On adic genus, Postnikov conjugates, and lambda-rings Donald Yau April 25, 2002 Department of Mathematics, MIT Room 2-588, 77 Massachusetts Av- enue, Cambridge, MA 02139, USA Abstract Sufficient conditions on a space are given which guarantee that the K-theory ring and the ordinary cohomology ring with coefficients over a principal ideal domain are invariants of, respectively, the adic genus and the SNT set. An independent proof of Notbohm's theorem on the classification of the adic genus of BS3 by KO-theory ~-rings is given. An immediate consequence of these results about adic genus is that for any positive integer n, the power series ring Z[[x1, . .,.xn]] admits uncountably many pairwise non-isomorphic ~-ring structures. MSC: 55P15; 55N15, 55P60, 55S25. Keywords: Adic genus, Postnikov conjugates, SNT, lambda-rings. 1 Introduction and statement of results In this paper we study some problems about the adic genus and the SNT set of a space. For a nilpotent finite type space X, we denote by Genus (X), call the adic genus of X, the set of homotopy types of nilpotent finite type spaces Y such that the p-completions of X and Y are homotopy equivalent for each prime p and also their rationalizations are homotopy equivalent. The adic genus of an infinite dimensional space X is often a very big set. For instance, Møller [17] proved that whenever G is a compact con- nected non-abelian Lie group, the adic genus of its classifying space BG is 1 uncountably large. It is an important problem to find computable homotopy invariants which can distinguish between spaces with the same adic genus. Such a result was achieved by Notbohm and Smith [19] (see also [11]). Re- call that a space X is said to be an adic fake Lie group of type G if (1) X = BX is a finite loop space and (2) BX lies in the adic genus of BG. Notbohm and Smith showed that if X is an adic fake Lie group of type G, where G is a simply-connected compact Lie group, then BX is homotopy equivalent to BG if and only if K(BX) ~=K(BG) as ~-rings. (Here K(Y ) denotes the complex K-theory of Y .) In fact, Notbohm [18] showed that K-theory ~-rings classify the genus of BG, provided G is a simply-connected compact Lie group. Note that what we call a ~-ring is what used to be called a "special" ~-ring. With this result in mind, a natural question is then the following: Is it really necessary to take into account the ~-operations in order to distinguish spaces in the adic genus of BG? To describe our answer to this question, we need the following definition. A filtered ring is a commutative ring R with unit together with a decreasing filtration R = I0 I1 I2 . . . of ideals such that Im In Im+n . A map of filtered rings is required to respect filtrations. The filtration in the K-theory of a space X is given by a skeletal filtration of X; that is, K(X) is filtered by the kernels In (X) = ker(K(X) ! K(Xn-1)) of the restriction maps, where Xn-1 denotes the (n - 1)-skeleton of X. By using the Cellular Approximation Theorem, it is easy to see that the filtered ring isomorphism type of a given space X is well-defined, though there are many different filtrations on X. Clearly, a map between two spaces induces a filtered ring map between the respective K-theories. Our first main result is then the following, which shows that for a tor- sion free classifying space, K-theory filtered ring cannot tell the difference 2 between spaces in its adic genus. Consequently, ~-operations are necessary in order to distinguish these spaces. Theorem 1.1. Let X be a simply-connected space of finite type whose in- tegral homology is torsion free and is concentrated in even dimensions and whose K-theory filtered ring is a finitely generated power series ring over Z. If a space Y belongs to the adic genus of X, then there exists a filtered ring isomorphism K(X) ~=K(Y ). For example, in Theorem 1.1 the space X can be BSp(n) (n 1), BSU(n) (n 2), or any finite