvature, that every manifold with
106: nonnegative {\sl shwci} that is not identically zero admits a metric with positive shwci.
107: \section{The H. Weyl curvature invariants}
108: Let
109: $\Lambda^{*}M=\bigoplus_{p\geq 0}\Lambda^{*p}M$ denote the ring of differential
110: forms
111: on $M$, where $M$ is as above. Considering the tensor product over the ring of smooth functions,
112: we define
113: ${\cal D}= \Lambda^{*}M\otimes \Lambda^{*}M=\bigoplus_{p,q\geq 0}
114: {\cal D}^{p,q}$ where $ {\cal D}^{p,q}= \Lambda^{*p}M \otimes
115: \Lambda^{*q}M$.
116: It is graded associative ring and called the ring of
117: double forms on $M$. \par
118: The ring of curvature structures on $M$ (\cite{Kulk}) is the ring ${\cal C}=\sum_{p\geq 0}
119: {\cal C}^p$ where ${\cal C}^p$ denotes symmetric elements in
120: ${\cal D}^{p,p}$. We denote by ${\cal C}_1$ (resp. ${\cal C}_2$, ${\cal C}_0$) the subring
121: of curvature structures satisfying the first (resp. the second, both the first and second) Bianchi identity. \par
122: The standard inner product and the Hodge star operator $*$ on $\Lambda^{*p}M$ can be extended in a
123: standard way to ${\cal D}$ and they satisfy the following properties, see \cite{Lab1} for the proof:
124: \begin{equation}\label{formula:a}g\omega=*c*\omega \end{equation}
125: for all $\omega\in {\cal D}$, where c denotes the contraction map.
126: Also for all $\omega_1, \omega_2\in {\cal D}$, we have
127: \begin{equation}\label{formula:b}<g\omega_1,\omega_2>=<\omega_1,c\omega_2>\end{equation}
128: that is the contraction map is the formal adjoint of the multiplication map by the metric
129: $g$. Furthermore, we have for all $\omega_1, \omega_2\in {\cal D}^{p,q}$
130: \begin{equation}\label{formula:c}<\omega_1,\omega_2>=*(\omega_1.*\omega_2)=*(*\omega_1.\omega_2)\end{equation}
131: and
132: \begin{equation}\label{formula:d}**=(-1)^{(p+q)(n-p-q)}Id\end{equation}
133: Where $Id$ is the identity map on ${\cal D}^{p,q}$.\par\medskip\noindent
134: Next, we define the H. Weyl curvature invariants: \par\medskip\noindent
135: {\bf Definition.}
136: {\sl The $2q$-Hermann Weyl curvature invariant, denoted $h_{2q}$, is
137: the complete contraction of the tensor $R^q$, precisely,
138: $$h_{2q}={1\over (2q)!}c^{2q}R^q$$
139: where $R^q$ denotes the multiplication of $R$ with itself $q$-times in the ring ${\cal C}$.
140: }
141: \par\medskip
142: Remark that $h_2={1\over 2}c^2R$ is one half the scalar curvature and if $n$ is even then $h_n$ is
143: (up to a constant) the Gauss-Bonnet integrand.
144: \par\medskip\noindent
145: Note that in \cite{Lab1}, it is proved that
146: \begin{equation}\label{formula:e}h_{2q}=
147: *{1\over (n-2q)!}g^{n-2q}R^q \end{equation}
148:
149: \subsection{Examples}
150: \begin{enumerate}
151: \item Let $(M,g)$ be with constant sectional curvature $\lambda$, then
152: $$R={\lambda\over 2}g^2\qquad {\text and}\qquad R^q={\lambda^q\over 2^q}g^{2q}$$
153: And therefore $h_{2q}$ is constant and equals to
154: $$h_{2q}=*{1\over (n-2q)!}g^{n-2q}R^q=*{\lambda^q\over 2^q(n-2q)!}g^{n}
155: ={\lambda^qn!\over 2^q(n-2q)!}$$
156: In particular,
157: \begin{equation} \label{formula:f}h_4={n(n-1)(n-2)(n-3)\over 4}\lambda^2\end{equation}
158: \par\bigskip\noindent
159: \item Let $(M,g)$ be a Riemannian product of two Riemannian manifolds
160: $(M_1,g_1)$ and $(M_2,g_2)$. If we index by $i$ the invariants of the
161: metric $g_i$ for $i=1,2$, then
162: $$R=R_1+R_2\quad {\rm and}\quad R^q=(R_1+R_2)^q=
163: \sum_{i=0}^qC_i^qR_1^iR_2^{q-i}$$
164: consequently, a starightforward calculation shows that
165: \begin{equation*}
166: \begin{split}
167: h_{2q}&={c^{2q}R^q\over (2q)!}=\sum_{i=0}^qC_i^q {c^{2q}\over (2q)!}
168: (R_1^iR_2^{q-i})\\
169: &=\sum_{i=0}^qC_i^q {c^{2i}R_1^i\over (2i)!}
170: {c^{2q-2i}R_2^{q-i}\over (2q-2i)!}\\
171: &=\sum_{i=0}^qC_i^q (h_{2i})_1 (h_{2q-2i})_2
172: \end{split} \end{equation*}
173: In particular,
174: \begin{equation}\label{formula:g}
175: h_4=(h_4)_1+{1\over 2}scal_1scal_2+(h_4)_2
176: \end{equation}
177: where $scal$ denotes the scalar curvature.
178: \par\bigskip\noindent
179: \item Let $(M,g)$ be a hypersurface of the Euclidean space. If $B$ denotes the second
180: fundamental form at a given point, then the Gauss equation shows that
181: $$R={1\over 2}B^2\qquad {\text and}\quad R^q={1\over 2^q}B^{2q}$$
182: Consequently, if $\lambda_1\leq \lambda_2\leq ...\leq \lambda_n$ denote the eigenvalues
183: of $B$, then the eigenvalues of $R^q$ are ${(2q)!\over 2^q}\lambda_{i_1}
184: \lambda_{i_2}...\lambda_{i_{2q}}$ where $ i_1<...<i_{2q}$.
