Subject: top question From: Wolfgang Ziller Date: Mon, 18 Dec 2006 23:10:02 -0500 Dear topology friends, I have an algebraic topology question which has applications in geometry (nonnegative curvature) It is a very concrete and special situation I am in. I have a family of SO(k) principal bundles P over CP2 , k>4 , with w_2 \ne 0 P arises from a concrete construction. All I know about P though is that SO(k) acts freely and I know the cohomology ring of P. From this knowledge can I can determine the bundle? By Dold Whitney's classification of bundles over simply connected 4-manifolds, such bundles are classified by w_2 , w_4 and p_1. p_1 can be interpreted as an integer by evaluating on a fundamental cycle. The structure group of course reduces to 4 dim bundles where w_4 is euler class mod 2. But my bundles have k>4 and from their construction there is no easy way to reduce the structure group explicitly. >From this though one can determine the restriction that need to be satisfied for these numbers since p_1=2k+2l e=k-l if w_2 = 0 and p_1 = 2k+2l+1 , e= k-l if w_2 \ne 0 where k,l are arbitrary integers. It is not hard to recognize w_2 and | p_1 | in the top of P : w_2\ne 0 iff P is simply connected and if p_1 \ne 0 then H4(P,Z) is a finite group of order |p_1|. If w_2=0 I am ok since then p_1= 2 mod 4 implies w_4 \ne 0 and p_1 = 0 mod 4 implies w_4 = 0 But when I have w_2 \ne 0 it only implies that if p_1 = 1 mod 4 then w_4 \ne 0 and if p_1 = 3 mod 4 then w_4=0. So my problem is that I cannot recognize the sign of p_1 from the coh ring . Is there some other information in the coh ring that will tell me what w_4 is? I assume that this is somehow encoded in the spectral sequence for the universal bundle (maybe with Z_2 coefficients). This is how one sees the above facts about w_2 and |p_1| for example Another question would be if you know two such bundles where p_1 differs by a sign concretely enough so I can see what to expect from their cohomology ring? Any advice would be appreciated. Thanks! Wolfgang