product of copies of these spaces and of infinite complex projective space. Remark 1.2. The proof of Theorem 1.1 is independent of Notbohm's result mentioned above. It consists of first showing the weaker statement that the filtered rings K(Y )=In (Y ) and K(X)=In (X) are isomorphic for all n suffi- ciently large. Since K(Y ) can be recovered from the quotients K(Y )=In (Y ) by taking inverse limits, a filtered ring analog of a classification result of Wilkerson [25] implies that to prove Theorem 1.1 it is sufficient to prove the triviality of a lim-1set. Remark 1.3. This theorem has a variant in which complex K-theory (resp. Z) is replaced with KO*-theory (resp. KO* = KO*(pt)), provided the inte- gral homology of X (which is simply-connected of finite type) is torsion free and is concentrated in dimensions divisible by 4 (e.g. X = BSp (n)). This variant admits a proof which is essentially identical with that of Theorem 1.1. Using Theorem 1.1 we will show the following KO-theory analog of Not- bohm's result (applied to BS3) mentioned above. Theorem 1.4. Let X and Y be spaces in the adic genus of BS3. Then X and Y are homotopy equivalent if and only if there exists a filtered ~-ring isomorphism KO*(X) ~=KO*(Y ). In Theorem 1.4, by a filtered ~-ring we mean a filtered ring R which is also a ~-ring for which the filtration ideals are all closed under the operatio* *ns 3 ~i (i > 0). The proof of of this theorem uses directly Rector's classification of the genus of BS3 [20] and the KO-theory analog of Theorem 1.1 when X = BS3. Our next result is a purely algebraic application of Theorem 1.1 to the study of ~-rings. Now since the adic genus of BSp(n), n any positive integer, is uncount- able [17], Theorems 1.1 and Notbohm's result alluded to above imply the following purely algebraic statement about ~-rings. Corollary 1.5. Let n be any positive integer. There exist uncountably many distinct isomorphism classes of ~-ring structures over the power series ring Z[[x1, . .,.xn]] on n indeterminates. Remark 1.6. As far as the author is aware, this result is new for every n. It exhibits a huge diversity of ~-ring structures over some rings. The author has recently obtained results about classifying filtered ~-ring structures over a filtered ring, exhibiting this set as a moduli space, and also results about how some filtered ~-rings give rise to unstable algebras over the mod p Steenrod algebra. Remark 1.7. Notice that the way we obtain Corollary 1.5 is by combining three topological results about spaces in the genus of classifying spaces and their K-theories, even though the statement of the result is purely algebraic. It would be nice if this statement can be demonstrated in an algebraic way. Remark 1.8. The analogous question of how many ~-ring structures the polynomial ring Z[x] on one indeterminate supports has been studied by Clauwens [7]. Employing the theory of commuting polynomials, he showed that there are essentially only two non-isomorphic ~-ring structures on the polynomial ring Z[x]. We, however, have not been able to establish any connections between Clauwens' result and our Corollary 1.5. As pointed out by Clauwens, the same question for the polynomial ring on n indeterminates, n > 1, is still open. 4 The last main result of this paper concerns the SNT set of a space. For a given nilpotent finite type space X, SNT (X) is by definition the set of homotopy types of nilpotent finite type spaces Y such that for every n the Postnikov approximations through dimension n of X and Y are homotopy equivalent. Spaces in the same SNT set are also known as spaces of the same n-type for all n (hence the notation SNT ). The SNT set of a space is closely related to its adic genus. Indeed, according to a result of Wilkerson [25], if X is a connected nilpotent finite type space, then there is an inclusion of pointed sets SNT (X) Genus (X). The adic genus of a space, when it is