185: Consequently,
186: $$h_{2q}=
187: {(2q)!\over 2^q}\sum_{1\leq i_1<...<i_{2q} \leq n}\lambda_{i_1}
188: ...\lambda_{i_{2q}}$$
189: So they coincide, up to a constant, with the symmetric functions in the eigenvalues
190: of $B$.\par\noindent
191: \par\medskip\noindent
192: \item Let $(M,g)$ be a conformally flat manifold. Then it is well known that
193: at each point
194: of $M$,
195: the Riemann curvature tensor is determined by a symmetric bilinear form $h$,
196: in the sens that $R=g.h$. Consequently, $R^q=g^qh^q$.\par
197: Let $\{e_1,..., e_n\}$ be an orthonormal basis of eigenvectors of $h$ and
198: $\lambda_1\leq \lambda_2\leq ...\leq \lambda_n$ denote the eigenvalues of $h$.
199: \par
200: Then it is not difficult to see that all the tensors $R^q$ are also
201: diagonalizable by the $2q$-vectors $e_{\scriptstyle i_1}\wedge ...\wedge e_{\scriptstyle i_{2q}}$, $i_1<...<i_{2q}$.
202: Their eigenvalues are of the form
203: $$R^q(e_1\wedge ...\wedge e_{2q},e_1\wedge ...\wedge e_{2q})=
204: (q!)^2\sum_{1\leq i_1<...<i_q \leq 2q}\lambda_{i_1}
205: ...\lambda_{i_q}$$
206: Consequently we get
207: \begin{equation*}
208: h_{2q}={(n-q)!q!\over (n-2q)!}
209: \sum_{1\leq i_1<...<i_q \leq n}\lambda_{i_1}
210: ...\lambda_{i_q}
211: \end{equation*}
212: \item Let $g_t=tg$ for $t>0$. If we index by $t$ the invariants of $g_t$ then
213: $$R_t=tR\qquad {\rm and} \qquad R_t^q=t^qR^q$$
214: and therefore
215: \begin{equation}
216: (h_{2q})_t={1\over t^q}h_{2q}
217: \end{equation}
218: \end{enumerate}
219: \par\medskip\noindent
220: Let us now recall some other useful facts from \cite{Lab1} which shall be used later.\par
221: Following Kulkarni we call the elements in $\ker c\subset D^{p,q}$
222: effective elements of $D^{p,q}$, and shall be denoted by $E^{p,q}$.
223: \par\noindent
224: Recall the following orthogonal decomposition
225: of $D^{p,q}$:
226: \begin{equation}\label{formula:h}
227: D^{p,q}=E^{p,q}\oplus gE^{p-1,q-1}\oplus g^2E^{p-2,q-2}\oplus ...
228: \oplus g^rE^{p-r,q-r}
229: \end{equation}
230: where $r={\min \{p,q\}}$.\par\noindent
231: With respect to the previous decomposition, if $\omega=\sum_{i=0}^p
232: g^{p-i}\omega_i\in {\cal C}_1^p$ and $n=2p$, then (see \cite{Lab1})
233: \begin{equation}\label{formula:i}
234: *\omega=\sum_{i=0}^p (-1)^i
235: g^{p-i}\omega_i
236: \end{equation}
237: Also let us recall the following lemma from \cite{Lab1}:
238: \begin{lemma}\label{lemma:gpq} Let $\omega_1\in E_1^r,\omega_2\in E_1^s$ be effectives then
239: \begin{equation*}
240: \begin{split}
241: <g^p\omega_1,g^q\omega_2>=& 0\quad {\text if}\quad (p\not =q)
242: \quad {\rm or}
243: \quad ( p=q \quad {\rm and}\quad r\not=s)\\
244: <g^p\omega_1,g^p\omega_2>=& p!\bigl(\prod_{i=0}^{p-1}(n-2r-i)\bigr)<\omega_1,\omega_2>
245: \quad {\text if}\quad p\geq 1
246: \quad {\text and}
247: \quad r=s
248: \end{split}
249: \end{equation*}
250: \end{lemma}
251: \section{The second H. Weyl curvature invariant}
252: With respect to the previous orthogonal decomposition \ref{formula:h},
253: the Riemann curvature tensor decomposes to
254: $R=\omega_2+g\omega_1+g^2\omega_0$, where
255: \begin{equation*}
256: \begin{split}
257: \omega_0=&{1\over 2n(n-1)}c^2R\\
258: \omega_1=&{1\over n-2}(cR-{1\over n}gc^2R)\\
259: \end{split}
260: \end{equation*}
261: and $\omega_2$ is the Weyl tensor, it is defined by the previous decomposition
262: of $R$.\par\smallskip
263: Corollary 6.5 in \cite{Lab1} shows that
264: \begin{equation}\label{formula:j}
265: h_4=
266: {1\over (n-4)!}[n!||\omega_0||^2-(n-2)!||\omega_1||^2+(n-4)!||\omega_2||^2]
267: \end{equation}
268: using lemma \ref{lemma:gpq} we can easily check that
269: \begin{equation}
270: \begin{split}
271: \|\omega_2\|^2&=\|R\|^2-{1\over n-2}\|cR\|^2+{1\over 2(n-1)(n-2)}\|c^2R\|^2\\
272: \|\omega_1\|^2&={1\over (n-2)^2}(\|cR\|^2-{1\over n}\|c^2R\|^2)\\
273: \|\omega_0\|^2&={1\over 4n^2(n-1)^2}\|c^2R\|^2\\
274: \end{split}
275: \end{equation}
276: and consequently using formula \ref{formula:j} we obtain another
277: useful expression for $h_4$ as follows:
278: \begin{equation}\label{formula:k}
279: h_4=\|R\|^2-\|cR\|^2+{1\over 4}\|c^2R\|^2
280: \end{equation}
281: The folowing theorem was first proved in \cite{Lab1}.