nontrivial, is often strictly bigger than its SNT set. There are many interesting spaces whose SNT sets are nontrivial (in fact, uncountable); see, for example, [9, 15, 22, 23]. Several homotopy invariants which can distinguish between spaces of the same n-type for all n have been found. These include [16] Aut(-) and WI (-), the group of homotopy classes of homotopy self-equivalences and the group of weak identities, and [24] End(-), the monoid of homotopy classes of self-maps. One might wonder if there are more familiar and more computable homotopy invariants, such as integral cohomology and K-theory, which can distinguish between spaces in the same SNT set. Of course, our Theorem 1.1 together with the above inclusion imply that if a space X is as in Theorem 1.1 and if Y lies in SNT (X), then the integral cohomology (resp. K-theory) rings of X and Y are isomorphic. Does this hold for more general spaces? The following result gives an answer to this question for the ordinary cohomology case. Theorem 1.9. Let be a principal ideal domain. Let X be a connected space of finite type whose ordinary cohomology ring with coefficients over is a finitely generated graded algebra over . If Y has the same n-type 5 as X for all n, then there exists an isomorphism H*(X; ) ~=H*(Y ; ) of cohomology rings. Remark 1.10. It should be emphasized that in Theorem 1.9 the hypothesis Y 2 SNT (X) cannot be weakened to the condition Y 2 Genus (X). Indeed, in [5] Bokor constructed two spaces, both two cell complexes, with the same genus but whose integral cohomology rings are non-isomorphic. This finishes the presentation of the results in this paper. The rest of the paper is organized as follows. In x2 we recall the definitions of a (filtered) ~-ring and of Adams operations. The proofs of Theorems 1.1, 1.4, and 1.9 are given in this order, one in each section, in x3 - x5. 2 ~-rings and Adams operations In this section, we recall the definitions of a (filtered) ~-ring and of Adams operations. The reader is referred to Atiyah and Tall [3] or Knutson [13] for more information about ~-rings. 2.1 ~-rings A ~-ring is a commutative ring R with unit together with functions ~i:R ! R for i = 0, 1, . .s.atisfying the following properties: For any elements r and s in R one has o ~0(r) = 1 o ~1(r) = r, ~n(1) = 0 for all n > 1 P n o ~n(r + s) = i=0 ~i(r)~n-i(s) o ~n(rs) = Pn(~1(r), . .,.~n(r); ~1(s), . .,.~n(s)) o ~n(~m (r)) = Pn,m(~1(r), . .,.~nm (r)) 6 Here the Pn and Pn,m are certain universal polynomials with integer coef- ficients. Note that in the literature (for example, Atiyah and Tall [3]) the terminology "special" ~-ring is used. A ~-ring map f :R ! S between two ~-rings is a ring map between the underlying rings which respects the operations ~i: f~i= ~if (i 0). 2.2 Filtered ~-rings A filtered ~-ring is a filtered ring R = (R, {In }) which is also a ~-ring such that the ideals In are all closed under the operations ~i (i > 0). A filtered ~-ring map is a ~-ring map which is also a filtered ring map. 2.3 Adams operations Given a ~-ring R, the Adams operations _k :R ! R for k = 1, 2, . .a.re defined inductively by the Newton formula: _k(a) - ~1(a)_k-1(a) + . .+.(-1)k-1~k-1(a)_1(a) = (-1)k-1k~k(a). The Adams operations satisfy the following properties. 1. All the _k :R ! R are ~-ring maps. 2. _1 = Idand _k_l = _kl for any k, l 1. 3. _p(a) ap (mod pR) for each prime p and element a in R. If R is a filtered ~-ring, then the Adams operations are all filtered ~-ring maps. 3 Proof of Theorem 1.1 In this section we prove Theorem 1.1, which consists of a few lemmas. Throughout this section we work in the category of filtered rings. 7 We will make use of the following observation, whose proof is a straight- forward adaptation of Wilkerson's proof of the classification theorem [25, Theorem I] of spaces of the same n-type for all n. Now let X be as in Theorem 1.1. Lemma 3.1 (Wilkerson). There is a bijection between the following two pointed sets: 1. The pointed set of isomorphism classes of filtered rings (R, {In}) with the properties: (a) The natural map R ! lim-nR=In is an isomorphism, and (b) R=In and K(X)=Kn(X) are isomorphic as filtered rings for all n > 0. 2. The pointed set lim-1nAut(K(X)=Kn(X)). Here Aut (-) denotes the group of filtered ring automorphisms, and the lim-1of a tower of not-necessarily abelian groups is as defined in Bousfield- Kan [6]. The two lemmas below will show that, for every space Y in the genus of X, the object K(Y ) lies in the first pointed set in Lemma 3.1. Lemma 3.2. For every space Y in the genus of X (as in Theorem 1.1), the natural map K(Y ) ! lim-K(Y )=Kj(Y ) j is an isomorphism. Proof.According to [4, 2.5 and 7.1] the natural map from K(Y ) to lim-jK(Y )=Kj(Y ) is an isomorphism if the following condition holds: lim-1rEp,rq= 0 for all pairs(p, q). (3.1) Here E*,r*is the Er-term in the K*-Atiyah-Hirzebruch spectral sequence (AHSS) for Y . This condition is satisfied, in particular, when the AHSS degenerates at the E2-term. Thus, to prove (3.1)it suffices to show that 8 H*(Y ; Z) is concentrated in even dimensions, since in that case there is no room for differentials in the AHSS. So pick an odd integer N. We must show that HN (Y ; Z) ~=0. (3.2) By the Universal Coefficient Theorem it suffices to show that the integral homology of Y is torsionfree and is concentrated in even dimensions, which hold because Y lies in the genus of X. __ This finishes the proof of Lemma 3.2. |__| Lemma 3.3. For every space Y in the genus of X (as in Theorem 1.1), the filtered rings K(Y )=Kn(Y ) and K(X)=Kn(X) are isomorphic for each n > 0. Proof.We have to show that for each j > 0 there is an isomorphism of filtered rings K(Y )=Kj(Y ) ~=K(X)=Kj(X). (3.3) It follows from the hypothesis that for each j > 0, the filtered r* *ing K(Y )=Kj(Y ) belongs to Genus (K(X)=Kj(X)), where Genus (R) for a fil- tered ring R is defined in terms of R Q and R bZin exactly the same way the genus of a space is defined. To finish the proof we will adapt two results of Wilkerson [26, 3.7 and 3.8], which we now recall. For a nilpotent finite type space X, Wilkerson showed that there is a surjection oe :Caut (X^0) ! Genus (X), Q where X^0is the rationalization of the formal completion X^= pX^pof X. Notice that each homotopy group ß*(X^0) is a Q bZ-module, and Caut(X^0) is by definition the group of homotopy classes of self-homotopy equivalences of X^0whose induced maps on homotopy groups are Q bZ-module maps. Note that the definitions Genus (-) and Caut (-) also make sense in both the categories of nilpotent groups and of filtered rings. For instance, if R = (R, {In }) is a filtered ring, then Caut(R Q Zb) is the group of filtered ring automorphisms of R Q bZwhich are also Q bZ-module maps. Now 9 if A is a finitely generated abelian group, then Wilkerson showed that for any class ['] 2 Caut (A Q bZ), the image oe([']) is isomorphic to A as groups; that is, the image of oe is constant at A. It is straightforward to adapt Wilkerson's proofs of these results to show that for each j the map oe :Caut ((K(X)=Kj(X)) Q bZ) ! Genus (K(X)=Kj(X)) is surjective and that the image of oe is constant at K(X)=Kj(X). In other words, the genus of K(X)=Kj(X) is the one-point set. Therefore, K(Y )=Kj(Y ) is isomorphic to K(X)=Kj(X). __ This finishes the proof of Lemma 3.3. |__| In view of Lemma 3.1, to finish the proof of Theorem 1.1 we are only left to show that the classifying object lim-1nAut(K(X)=Kn(X)) is the one point set. To do this, it suffices to show that almost all the structure maps in the tower are surjective. This is shown in the following lemma. Lemma 3.4. The maps Aut(K(X)=Kj+1(X)) ! Aut(K(X)=Kj(X)) are surjective for all j sufficiently large. Proof.First note that by hypothesis the K-theory filtered ring of X has the form K(X) = Z[[c1, . .,.cn]] (3.4) in which the generators ci are algebraically independent over Z and Kj(X) is the ideal generated by the monomials of filtrations at least j. Suppose that di is the largest integer k for which ci lies in filtration k. Let N be the integer N = max {di:1 i n} + 1. 