282: \begin{theorem}
283: Let $(M,g)$ be a Riemannian manifold of dimension $\geq 4$.
284: \begin{enumerate}
285: \item If $(M,g)$ is an Einstein manifold then $h_4\geq 0$. Furthermore $h_4\equiv 0$ if and only if
286: $(M,g)$ is flat.
287: \item
288: If $(M,g)$ is conformally flat with zero scalar curvature then
289: $h_4\leq 0$. Furthermore $h_4\equiv 0$ if and only if
290: $(M,g)$ is flat.
291: \end{enumerate}
292: \end{theorem}
293: {\bf Proof.} If $(M,g)$ is conformally flat then $\omega_2=0$ and then
294: \begin{equation*}
295: \begin{split}
296: h_4=&{1\over (n-4)!}[n!||\omega_0||^2-(n-2)!||\omega_1||^2]\\
297: =&{n-3\over n-2}[{n\over 4(n-1)}||c^2R||^2-||cR||^2]
298: \end{split}
299: \end{equation*}
300: From which is clear that if $c^2R=0$ then $h_4\leq 0$ and $h_4\equiv 0$
301: if and only if the metric
302: is Ricci flat and hence is flat. This proves the first part of the
303: theorem.\par
304: Next, if $(M,g)$ is Einstein then $\omega_1=0$ and hence
305: \begin{equation*}\begin{split}
306: h_4=&{1\over (n-4)!}[n!||\omega_0||^2+(n-4)!||\omega_2||^2]\\
307: =&||R||^2+{n-4\over 4n}(c^2R)^2
308: \end{split}
309: \end{equation*}
310: From which it is clear that $h_4\geq 0$ and $h_4\equiv 0$ if and only if the metric
311: is flat. This completes the proof of the theorem.
312: \ppp
313: \par\medskip\noindent
314: Recall that (see \cite{Lab3,Lab4}) the $p$-curvature of $(M,g)$, denoted $s_p$ for $1\leq p\leq n-2$, is a function defined on the $p$-Grassmanian bundle of the manifold. Its value at a
315: tangent $p$-plane $P$ is the avearge of the sectional curvatures of all 2-planes orthogonal
316: to $P$. In particular $s_0$ is the scalar curvature and $s_{n-2}$ is twice the sectional curvature.\par
317: The following theorem provides a relation between the positivity of the $p$-curvature
318: and the {\sl shwci}.:\par\medskip\noindent
319: {\bf Theorem A.} {\sl
320: Let $(M,g)$ be a Riemannian manifold of dimension $n\geq 4$ and with
321: nonnegative (resp. positive) $p$-curvature such that $p\geq {n\over 2}$,
322: then the {\sl shwci} of $(M,g)$ is nonnegative (resp. positive).
323: Furthermore, it
324: vanishes if and only if the manifold is flat.}
325: \par\medskip\noindent
326: {\bf Proof.} Suppose $n=2(k+2)$ is even, $k\geq 0$.
327: Since
328: $$R=\omega_2+g\omega_1+g^2\omega_0$$
329: then
330: $$g^kR=g^k\omega_2+g^{k+1}\omega_1+g^{k+2}\omega_0$$
331: and by formula \ref{formula:i} we have
332: $$*g^kR=g^k\omega_2-g^{k+1}\omega_1+g^{k+2}\omega_0$$
333: On the other hand since $s_{k+2}\geq0$, then both the tensors $g^kR$ and $*g^kR$ are with positive
334: sectional curvature, hence
335: $$[g^k\omega_2+g^{k+2}\omega_0](e_{i_1},...,e_{i_{k+2}},e_{i_1},...,e_{i_{k+2}})
336: \geq g^{k+1}\omega_1(e_{i_1},...,e_{i_{k+2}},e_{i_1},...,e_{i_{k+2}})$$
337: and
338: $$[g^k\omega_2+g^{k+2}\omega_0](e_{i_1},...,e_{i_{k+2}},e_{i_1},...,e_{i_{k+2}})
339: \geq-g^{k+1}\omega_1(e_{i_1},...,e_{i_{k+2}},e_{i_1},...,e_{i_{k+2}})$$
340: for all orthonormal vectors $e_{i_1},...,e_{i_{k+2}}$,
341: and therefore
342: $$[g^k\omega_2+g^{k+2}\omega_0](e_{i_1},...,e_{i_{k+2}},e_{i_1},...,e_{i_{k+2}})
343: \geq |g^{k+1}\omega_1(e_{i_1},...,e_{i_{k+2}},e_{i_1},...,e_{i_{k+2}})|$$
344: Consequently, using formulas \ref{formula:e} and \ref{formula:c},
345: we get
346: $$h_4=*{1\over (n-4)!}g^{n-4}R^2=*{1\over (2k)!}(g^kR.g^kR)={1\over (2k)!}<g^kR,*g^kR>$$
347: and hence using lemma \ref{lemma:gpq} and considering an orthonormal basis
348: diagonalizing $cR$, we obtain
349: \begin{equation*}
350: \begin{split}
351: (2k)!h_4=&<g^k\omega_2+g^{k+2}\omega_0,g^k\omega_2+g^{k+2}\omega_0>-<g^{k+1}\omega_1,g^{k+1}\omega_1>
352: \\
353: \geq &\sum_{\scriptstyle i_1<...<i_{k+2}} \bigl[(g^k\omega_2+g^{k+2}\omega_0)(e_{i_1},...,e_{i_{k+2}},e_{i_1},...,e_{i_{k+2}})\bigr]^2
354: - ||g^{k+1}\omega_1||^2 \\
355: \geq & \sum_{\scriptstyle i_1<...<i_{k+2}}
356: \bigl[g^{k+1}\omega_1(e_{i_1},...,e_{i_{k+2}},e_{i_1},...,e_{i_{k+2}})\bigr]^2-
357: ||g^{k+1}\omega_1||^2=0
358: \end{split}
359: \end{equation*}
360: The same proof works for strict inequality. Also it is clear that
361: if $s_{k+2}\geq 0$ and $h_4\equiv 0$ then $s_{k+2}\equiv 0$ so that the
362: metric is flat.\par
363: To complete the proof, note that if the dimension of the manifold
364: $n=2p+1\geq 5$ is odd then one can consider the
365: product $M\times S^1$. It is of even dimension $2(p+1)$ and with nonnegative (resp. positive)
366: $(p+1)$-curvature therefore by formula
367: \ref{formula:g} we have
368: $$h_4(M)=h_4(M\times S^1)\geq 0 (\text{ resp. } >0)$$
369: this completes the proof of the theorem.\ppp
370: \begin{corollary}
371: \begin{enumerate}
372: \item A Riemannian manifold of dimension $\geq 4$ and with
373: nonnegative (resp. positive) sectional curvature is with
374: nonnegative (resp. positive) {\sl shwci}.