10 We will show that the structure maps Aut(K(X)=Kj+1(X)) ! Aut(K(X)=Kj(X)) are surjective for all j > N. So fix an integer j > N and pick a filtered ring automorphism oe of K(X)=Kj(X). We must show that oe can be lifted to a filtered ring auto- morphism of K(X)=Kj+1(X). For 1 i n pick any lift of the element oe(ci) to K(X)=Kj+1(X) and call it ^oe(ci). Since there are no relations among the ci in K(X), it is easy to see that ^oeextends to a well-defined filtered ri* *ng endomorphism of K(X)=Kj+1(X), and it will be a desired lift of oe once it is shown to be bijective. To show that ^oe:K(X)=Kj+1(X) ! K(X)=Kj+1(X) is surjective, it suffices to show that the image of each ci in K(X)=Kj+1(X) lies in the image of ^oe, since K(X)=Kj+1(X) is generated as a filtered ring by the images of the ci. So fix an integer i with 1 i n. We know that there exists an element gi2 K(X)=Kj(X) such that oe(gi) = ci. (3.5) Pick any lift of gi to K(X)=Kj+1(X), call it gi again, and observe that (3.5) implies that ^oe(gi) = ci+ ffi (3.6) in K(X)=Kj+1(X) for some element ffi 2 Kj(X)=Kj+1(X). We will alter gi to obtain a ^oe-pre-image of ci as follows. Observe that the ideal Kj(X)=Kj+1(X) is generated by certain mono- mials in c1, . .,.cn. Namely, the monomials ci = ci11. .c.inn, i = (i1, . .,.in) 2 Jj (3.7) where Jj is the set of ordered n-tuples i = (i1, . .,.in) of nonnegative intege* *rs satisfying Xn d . i = dlil = j. l=1 11 Thus, for every element i in the set Jj, there exists a corresponding integer ai such that we can write the element ffi as the sum X ffi = aici. (3.8) i 2 Jj Now define the element ~giin K(X)=Kj+1(X) by the formula X ~gidef=gi - aigi, where gi= gi11. .g.inn. (3.9) i 2 Jj We claim that ~giis a ^oe-pre-image of ci. That is, we claim that ^oe(~gi) = ci in K(X)=Kj+1(X). (3.10) In view of (3.6), (3.8), and (3.9), it clearly suffices to prove the following equality for each element i in Jj: i j ^oegi = ci in K(X)=Kj+1(X). (3.11) Now in the quotient K(X)=Kj+1(X), one computes i j Yn ^oegi = ^oe(gj)ij j=1 Yn = (cj + ffj)ij by (3.6) j=1 = ci + (terms of filtrations> j) = ci. This proves (3.11), and hence (3.10), and therefore ^oeis surjective. It remains to show that o^eis injective. Since any surjective endo- morphism of a finitely generated abelian group is also injective and since K(X)=Kj+1(X) is a finitely generated abelian group, it follows that ^oeis in- jective as well. Thus, ^oeis a filtered-ring automorphism of K(X)=Kj+1(X) and is a lift of oe. __ This finishes the proof of Lemma 3.4. |__| This proof of Theorem 1.1 is complete. 12 4 Proof of Theorem 1.4 In this section we prove Theorem 1.4. The arguments in this section, espe- cially Lemma 4.2 below, are inspired by Rector's [21, x4]. We begin by noting that an argument entirely similar to the proof of Theorem 1.1 implies that whenever X belongs to Genus (BS3), one has KO*(X) ~=KO*[[x]] as filtered rings, where x 2 KO44(X) is a representative of an integral genera- tor x4 2 H4(X; Z) = E4,20in the KO*-Atiyah-Hirzebruch spectral sequence for X. Here KOab(X) denotes the subgroup of KOa(X) consisting of ele- ments u which restrict to 0 under the natural map KOa(X) ! KOa(Xb-1). Such an element u is said to be in degree a and filtration b. Now we recall the relevant notations, definitions, and results regarding Rector's classification of the genus of BS3 [20]. Let , 2 ß-4KO and bR 2 ß-8KO be the generators so that ,2 = 4bR . As usual, denote by _k (k = 1, 2, . .).the Adams operations. Since X is homotopy equivalent to S3, it follows as in [21, x4] that there exists an integer a, depending on the choice of the representative x, such that the following statements hold. 1. _2(,x) = 4,x + 2abR x2 (mod KO09(X)). 2. The integer a is well-defined (mod 24). This means that if x0is another representative of x4 with corresponding integer a0, then a a0 (mod 24), and if x4 is replaced with -x4, then a will be replaced with -a. We can therefore write a(X) for a. 3. a(X) 1, 5, 7, or 11 (mod 24). The last condition above follows from the examples constructed by Rector in [21, x5] and James' result [10] which says that there are precisely eight ho- motopy classes of homotopy-associative multiplications on S3. These eight classes can be divided into four pairs with each pair consisting of a homotopy class of multiplication and its inverse. 13 Rector's invariant (X=p) for p an odd prime is defined as follows [20]. The Adem relation P 1P 1= 2P 2implies that P 1__x4= 2__x(p+1)=24 in H*(X; Z=p), where __x4is the mod p reduction of the integral generator x4. Then (X=p) 2 { 1} is defined as the sign on the right-hand side of this equation. The invariant (X=2) and a canonical choice of orientation of the integral generator are given as follows. Using the mod 24 integer a(X), define 8 >>>(1, 1) ifa(X) 1 mod 24; >>> < (1, -1) ifa(X) 5 mod 24; ((X=2), (X=3)) = (4.1) >>>(-1, 1) ifa(X) 7 mod 24; >>> : (-1, -1) ifa(X) 11 mod 24. The orientation is then chosen so that (X=3) is as given in (4.1). This definition of Rector's invariants coincides with the original one (cf. [14, x9]* *). Now we can recall the classification theorem of the genus of BS3 [20]. Theorem 4.1 (Rector). The (X=p) for p primes provide a complete list of classification invariants for the (adic) genus of BS3. Any combination of values of the (X=p) can occur. If X is BS3 then (X=p) = 1 for any prime p. From now on in this section, X and Y are as in Theorem 1.4. Now we can prove Theorem 1.4. Proof of Theorem 1.4.In view of Theorem 4.1, to prove Theorem 1.4 it suffices to show that if there exists a filtered ~-ring isomorphism oe :KO*(X) ~=KO*(Y ), then a(X) a(Y ) (mod 24) 14 and (X=p) = (Y=p) for all odd primes p. We will prove these in Lemmas 4.2 and 4.3 below, __ thereby proving Theorem 1.4. |__| As explained above, KO*(X) = KO*[[x]] and KO*(Y ) = KO*[[y]] with x 2 KO44(X) and y 2 KO44(Y ) representing, respectively, the integral gen- erators x4 2 H4(X; Z) and y4 2 H4(Y ; Z). Lemma 4.2. If there exists a filtered ~-ring isomorphism ~= * oe :KO*(X) -! KO (Y ), then a(X) a(Y ) (mod 24). Proof.Since oe is a ring isomorphism, we have oe(,x) = ",y + oe2bR y2 (mod KO09(Y )) for some integer oe2 and " 2 { 1}. Computing modulo KO09(X) we have 4oe(bR x2)= oe(,x)2 = ,2y2 = 4bR y2. Therefore, one has oe(bR x2) = bR y2 (mod KO09(Y )). First we claim that there is an equality a(X) = 6oe2 + ä (Y ). (4.2) To prove (4.2)we will compute both sides of the equality oe_2(,x) = _2oe(,x) (mod KO09(Y )). 15 Working modulo KO09(Y ) we have, on the one hand, oe_2(,x) = oe(4,x + 2a(X)bR x2) = 4(",y + oe2bR y2) + 2a(X)bR y2 = 4",y + (4oe2 + 2a(X))bR y2. On the other hand, still working modulo KO09(Y ), we have _2oe(,x) = "_2(,y) + oe2_2(bR y2) = "(4,y + 2a(Y )bR y2) + oe2(24bR y2) = 4",y + (16oe2 + 2ä (Y ))bR y2. Equation (4.2)now follows by equating the coefficients of bR y2. In view of (4.2), to finish the proof of Lemma 4.2 it is enough to establish oe2 0 (mod 4). (4.3) To prove (4.3), note that since oe is a KO*-module map, we have ,oe(x) = oe(,x). Since oe is a ring isomorphism, we also have oe(x) = "0y + oe02,y2 (mod KO49(Y )) for some integer oe02and "02 { 1}. Therefore, working modulo KO09(Y ) we have ,oe(x) = "0,y + oe02,2y2 = "0,y + 4oe02bR y2 = ",y + oe2bR y2. In particular, by equating the coefficients of bR y2 we obtain oe2 = 4oe02, thereby proving (4.3). __ This completes the proof of Lemma 4.2. |__| 16 Lemma 4.3. If there exists a filtered ~-ring isomorphism ~= * oe :KO*(X) -! KO (Y ), then (X=p) = (Y=p) for each odd prime p. Proof.It follows from Theorem 1.1 that K*(X) ~=K*[[ux]] with ux 2 K44(X) a representative of the integral generator x4 2 H4(X; Z) = E4,02in the K*-Atiyah-Hirzebruch spectral sequence. Moreover, we may choose ux so that c(x) = ux, where c: KO*(X) ! K*(X) is the complexification map. Similar remarks apply to Y so that K*(Y ) ~=K*[[uy]]. Now denote by b 2 ß-2K the Bott element and let p be a fixed odd prime. We first claim that _p(b2ux) = (b2ux)p + 2(X=p) p (b2ux)(p+1)=2+ p wx + p2x0 (4.4) for some wx 2 K02p+3(X) and some x0 2 K04(X). To see this, note that since b2ux 2 K04(X), it follows from Atiyah's theorem [2, 5.6] that _p(b2ux) = (b2ux)p + p x1 + p2x0 for some xi2 K04+2i(p-1)(X) (i = 0, 1). Moreover, one has __x 1_____2 1 = P b ux, 17 where __zis the mod p reduction of z and P 1is the Steenrod operation of degree 2(p - 1) in mod p cohomology. Thus, to prove (4.4)it is enough to show that x1 = 2(X=p)(b2ux)(p+1)=2+ wx + p zx (4.5) for some wx 2 K02p+3(X) and some zx 2 K02p+2(X). Now in H*(X; Z) Z=p we have __x 1_____2 1 = P b ux = P 1__x4 = 2(X=p) __x(p+1)=24 _____(p+1)=2 = 2(X=p) b2ux . Now (4.5) follows immediately. As remarked above, this also establishes (4.4). Now the ~-ring isomorphism oe induces via c a ~-ring isomorphism ~= * oec: K*(X) -! K (Y ). By composing oec with a suitable ~-ring automorphism of K*(Y ) if necessary, we obtain a ~-ring isomorphism ~= * ff: K*(X) -! K (Y ) with the property that ff(b2ux) = b2uy + higher terms inb2uy. (4.6) Using (4.4)and (4.6)it is then easy to check that ff_p(b2ux) = 2(X=p) p (b2uy)(p+1)=2 (mod K02p+3(Y ) and p2)(4.7) and _pff(b2ux) = 2(Y=p) p (b2uy)(p+1)=2 (mod K02p+3(Y ) and p2).(4.8) Since ff_p = _pff it follows from (4.7)and (4.8)that 2(X=p) p 2(Y=p) p (mod p2), 18 or, equivalently, 2(X=p) 2(Y=p) (mod p). But p is assumed odd, and so (X=p) (Y=p) (mod p). Hence (X=p) = (Y=p), as desired. __ This finishes the proof of Lemma 4.3. |__| The proof of Theorem 1.4 is complete. 5 Proof of Theorem 1.9 In this final section we give the proof of Theorem 1.9. The argument is somewhat similar to the proof of Theorem 1.1. All cohomology groups and rings have coefficients over a fixed principal ideal domain . Throughout this section we are working in the category of (non-negatively) graded algebras over . We need the following two lemmas. Lemma 5.1. For each n > 0, there is an isomorphism H n (X) ~= H n (Y ) of graded algebras over . Proof.This follows immediately from the assumption that X and Y have __ the same n-type for all n. |__| By hypothesis the cohomology ring of X has the form H*(X) = [x1, . .,.xs]=J for some homogeneous generators xi (1 i s) and some homogeneous ideal J. Suppose that |xi| = di. Note that since the polynomial ring 19 [x1, . .,.xs] is Noetherian, the ideal J is generated by finitely many poly- nomials, say, fm (1 m k). We can thus choose an integer N which is strictly greater than the di (1 i s) and the degree of any nontrivial monomial in any fm (1 m k). Lemma 5.2. For each j > N, the natural map j+1 j Aut H (X) -! Aut H (X) is surjective, where Aut (-) denotes the group of graded -algebra automor- phisms. Proof.Let j be any integer strictly greater than N and let oe be an au- tomorphism of the graded algebra H j (X). We want to show that oe can be lifted to an automorphism of H j+1 (X). Now oe(fm ) is 0 in H j (X) for m = 1, . .,.k. But since each oe(xi) is homogeneous of degree di, each monomial of oe(fm ) = fm (oe(x1), . .,.oe(xs)) still has degree less than N, and in this degree the natural map H j+1 (X) ! H j (X) is an isomorphism. Therefore, there is a well-defined graded algebra map ^oe:H j+1 (X) ! H j+1 (X) satisfying ^oe|H j (X) = oe :H j (X) ! H j (X). We will be done once we show that ^oeis bijective. It is clear that ^oeis surjective because H j+1 (X) is generated as an algebra by the xi (1 i s) and each xi is in the image of ^oe, since this is true for oe. Finally, since H j+1 (X) is a finitely generated -module, the injectivity of ^oenow follows from its surjectivity. Therefore, ^oeis an automorphism of H j+1 (X) and is a lift of oe. __ This finishes the proof of Lemma 5.2. |__| 20 Proof of Theorem 1.9.This is now an immediate consequence of Lemma 5.1, Lemma 5.2, and the analog of Wilkerson's Theorem 3.1 in the context __ of (non-negatively) graded -algebras. |__| Acknowledgements Most of the results in this paper come from the author's Ph.D. thesis at MIT. The author would like to express his sincerest gratitude to his advisor, Professor Haynes Miller, for superb guidance and encouragement. References [1]J. F. Adams, Localisation and completion, University of Chicago Press, Chicago, Ill., 1975. [2]M. F. Atiyah, Power operations in K-theory, Quart. J. Math. Oxford 17 (1966) 163-193. [3]M. F. Atiyah and D. O. Tall, Group representations, ~-rings and the J-homomorphism, Topology 8 (1969) 253-297. [4]J. Michael Boardman, Conditionally convergent spectral sequences, Contemp. Math. 239, Amer. Math. Soc., Providence, RI, 1999, p. 49-84. [5]I. Bokor, On the connection between the topological genus of certain polyhedra and the algebraic genus of their Hilton-Hopf quadratic forms, Publ. Mat. 34 (1990) 323-333. [6]A. K. Bousfield and D. M. Kan, Homotopy limits, completions and localizations, Lecture Notes in Mathematics, Vol. 304, Springer-Verlag, Berlin-New York, 1972. [7]F. J. B. J. Clauwens, Commuting polynomials and ~-ring structures on Z[x], J. Pure Appl. Algebra 95 (1994) 261-269. [8]E. Dror, W. G. Dwyer, and D. M. Kan, Self-homotopy equivalences of Postnikov conjugates, Proc. Amer. Math. Soc. 74 (1979) 183-186. 21 [9]J. R. Harper and J. Roitberg, Phantom maps and spaces of the same n-type for all n, J. Pure Appl. Algebra 80 (1992) 123-137. [10]I. M. James, Multiplication on spheres. II, Trans. Amer. Math. Soc. 84 (1957) 545-558. [11]A. Jeanneret and A. Osse, The K-theory of p-compact groups, Com- ment. Math. Helv. 72 (1997) 556-581. [12]D. C. Johnson and W. S. Wilson, BP operations and Morava's extraor- dinary K-theories, Math. Z. 144 (1975) 55-75. [13]D. Knutson, ~-rings and the representation theory of the symmetric group, Lecture Notes in Mathematics, Vol. 308. Springer-Verlag, Berlin- New York, 1973. [14]C. A. McGibbon, The Mislin genus of a space, The Hilton Sympo- sium 1993 (Montreal, PQ), 75-102, CRM Proc. Lecture Notes, 6, Amer. Math. Soc., Providence, RI, 1994. [15]C. A. McGibbon and J. M. Møller, On spaces with the same n-type for all n, Topology 31 (1992) 177-201. [16]__-, How can you tell two spaces apart when they have the same n-type for all n?, Adams Memorial Symposium on Algebraic Topology, 1 (Manchester, 1990), 131-143, London Math. Soc. Lecture Note Ser., 175, Cambridge Univ. Press, Cambridge, 1992. [17]J. M. Møller, The normalizer of the Weyl group, Math. Ann. 294 (1992) 59-80. [18]D. Notbohm, Maps between classifying spaces and applications, J. Pure Appl. Algebra 89 (1993) 273-294. [19]D. Notbohm and L. Smith, Fake Lie groups and maximal tori. I - III, Math. Ann. 288 (1990) 637-661, 663-673 and 290 (1991) 629-642. 22 [20]D. L. Rector, Loop structures on the homotopy type of S3, Symposium on Algebraic Topology (Battelle Seattle Res. Center, Seattle, Wash., 1971), pp. 99-105. Lecture Notes in Math., Vol. 249, Springer, Berlin, 1971. [21]__-, Subgroups of finite dimensional topological groups, J. Pure Appl. Algebra 1 (1971) 253-273. [22]Y. Shitanda, Uncountably many loop spaces of the same n-type for all n. I and I, Yokohama Math. J. 41 (1993) 17-24 and 41 (1994) 85-93. [23]__-, Uncountably many infinite loop spaces of the same n-type for all n, Math. J. Okayama Univ. 34 (1992) 217-223. [24]__-, Spaces of the same clone type, J. Math. Soc. Japan 48 (1996) 705-714. [25]C. Wilkerson, Classification of spaces of the same n-type for all n, Proc. Amer. Math. Soc. 60 (1976) 279-285. [26]__-, Applications of minimal simplicial groups, Topology 15 (1976) 111-130. [27]__-, Lambda-rings, binomial domains, and vector bundles over CP (1), Comm. Algebra 10 (1982) 311-328. 23