375: Furthermore, $h_4\equiv 0$
376: if and only if the metric is flat.
377: \item A Riemannian manifold of dimension $\geq 8$ and with
378: nonnegative (resp. positive) isotropic curvature is with
379: nonnegative (resp. positive) {\sl shwci}.
380: Furthermore, $h_4\equiv 0$
381: if and only if the metric is flat.
382: \end{enumerate}
383: \end{corollary}
384: {\bf Proof.} Straightforward since positive sectional curvature implies positive $p$-curvature
385: and positive isotropic curvature implies the positivity of the $p$-curvature for all $p\leq n-4$,
386: see \cite{Lab2}.\ppp
387: \par\smallskip\noindent
388: {\bf Remarks.}
389: \begin{enumerate}
390: \item If the dimension of the manifold is even, say $n=2q$, Hopf conjecture states that if the sectional curvature is positive then so is the Gauss-Bonnet integrand that is $h_{2q}$. Then one can ask the more general question:\par\smallskip
391: {\sl Does positive sectional curvature implies positive $h_{2k}$, for all $2\leq 2k\leq n$?}\par\smallskip
392: This is now true for $k=1,2$ by the previous theorem, and it remains an open question for $k\geq 3$.
393: \item Theorem A generalizes a
394: result of Thorpe \cite{Thorpe} for the dimension $n=4$.
395: \end{enumerate}
396: \par\medskip
397: It results from the previous corollary that Lie groups with a biinvariant metric and normal homogeneous
398: Riemannian manifolds are with nonnegative {\sl shwci}. Furthermore
399: using our previous results on the $p$-curvature \cite{Lab3},
400: \cite{Lab4} and the above theorem we can easily prove the following
401: corollaries:
402: \begin{corollary}
403: \begin{enumerate}
404: \item Let $G$ be a compact connected Lie group with rank $r$ such that $r<[{\dim G+1\over 2}]$ endowed
405: with a biinvariant metric $b$ then $(G,b)$ is with positive {\sl shwci}.\par
406: In particular, if $G$ is simple then it is with positive {\sl shwci}.
407: \item If $G/H$ is a normal homogeneous Riemannian manifold such that the rank r of $G$ satisfies
408: $r<[{\dim(G/H)+1\over 2}]$ then it is with positive {\sl shwci}.
409: \end{enumerate}
410: \end{corollary}
411: \begin{corollary}
412: If a compact manifold $M$ admits a smooth action of a compact connected simple Lie group
413: with rank $r$ satisfying $r>[{\dim M+1\over 2}]$ then it admits a metric with positive {\sl shwci}.
414: \end{corollary}
415: \section{Proof of theorem B}
416: Let $ (M,g) $ and $ \left(B,\check g \right) $ be two Riemannian manifolds,
417: and let $ \pi : (M ,g)\rightarrow (B ,\check g)$ a Riemannian submersion.
418: We define, for every $t \in {\bf R}$, a new Riemannian metric $g_t$
419: on the manifold $M$ by multiplying the metric $g$ by $t^2$ in the vertical
420: directions. Recall that $\forall m\in M$, we have a natural orthogonal
421: decomposition of the tangent space at m
422: $$T_mM={\cal V}_m\oplus {\cal H}_m$$
423: where ${\cal V}_m$ is the tangent to the fiber at m and ${\cal H}_m$
424: is the horizontal space, so that
425: \begin{eqnarray*}
426: g_t \mid {\cal V}_m &=&t^2g\\
427: g_t \mid {\cal H}_m &=& g\\
428: g_t({\cal V}_m, {\cal H}_m)&=&0
429: \end{eqnarray*}
430: Note that in this case, $\pi :
431: (M,g_t)
432: \rightarrow (B,\check g)$ is still a Riemannian submersion with the same
433: horizontal and vertical distributions (see \cite{Besse}, \cite{Lab3}).\par \medskip
434: \noindent
435: In the following we shall index by $t$ all the invariants of the metric
436: $g_t$, and in the case case
437: $t=1$ we omitt the index 1.\par
438: Also, We make under a hat `` $\hat {} $ '' (resp. under a
439: check `` $ \check {} $ ''
440: ) the invariants of the fibers with the induced metric
441: (resp. of the basis $B$ ).\par\medskip\noindent
442: Using lemma 2.1 in \cite{Lab3} it is easy to show that for all $g_t$-unit tangent vectors
443: $e_1,e_2,e_3,e_4$ we have
444: $$R_t(e_1,e_2,e_3,e_4)=O({1\over t})\qquad
445: \text {if one of these vecotors is horizontal}$$
446: and that
447: $$R_t(e_1,e_2,e_3,e_4)={1\over t^2}{\hat R}(te_1,te_2,te_3,te_4)+O(1)\qquad
448: {\text{ if the four vectors are vertical}}$$
449: Consequently, if $\{e_1,e_2,...,e_n\}$ is a $g_t$-orthonormal basis such that $\{e_1,...,e_q\}
450: \in {\cal V}_m$ and $\{e_{q+1},...,e_n\}
451: \in {\cal H}_m$, then
452: \begin{equation*}
453: \begin{split}
454: (||R_t||_t)^2=&\sum_{1\leq i<j\leq n,1\leq k<l\leq n}[R_t(e_i,e_j,e_k,e_l)]^2\\
455: =& {1\over t^4}\sum_{1\leq i<j\leq q,1\leq k<l\leq q}[{\hat R}(te_i,te_j,te_k,te_l)]^2+
456: O({1\over t^2})\\
457: =&{1\over t^4} \|\hat R \|^2+O({1\over t^2})\\
458: (||Ric_t||_t)^2=&\sum_{1\leq i,j\leq n}[Ric_t(e_i,e_j)]^2\\
459: =& {1\over t^4}\sum_{1\leq i,j\leq q}[{\hat Ric}(te_i,te_j)]^2={1\over t^4}||\hat Ric||^2+
460: O({1\over t^2})\\
461: (||scal_t||_t)^2=&{1\over t^4}||\hat scal||^2+O({1\over t^2})\\
462: \end{split}
463: \end{equation*}
464: Therefore, at the point $m$ we have:
465: \begin{equation}\label{h4MMM}
466: (h_4)_t= {1\over t^4}\hat{h_4}+O({1\over t^2})
467: \end{equation}
468: This completes the proof of theorem B since the total space is compact. \ppp
469: \par\medskip\noindent
470: \begin{corollary}
471: \begin{enumerate}
472: \item The product $S^p\times M$ of an arbitrary compact manifold M
473: with a sphere $S^p,p\geq 4$ admits a Riemannian metric with positive {\sl shwci}.
474: \item If a compact manifold admits a Riemannian foliation such that the
475: leaves are with
476: positive {\sl shwci} then the manifold admits a Riemannian metric with
477: positive {\sl shwci}.
478: \end{enumerate}
479: \end{corollary}
480: \par\medskip\noindent
481: {\bf Proof.} The first part is straightforward, to prove the second one it suffices to
482: notice that the proof of the previuos theorem works also in the case of local
483: Riemannian submersions. \ppp
484: \par\medskip\noindent
485: \begin{corollary} If a compact manifold M admits a free and smooth action
486: of a compact connected Lie group $G$ with rank $r$ such that $r<[{\dim G+1\over 2}]$
487: then the manifold M admits a Riemannian metric with
488: positive {\sl shwci}. \end{corollary}
489: \par\medskip\noindent
490: {\bf Proof.} The canonical projection $M\rightarrow M/G$ is in this case a
491: smooth submersion. Let the fibers be equipped with a biinvariant metric from
492: the group $G$ via the canonical inclusion ${\cal G}\subset T_mM$.
493: \par\noindent
494: Using any G-invariant metric on M, we define the horizontal distribution
495: to which we lift up an arbitrary metric from the basis $M/G$. Thus
496: we have defined a metric on M such that the projection $M\rightarrow M/G$
497: is a Riemannian submersion.\par\noindent
498: Finally, since the group $G$ with a biinvariant metric is
499: with positive {\sl shwci} then so are the fibers with the induced metric,
500: and we conclude using the previous theorem. \ppp \par\medskip\noindent
501: {\bf Remark.} All simple Lie groups satisfy the property $r<[{\dim G+1\over 2}]$.
502:
503: \section{Proof of Theorem C}
504: We proceed as in Gromov-Lawson's proof for the case of scalar curvature \cite{GroLaw}.\par
505: Let $(X,g)$ be a compact $n$-dimensional Riemannian manifold with positive {\sl shwci} and let
506: $S^m\subset X$ be an embedded sphere of codimension $q$ and with trivial normal bundle $N\equiv
507: S^m\times {\bf R}^q$. There exists $r_0>0$ such that the exponential map ${\rm exp}:
508: S^m\times D^q(r_{0})\rightarrow X$ is an embedding, where for every $x\in S^m$, $\{x\} \times D^q(r_{0})$ denotes the closed Euclidean ball in ${\bf R}^q\equiv \{ x\}\times {\bf R}^q$. Let ${\rm exp}^*g$ denotes the pull back of
509: the metric $g$ to the normal sub-bundle $S^m\times D^q(r_{0})$.
510: \par\smallskip
511: Another natural metric on the normal bundle is the metric $g^\nabla$
512: defined using the normal connection $\nabla$, that is the metric compatible
513: with the normal connection and such that the projection
514: $ S^m\times D^q(r)\rightarrow S^m $
515: is a Riemannian submersion.
516: We shall denote also by $g^\nabla$ its restriction to the sub-bundles
517: $S^m\times D^q(r)$ and $\partial (S^m\times D^q(r))=S^m\times S^{q-1}(r)$.
518: \par\smallskip
519: Recall that at each $(p,v)\in S^m\times D^q(r)$ we have a natural
520: $g^\nabla$-orthogonal decomposition of the tangent space
521: into vertical and horizontal subspaces, namely
522: \begin{equation}\label{riem:subm} T_{(p,v)} S^m\times D^q(r)
523: ={\cal V}_{(p,v)}+{\cal H}_{(p,v)} \end{equation}
524: where ${\cal V}_{(p,v)}$ is the tangent space
525: to the fiber (over $p$) $D^q(r)$ at $v$.
526: These two metrics are tangent to the order two in the directions tangent to
527: $D^q$, precisely
528: with respect to the decomposition \ref{riem:subm} we have (see \cite{Lab5})
529: \begin{equation}\label{bbb}
530: \left( \begin{matrix} \displaystyle g^\nabla + 0(r^2 ) &\displaystyle
531: & \displaystyle & g^\nabla + 0(r ) \cr\displaystyle
532: g^\nabla + 0(r) &\displaystyle & \displaystyle & g^\nabla + 0(r )\cr
533: \end{matrix}\right)
534: \end{equation}
535: \noindent
536: {\bf Remark.} Note that in \cite{GroLaw} in the begining of the proof of Lemma 2 page 430, it is claimed that
537: the former metrics are sufficiently close in the $C^2$-topology. But in general this is only true for the directions
538: tangent to $S^{q-1}(r)$, a detailed study of the behavior of these two metrics will appear in a separate forthcoming paper \cite{Lab5}. The same error is also in \cite{Lab4}. A short proof of this fact is as follows:\par
539: With respect to the metric $g^\nabla$, the sphere $S^m \hookrightarrow S^m\times D^q$ is totally geodesic (since for a Riemannian submersion the horizontal lift of a geodesic is a geodesic). But on the other side, the sphere $S^m \hookrightarrow S^m\times D^q$ is totally geodesic
540: for the metric ${\rm exp}^*g$ only if the sphere $S^m$ is totally geodesic in $(X,g)$.\par\smallskip
541: However this does not affect the corresponding conclusions in both papers
542: (after minor changes)
543: since the curvatures in question
544: (that is the scalar curvature and the $p$-curvatures, $p\leq q-3$) of these
545: two metrics on the bundles $S^m\times S^{q-1}(r)$ are high and close enough
546: as $r\rightarrow 0$.\par\medskip
547: Now, it is easy to see that the second fundamental form of $S^m\times S^{q-1}(r)$
548: in
549: $S^m\times D^q(r)$ with respect to the decomposition \ref{riem:subm}
550: is of the form
551: \begin{equation}\label{aaa}
552: \left( \begin{matrix} -{{\rm Id} \over r} &
553: 0 \cr
554: 0 & 0 \cr
555: \end{matrix} \right)
556: \end{equation}
557:
558: Consequently, using formulas \ref{bbb} and \ref{aaa} one can deduce without
559: difficulties that
560: the second fundamental form of $S^m\times S^{q-1}(r)$ in
561: $S^m\times D^q(r)$ with respect to the metric ${\rm exp}^*g$ is of the form (with respect to the decomposition \ref{riem:subm}):
562: \begin{equation}\label{ccc}
563: \left( \begin{matrix} - {Id \over r} + O(r) &
564: & O(1) \cr & &
565: \cr O(1) & & O(1)\cr
566: \end{matrix} \right) \end{equation}
567: \par
568: Note that since the second fundamental form is a continuous function, then it still has the form \ref{ccc} with respect to the following ${\rm exp}^{*}g$-orthogonal decomposition:
569: \begin{equation}\label{riem:submbis} T_{(p,v)} S^m\times D^q(r)
570: ={\cal V}_{(p,v)}\oplus {\cal H}'_{(p,v)} \end{equation}
571: where ${\cal V}_{(p,v)}$ is as in \ref{riem:subm} and the distribution ${\cal H}'$ is defined
572: by the previous orthogonal decomposition. Note that as $r\rightarrow 0$, the distribution
573: ${\cal H}'$ converges to the distribution
574: ${\cal H}$ defined by the decomposition \ref{riem:subm}.
575: \par\medskip
576: Now we define a hypersurface $M$ in the product
577: $ S^m\times D^q(r_{0})$ endowed with the product metric ${\rm exp}^{*}g \times {\bf R}$ by
578: the relation
579: \[
580: M=\left\{ ((x,v),t)\in S^m\times D^q(r_{0})\times {\bf R}\quad /\quad (\Vert v\Vert ,t)\in
581: \gamma \right\}
582: \]
583: where $\gamma $ is a curve whose graph in the $(r,t)$-plane as pictured
584: below:
585: \begin{figure}[bht]
586: \begin{center}
587: \epsfig{figure=fig_2.eps,height=50mm,width=57mm,clip=,angle=0,
588: silent=,bbllx=6mm,bblly=245mm,bburx=63mm,bbury=293mm}
589: \end{center}
590: \end{figure}
591:
592: The important points about $\gamma $ is that it is tangent to the $r$-axis
593: at $t=0$ and is constant for $r=\varepsilon >0$. Thus the induced metric on
594: $M$ extends the metric ${\rm exp}^{*}g$ on $S^m\times D^q(r_{0})$ near its boundary and
595: finishes with the product metric
596: $\bigl(\partial (S^m\times D^q(\varepsilon )),{\rm exp}^{*}g\bigr)
597: \times {\bf R}=\bigl(S^m\times S^{q-1}(\varepsilon ),{\rm exp}^{*}g\bigr) \times
598: {\bf R} $.\par\medskip
599: Next, we evaluate the {\sl shwci} of the hypersurface $M$.\par
600: For each $m\in M$, we have the following ${\rm exp}^{*}g$-orthogonal decomposition of $T_mM$
601: \begin{equation}\label{ddd}
602: T_mM={\bf R}\tau \oplus {\cal V}_m\oplus {\cal H}'_m
603: \end{equation}
604: where $\tau$ is the unit tangent vector to the curve $\gamma$ in the $(r,t)$-plane and
605: ${\cal V}_m, {\cal H}'_m$ are as in \ref{riem:submbis}.\par\medskip
606: It results by a straightforward computation using \ref{ccc} that the the second fundamental form of the hypersurface $M$
607: has the following
608: the form
609: (with respect to the decomposition \ref{ddd})
610: \begin{equation}\label{matrix}
611: \left( \begin{matrix}
612: k & 0 & ... & 0 \cr
613: 0 & & & \cr
614: & & \left(- {Id \over r} +
615: O(r)\right) {\sin} \theta &
616: \bigl( O(1){\sin} \theta \bigr)\cr
617: \vdots &
618: & & \cr
619: & & \bigl( O(1){\rm sin} \theta \bigr)
620: & \bigl( O(1){\rm sin} \theta \bigr)\cr
621: 0 & & &
622: \cr \end{matrix}\right)
623: \end{equation}
624: where $k$ denotes the curvature of the curve $\gamma$ in the $(r,t)$-plane
625: and $\theta$ denotes the angle between the normal to $M$ and the $t$-axis at the corresponding point.\par\medskip
626: Then a long but direct computation using the Gauss equation and the previous formula \ref{matrix} shows that the curvatures of $M$ have the form :
627: \begin{equation*}
628: \begin{split}
629: \| R^M\|^2=&\|R^{\scriptstyle {S^p}\times {D^q}}\|^2+{(q-1)(q-2)\over 2r^4}\sin^4\theta+(q-1){k^2\over r^2}\sin^2\theta+O({1\over r^2})\sin\theta\\
630: \|{\rm Ric}^M\|^2=&\|{\rm Ric}^{\scriptstyle{ S^p\times D^q}}\|^2+{(q-1)(q-2)^2\over r^4}\sin^4\theta+q(q-1){k^2\over r^2}\sin^2\theta\\
631: &{\hfill -{(q-1)(q-2)^2k\over r^3}\sin^3\theta +O({1\over r^2})\sin\theta}\\
632: \|{\rm scal}^M\|^2=&\|{\rm scal}^{\scriptstyle S^p\times D^q}\|^2+{(q-1)^2(q-2)^2\over r^4}\sin^4\theta+4(q-1)^2{k^2\over r^2}\sin^2\theta\\
633: &{\hfill -2{(q-1)^2(q-2)k\over r^3}\sin^3\theta +O({1\over r^3})\sin\theta}\\
634: \end{split}
635: \end{equation*}
636: where we supposed that the curve $\gamma$ has its curvature of the form $k=O({1\over r})$.\par
637: \noindent
638: Consequently, we can evaluate the {\sl shwci} of $M$ as follows
639: \begin{equation}\label{h4M}\begin{split}
640: h_4^M=h_4^{\scriptstyle S^p\times D^q}+&{(q-1)(q-2)(q-3)(q-4)\over 4r^4}\sin^4\theta\\
641: &-{(q-1)(q-2)(q-3)k\over 2r^3}\sin^3\theta +O({1\over r^3})\sin\theta
642: \end{split}
643: \end{equation}
644: Next we shall show that it is possible to choose the curve $\gamma$ so that the metric induced
645: on $M$ has positive {\sl shwci} at all points $m\in M$.\par\smallskip
646: Formula \ref{h4M} shows that for $\theta=0$ we have $h_4^M=h_4^{\scriptstyle S^p\times D^q}$ is
647: positive, and then there exists an angle $\theta_0>0$ such that for all $0<\theta\leq \theta_0$
648: the {\sl shwci} of $M$ is positive.\par
649: then we continue with a straight line ($k=0$) of angle $\theta_0$, say $\gamma_1$, until the term
650: ${(q-1)(q-2)(q-3)(q-4)\over 4r^4}\sin^4\theta_0$ is strongly dominating.\par
651: On the other hand, when $\theta=\pi/2$ then $k=0$ and $r=\epsilon$ we have
652: \begin{equation}\label{h4MM}
653: h_4^M={(q-1)(q-2)(q-3)(q-4)\over 4\epsilon^4}+O({1\over \epsilon^3})
654: \end{equation}
655: which is positive as $\epsilon$ is small enough and $q\geq 5$.\par\smallskip
656:
657: We now choose $r_1>0$ small and consider the point $(r_1,t_1)\in \gamma_1$. Then we bend
658: the straight line $\gamma_1$, begining at this point, with a curvature $k(s)$ of the following form
659: \begin{figure}[bht]
660: \begin{center}
661: \epsfig{figure=fig_1.eps,height=70mm,width=71mm,clip=,angle=0,
662: silent=,bbllx=2mm,bblly=222mm,bburx=73mm,bbury=293mm}
663: \end{center}
664: \end{figure}
665:
666: where the variable $s$ denotes the arc length along the curve. \par
667: Since $q\geq 5$, formula \ref{h4M} shows that
668: \begin{equation}\label{eee}
669: h_4^M\geq h_4^{\scriptstyle S^p\times D^q}+{(q-1)(q-2)(q-3)\over 2r^3}\sin^3\theta
670: ({\sin\theta\over 2r}-k)+O({1\over r^3})\sin\theta
671: \end{equation}
672: Then it is clear that
673: that the hypersurface $M$ will continue to have $h_4^M>0$, since $k<\frac{\sin\theta_0}{2r_1}
674: <\frac{\sin\theta}{2r}$.\par
675: After this first bending, we have
676: $\Delta r\leq \Delta s={r_1\over 2}$ and then $r\geq r_1-\Delta r\geq r_1-{r_1\over 2}>0$,
677: consequently the curve will not cross the $t$-axis.\par
678: On the other hand, $\Delta \theta =\int kds\approx {\sin\theta_0\over 4}$ is independent
679: of $r_1$. Clearly, by scaling down the curvature $k$, we can produce any
680: $\Delta \theta$ such that $0<\Delta \theta\leq {\sin\theta_0\over 4}$.\par
681: Our curve now continues with a new straight line $\gamma_2$ with angle
682: $\theta_1=\theta_0
683: +\Delta \theta$. By repeating this process finitely many times we can
684: achieve a total bend
685: of ${\pi\over 2}$.\par\smallskip
686: Let $g_\epsilon$ denote the induced metric from ${\rm exp}^*g$ on
687: $\partial(S^p\times D^q(\epsilon))=S^p\times S^{q-1}(\epsilon) $,
688: and recall that the new metric defined on $M$ is the old metric when $t=0$,
689: and finishes with the product
690: metric $g_\epsilon\times {\bf R}$. In the following we shall deform the
691: product metric $g_\epsilon \times {\bf R}$ on $ S^p\times S^{q-1}(\epsilon) $,
692: to the standard product metric through metrics with positive {\sl shwci}. This will
693: be done in two steps:\par
694: Step1: We deform The metric $g_\epsilon$ on
695: $S^m\times S^{q-1}(\epsilon) $ to the standard product metric
696: $S^m(1)\times S^{q-1}(\epsilon) $ through metrics with positive {\sl shwci}, as follows:\par
697: First, the metric $g_\epsilon$ can be homotoped through metrics
698: with $h_4>0$ to the normal metric $g^\nabla$ since their {\sl shwci} are
699: respectively high and close enough, see formulas \ref{h4MM} and \ref{h4MMM}.
700: \par
701: Then, for $\epsilon$ small enough, we can deform the normal metric $g^\nabla$
702: on
703: $S^m\times S^{q-1}(\epsilon )$ through Riemannian submersions to a new
704: metric where
705: $S^p$ is the standard sphere $S^p(1)$, keeping the horizontal distribution
706: fixed.\par\noindent
707: This deformation keeps $h_4>0$ as far as $\epsilon$ is small enough,
708: see formula \ref{h4MMM}.\par\smallskip
709: Finally, we deform the horizontal distribution to the standard one and
710: again by the same formula \ref{h4MMM}
711: this can be done keeping $h_4>0$. \par\medskip
712: Step2: Let us denote by $ds_t^2, 0\leq t\leq 1$, the previous family of
713: deformations on $S^m\times S^{q-1}(\epsilon) $. They are all with positive
714: {\sl shwci}. Where $ds_0=g_\epsilon$ and $ds_1$ is the standard product metric.\par
715: \noindent
716: It is clear that the metric $ds_{{t\over a}}^2+dt^2, 0\leq t\leq a$, glues together
717: the two metrics $ds_0\times {\bf R}$ and $ds_1\times {\bf R}$. Furthermore,
718: there exists $a_0>0$ such that for all $a\geq a_0$ the metric
719: $ds_{{t\over a}}^2+dt^2$ on $S^m\times S^{q-1}(\epsilon)\times [0,a] $
720: is with positive {\sl shwci}. In fact, via a change of variable, this is equivalent
721: to the existence of $\lambda_0>0$ such that for all $0<\lambda\leq \lambda_0$,
722: the metric $\lambda^2ds_t^2+dt^2$ is with positive {\sl shwci}. This is already known to be true
723: again by formula \ref{h4MMM}. \ppp
724: \begin{corollary} Let G be a finitely presented group.
725: Then for every $n\geq 6$, there exists a compact n-manifold M with positive
726: {\sl shwci} such that $\pi_1(M)=G$.
727: \end{corollary}
728: {\bf Proof.} Let $G$ be a group which has a presentation
729: consisting of $k$ generators $x_{1},x_{2},...,x_{k}$ and $l$ relations
730: $r_{1},r_{2},...,r_{l}$.
731: \par\noindent Let $S^{1}\times S^{n-1}$ be endowed with the standard product
732: metric which is with positive {\sl shwci} ($n-1\geq 4$). Remark that the
733: fundamental group of $S^{1}\times S^{n-1}$ is infinite cyclic. Hence by
734: taking the connected sum $N$ of $k$-copies of $S^{1}\times S^{n-1}$ we
735: obtain an orientable compact $n$-manifold with positive {\sl shwci} (since
736: this operation is a surgery of codimension $n\geq 5$). By Van-Kampen
737: theorem, the fundamental group of $N$ is a free group on $n$-generators,
738: which we may denote by $x_{1},x_{2},...,x_{k}$.
739: \par\noindent We now perform surgery $l$-times on the manifold $N$,
740: killing in
741: succession the elements $r_{1},r_{2},...,r_{l}$. The result will be a
742: compact, orientable $n$-manifold $M$ with positive {\sl shwci} (since the
743: surgery is of codimension $n-1\geq 5$) such that $\pi _{1}(M)=G$, as
744: required. \ppp
745:
746:
747:
748:
749:
750:
751:
752:
753:
754:
755: \bigskip\noindent
756:
757:
758: \begin{thebibliography}{99}
759: \bibitem{Besse} Besse, A. L., \emph{ Einstein manifolds} Springer-Verlag (1987).
760: \bibitem{GroLaw} Gromov, M., Lawson, H. B., \emph{ The classification of simply connected manifolds of positive scalar curvature}, Annals of Math 111, 423-434 (1980).
761: \bibitem{Kulk} Kulkarni, R. S., \emph{On Bianchi Identities}, Math. Ann. 199, 175-204 (1972).
762: Journal of math (1974).
763: \bibitem{Lab1} Labbi, M.-L., \emph{ Double forms, curvature structures and the
764: $(p,q)$-curvatures}, to appear.
765: \bibitem{Lab2} Labbi, M.-L., \emph{ On compact manifolds with positive isotropic curvature},
766: Proceedings of the American Mathematical Society, Volume 128, Number 5, Pages 1467-1474 (2000).
767: \bibitem{Lab3} Labbi, M.-L., \emph{ Actions des groupes de Lie presques simples et positivit\'e
768: de la $p$-courbure}, Annales de la facult\'e des Sciences de Toulouse, volume 5, Number 2, (1997)
769: 263-276.
770: \bibitem{Lab4} Labbi, M.-L., \emph{ Stability of the $p$-curvature positivity under surgeries and manifolds with positive Einstein tensor,} Annals of Global Analysis and Geometry, 15 (1997) 299-312.
771: \bibitem{Lab5} Labbi, M.-L., \emph{ On two natural Riemannian metrics on a
772: tubular neighberhood of an embedded submanifold}, to appear.
773: \bibitem{Thorpe} Thorpe, J. A., \emph{ Some remarks on the Gauss-Bonnet integral}, Journal of
774: Mathematics and Mechanics, Volume 18, Number 8 (1969).
775: \end{thebibliography}
776:
777:
778:
779:
780:
781:
782:
783:
784:
785:
786:
787:
788:
789:
790:
791:
792:
793:
794:
795:
796:
797: \end{document}
798: