Alg`ebre homologique/Homological Algebra The 2

Transcription

Alg`ebre homologique/Homological Algebra The 2
Algèbre homologique/Homological Algebra
The 2-torsion in the K-theory of the Integers
Charles Weibel
Abstract — Using recent results of Voevodsky, Suslin-Voevodsky and BlochLichtenbaum, we completely determine the 2-torsion subgroups of the K-theory of
the integers Z. The result is periodic of order 8, and there are no 2-torsion elements
except those which have been known for over 20 years. There is no 2-torsion except
for the Z/2 summands in degrees 8n + 1 and 8n + 2, the Z/16 in degrees 8n + 3
and the image of the J-homomorphism in degrees 8n + 7. In particular, the 2-part
of ζ(1 − 2n) is twice the 2-part of the ratio |K4n−2 (Z)|/|K4n−1 (Z)| for all n > 0.
This corrects a conjecture of Lichtenbaum.
La torsion 2-primaire de la K-théorie de Z
Résumé — A partir de résultats récents de Voevodsky, Suslin-Voevodsky et
Bloch-Lichtenbaum, nous déterminons la torsion 2-primaire de la K-théorie de Z,
qui est périodique de période 8. Il n’existe pas d’autres éléments 2-primaires que
ceux connus depuis 20 ans. La torsion 2-primaire se réduit à Z/2 en degrés 8n + 1
et 8n + 2, à Z/16 en degré 8n + 3, et à l’image de l’homomorphisme J en degré
8n + 7. La partie 2-primaire de ζ(1 − 2n) est donc le double de la partie 2-primaire
de |K4n−2 (Z)|/|K4n−1 (Z)| pour tout n > 0. Ce résultat corrige un conjecture de
Lichtenbaum.
K0 (Z) = Z
K1 (Z) = Z/2
K2 (Z) = Z/2
K3 (Z) = Z/48
K4 (Z) = 0
K5 (Z) = Z ⊕ (3-torsion)
K6 (Z) = (odd)
K7 (Z) = Z/240 ⊕ (odd)
K8n (Z) = (odd) for n ≥ 1
K8n+1 (Z) = Z ⊕ Z/2 ⊕ (odd) for n ≥ 1
K8n+2 (Z) = Z/2 ⊕ (odd)
K8n+3 (Z) = Z/16 ⊕ (odd)
K8n+4 (Z) = (odd)
K8n+5 (Z) = Z ⊕ (odd)
K8n+6 (Z) = (odd)
K8n+7 (Z) = (Z/wi ) ⊕ (odd), i = 4(n + 1).
Table 1. The K-theory of Z. La K-théorie de Z.
Version française abrégée – Les résultats récents de Voevodsky [V] nous permettent de déterminer la torsion 2-primaire de la K-théorie de Z, qui est périodique
de période 8. Il n’existe pas d’autres éléments 2-primaires que ceux connus depuis
20 ans [Q] [Br]. Comme les groupes Kj (Z) sont de type fini [Q1], et que leur rang
est connu par [Bo], nous pouvons énoncer le résultat suivant.
Théorème 1. Les groupes K∗ (Z) sont données dans la Table 1 ci-dessus, où wi
dénote la plus grande puissance de 2 qui divise 4i, et le symbole “(odd)” dans la
table représente un groupe fini d’ordre impair.
Les premières assertions non déjà connues sont que K4 (Z) = 0 et que la seule
torsion dans K5 (Z) est 3-primaire. Cela découle des résultats de Lee-Szczarba et
Soulé [Sou], de Rognes [R], et de notre calcul qui montre qu’il n’existe pas de torsion
2-primaire.
Comme la partie 2-primaire de ζ(1 − 2n) = (−1)n Bn /2n est 12 wi , cela implique:
Typeset by AMS-TEX
1
2
Corollaire 2. La partie 2-primaire de la formule conjecturée par Lichtenbaum [L]
est correcte à un facteur 2 près:
1
1
|K4n−2 (Z)|
=
= ζ(1 − 2n)
|K4n−1 (Z)|
w2n
2
à un facteur impair près.
Pour obtenir ces résultats, nous utilisons la suite spectrale
(†)
p−q
E2pq = Het
(F ; Z/2) ⇒ K−p−q (F ; Z/2)
(p ≤ 0, p ≥ q).
C’est la suite spectrale de Bloch et Lichtenbaum [BL], le terme E2pq étant déduit
de l’isomorphisme
j
Het (F ; Z/2), j ≤ i
i
∼
CH (F, 2i − j; Z/2) =
0,
j > i,
voir les travaux récents de Voevodsky [V] et Suslin [S1] [SV].
Nous commençons par calculer la suite spectrale (†) pour R et pour les corps
i
i
locaux Qp . Comme Het
(Q; Z/2) → Het
(R; Z/2) est un isomorphisme pour i 6= 1, 2
(voir [T]), nous pouvons utiliser ces calculs pour déterminer les différentielles dans
la suite spectrale pour Q.
Théorème 5. Pour n ≥ 0, la K-théorie modulo 2 de Q est donnée par la Table 2
ci-dessous.
La comparaison des suites exactes de localisation en K-théorie et en cohomologie
étale, pour Q et pour les Qp , donne le résultat suivant.
Théorème 7. Pour n ≥ 0, la K-théorie modulo 2 de Z[ 12 ] est donnée par la Table 3
ci-dessous.
∼ Kj (Z[ 1 ]; Z/2) pour j ≥ 2.
Ce résultat implique le Théorème 1, car Kj (Z; Z/2) =
2
En effet, tous les termes de la Table 3 proviennent des facteurs connus de K∗ (Z).
Le Théorème 1 implique aussi que Kj (Z[ 12 ]) ⊗ Ẑ2 est la K-théorie étale dyadique
Kjet (Z[ 12 ]). Bökstedt, et plus tard Dwyer et Friedlander [DF1, Proposition 4.2], ont
remarqué que cela implique le résultat suivant.
Corollaire 8. Soit K(R)2̂ le complété dyadique de l’espace de la K-théorie K(R)
d’un anneau R. On fixe un plongement de Ẑ3 dans le corps C des nombres complexes. Alors le carré suivant est homotopiquement cartésien:
K(Z[ 12 ])2̂ −−−−→


y
K(Ẑ3 )2̂ −−−−→
K(R)2̂


y
K(C)2̂ .
The recent spectacular results of Voevodsky [V] allow us to completely determine
the 2-torsion in the K-theory groups Kj (Z) associated to the integers Z. The result
is that there are no new elements; all the 2-torsion has been known for over 20 years.
Since each group Kj (Z) is finitely generated by [Q1], and their ranks were computed
by Borel [Bo], our 2-primary calculation yields the following assertion.
3
Theorem 1. The groups K∗ (Z) are given in Table 1. Here wi = wi (Q) is defined
(for even i) to be the largest power of 2 dividing 4i. The symbol “(odd)” in the table
denotes a finite group of odd order.
Of course there is nothing new up to K3 (Z). The first new entry is the assertion
that K4 (Z) = 0 and that K5 (Z) has at most 3-torsion. This follows from our
calculation that there is no 2-torsion in K4 (Z) or K5 (Z), because Lee-Szczarba and
Soulé [Sou] showed that there is no p-torsion in K4 (Z) or K5 (Z) for p > 3, and
Rognes [R] has shown that there is no 3-torsion in K4 (Z) either.
Essentially, the 2-torsion summands of Kj (Z) all come from the stable homotopy
groups πjs of spheres via the natural map πjs → Kj (Z), and were found by Quillen
in [Q2]. When j is 8n + 1 or 8n + 2, the Z/2-summand is generated by the image
of Adams’ element µj . When j = 8n + 3 the 2-Sylow subgroup of J(πj O) is cyclic
of order 8, and is contained in a Z/16 summand of Kj (Z); this result is due to
Browder [Br], although Quillen proved in [Q2] that J(πj O) injects into Kj (Z).
When j = 8n + 7 there is a cyclic summand of Kj (Z) isomorphic to the subgroup
J(πj O) of πjs . The 2-Sylow subgroup of J(π8n+7 O) is isomorphic to Z/wi , where
i = 4(n + 1) and wi is defined in Theorem 1; see [C, p. 284].
The number wi also arises in the following ways; see [W, Theorem 6.7]. If i = 2n
0
∼
then: (1) for ν 0, Het
(Q, µ⊗i
2ν ) = Z/wi ; (2) the 2-part of the denominator of
1
n
ζ(1 − 2n) = (−1) Bn /2n is 2 w2n . The first of these is used to detect the image of
J, while the second lets us relate the orders of K-groups to the zeta function.
Corollary 2. The 2-part of the formula conjectured by Lichtenbaum in [L] holds
up to a factor of 2:
|K4n−2 (Z)|
1
1
≡
≡ ζ(1 − 2n)
|K4n−1 (Z)|
w2n
2
(up to odd torsion).
We now turn to the derivation of these results. Voevodsky proved [V] that the
j
Galois symbol KjM (F )/2KjM (F ) → Het
(F ; Z/2) is an isomorphism for every field
F of characteristic 6= 2 and every j. Suslin and Voevodsky have shown in [S1] and
[SV] that this implies that Bloch’s higher Chow groups (with coefficients mod 2)
[B] for a field F are:
i
CH (F, 2i − j; Z/2) =
j
Het
(F ; Z/2), j ≤ i
0,
j > i.
Bloch and Lichtenbaum have defined a third quadrant spectral sequence in [BL],
converging to groups K∗ (F ). Using K-theory with coefficients modulo 2 in their
construction yields a similar third quadrant spectral sequence; using the SuslinVoevodsky formula above, we may rewrite it as
(†)
p−q
E2pq = Het
(F ; Z/2) ⇒ K−p−q (F ; Z/2)
(p ≤ 0, p ≥ q).
Note that every column E2p∗ of (†) equals the mod 2 étale cohomology of F .
4
Example 3 (Local fields). Let Fv be a local field with residue field kv of chari
acteristic 6= 2. By [T] we know that Het
(Fv ; Z/2) is: Z/2 for i = 0, 2, (Z/2)2
for i = 1, and 0 for i > 2. Thus the spectral sequence (†) degenerates to yield
the well-known calculation that for j > 0 the group Kj (Fv ; Z/2) is Z/2 ⊕ Z/2 for
all j > 0. Comparing with an unramified extension of Fv , we see that the tame
symbol K2i (Fv ; Z/2) → K2i−1 (kv ; Z/2) = Z/2 is the projection onto the summand
2
Het
(Fv ; Z/2), while the tame symbol K2i+1 (Fv ; Z/2) → K2i (kv ; Z/2) = Z/2 may
1
0
be identified with the canonical map Het
(Fv ; Z/2) → Het
(kv ; Z/2).
When F = R, every column of the spectral sequence (†) is the polynomial ring
∗
1
Het
(R; Z/2) = Z/2[η], where η is the nonzero element in Het
(R; Z/2). From [S] we
know that Kj (R; Z/2) is: Z/2 for j ≡ 0, 1, 3, 4 (mod 8); Z/4 for j ≡ 2 (mod 8); 0
for j ≡ 5, 6, 7 (mod 8). Thus the d2 differential out of E2−2,−2 must be nonzero.
Now multiplication by η commutes with the differentials, and induction yields the
following result.
Proposition 4. When F = R, the spectral sequence (†) degenerates at E 3 = E∞
to the mod 2 K-theory of R. Indeed, the differentials E2−p,q → E22−p,q−1 are isomorphisms for all p ≡ 2, 3 (mod 4) and all q ≤ −p.
e 1 (Q; Z/2)
We are now ready to attack the rational numbers Q. Let us write H
1
1
2
e (Q; Z/2) for the
for the kernel of the surjection Het (Q; Z/2) → Het (R; Z/2) and H
2
2
kernel of the surjection Het
(Q; Z/2) → Het
(R; Z/2).
Theorem 5. For n ≥ 0, the mod 2 K-theory of Q is given by Table 2. Here the
notation H o Z/2 denotes the nontrivial extension of Z/2 by the group H.
1
K8n+1 (Q; Z/2) = Het
(Q; Z/2)
1
K8n+3 (Q; Z/2) = Het
(Q; Z/2)
1
e
K8n+5 (Q; Z/2) = H (Q; Z/2)
e 1 (Q; Z/2)
K8n+7 (Q; Z/2) = H
2
K8n+2 (Q; Z/2) = Het
(Q; Z/2) o Z/2
2
K8n+4 (Q; Z/2) = Het
(Q; Z/2)
2
e
K8n+6 (Q; Z/2) = H (Q; Z/2)
e 2 (Q; Z/2) ⊕ Z/2
K8n+8 (Q; Z/2) = H
Table 2. The mod 2 K-theory of Q. La K-théorie modulo 2 de Q.
i
i
Proof. We know from [T] that the natural map Het
(Q; Z/2) → Het
(R; Z/2) is an
isomorphism for i 6= 1, 2. The differentials in (†) for F = Q are thus determined by
the differentials when F = R, and we can read off the result from E3 = E∞ . The
two extension problems are resolved by comparison with F = R and F = Qp ; cf.
[Br].
In order to pass from Q to Z[ 21 ] we need to break up the localization sequence:
Lemma 6. For each j > 0 there is a short exact sequence
1
∂
0 → Kj (Z[ ]; Z/2) → Kj (Q; Z/2) −
→ ⊕p6=2 Kj−1 (Z/p; Z/2) → 0.
2
Proof. Fix an even number j. Then the Universal Coefficient Theorem implies that
Kj (Q; Z/2) maps onto ⊕Kj−1 (Z/p; Z/2). In addition, since Kj (Z/p; Z/2) = Z/2
for all p 6= 2, the natural map from ⊕p6=2 Kj (Z/p; Z/2) = ⊕p6=2 Z/2 to Kj (Z[ 12 ]; Z/2)
factors through Pic(Z[ 12 ])/2 = 0.
5
Theorem 7. For n ≥ 0, the mod 2 K-theory of Z is given by Table 3.
K8n+1 (Z[ 12 ]; Z/2) = (Z/2)2
K8n+3 (Z[ 12 ]; Z/2) = (Z/2)2
K8n+5 (Z[ 12 ]; Z/2) = Z/2
K8n+7 (Z[ 12 ]; Z/2) = Z/2
K8n+2 (Z[ 12 ]; Z/2) = Z/4
K8n+4 (Z[ 12 ]; Z/2) = Z/2
K8n+6 (Z[ 12 ]; Z/2) = 0
K8n+8 (Z[ 12 ]; Z/2) = (Z/2)
Table 3. The mod 2 K-theory of Z[ 21 ]. La K-théorie modulo 2 de Z[ 21 ].
1
1
Proof. Since Het
(Q; Z/2) = Q∗ /Q∗2 and Het
(Z[ 12 ]; Z/2) = Z[ 12 ]∗ /Z[ 21 ]∗2 ∼
= (Z/2)2 ,
we have a short exact sequence
1
1
1
0 → Het
(Z[ ]; Z/2) → Het
(Q; Z/2) → ⊕p6=2 Z/2 → 0.
2
Comparing with the direct sum of the corresponding sequences for the local fields
Qp allows us to deduce the calculation of Kj (Z[ 12 ]; Z/2) for odd j. The calculation
for j even follows similarly from the commutative diagram (in which the left vertical
isomorphism follows from class field theory).
2
e et
H
(Q; Z/2)

∼
y=
−−−−→
∂
K2i (Q; Z/2)


y
−−−−→ ⊕p6=2 K2i−1 (Z/p; Z/2)

=
y
∂
2
⊕p6=2 Het
(Qp ; Z/2) −−−−→ ⊕p6=2 K2i (Qp ; Z/2) −−−−→ ⊕p6=2 K2i−1 (Z/p; Z/2)
Theorem 7 implies Theorem 1. Indeed, Kj (Z; Z/2) ∼
= Kj (Z[ 12 ]; Z/2) for j ≥ 2,
and the known summands of Kj (Z) account for all of Kj (Z[ 12 ]; Z/2) in this range.
Theorem 1 also shows that Kj (Z[ 12 ])⊗ Ẑ2 is the 2-adic étale K-theory Kjet (Z[ 12 ])2̂ .
As observed by Bökstedt, and again by Dwyer and Friedlander in [DF1, Proposition 4.2], it also implies the following topological result. Let K(R)2̂ denote the
2-completion of the algebraic K-theory space K(R) of a ring R.
Corollary 8. Choose an embedding of Ẑ3 in the complex numbers C representing
the Brauer lifting of K(Z/3) → BU . Then the square
Z[ 12 ] −−−−→


y
R


y
Ẑ3 −−−−→ C
induces a homotopy cartesian square upon 2-adic completion.
K(Z[ 12 ])2̂ −−−−→


y
K(Ẑ3 )2̂ −−−−→
K(R)2̂


y
K(C)2̂
6
Example 9. Another case when the spectral sequence (†) degenerates is when F
2
is a totally imaginary number field. In this case Kj (F ; Z/2) is Z/2 ⊕ Het
(F ; Z/2)
1
for j > 0 even, and Het (F ; Z/2) for j odd. That is, Kj (F ; Z/2) equals the étale
K-theory Kjet (F ; Z/2) of Dwyer and Friedlander [DF]. However, the results of [DF]
√
do not apply here unless F contains −1.
When the primes over 2 generate the class group of F , the methods used for
Z allow us to calculate
√ the K-theory of the ring of integers in F . For example,
when F = Q(i), i = −1, we obtain the following result. Recall that the groups
K0 (Z[i]) = Z, K1 (Z[i]) = Z/4, K2 (Z[i]) = 0 and K3 (Z[i]) = Z ⊕ Z/24 are known.
Theorem 10. For all n ≥ 2, K2n (Z[i]) is a finite group of odd order, while
K2n−1 (Z[i]) = Z ⊕ Z/wn ⊕ (odd).
As in Theorem 1, wn = wn (Q(i)) denotes the largest power of 2 dividing 4n. (For
example, if n is odd then wn = 4.)
References
[B]
[BL]
[Bo]
[Br]
[DF]
[DF1]
[L]
[Q]
[R]
[Sou]
[S]
[S1]
[SV]
[T]
[V]
[W]
S. Bloch, Algebraic Cycles and higher K-theory, Advance Math. 61 (1986), 267–304.
S. Bloch and S. Lichtenbaum, A spectral sequence for motivic cohomology, preprint (1994).
A. Borel, Cohomologie réele stable des groupes S-arithmétiques classiques, C.R. Acad. Sci.
Paris 271 (1970), A1156–A1158.
W. Browder, Algebraic K-theory with coefficients Z/p, Lecture Notes in Math., vol. 657,
Springer-Verlag, 1978, pp. 40–84.
W. Dwyer and E. Friedlander, Algebraic and étale K-theory, Trans. AMS 292 (1985),
247–280.
W. Dwyer and E. Friedlander, Conjectural calculations of general linear group homology,
AMS Contemp. Math. 55 (1986), 135–147.
S. Lichtenbaum, Values of zeta-functions, étale cohomology, and algebraic K-theory, Lecture Notes in Math., vol. 342, Springer-Verlag, 1973, pp. 489–501.
J
D. Quillen, A letter from Quillen to Milnor on Im(πi O −
→ πis → Ki Z), Lecture Notes in
Math., vol. 551, Springer-Verlag, 1976, pp. 182–188.
J. Rognes, K4 (Z) is the trivial group, preprint (1994).
C. Soulé, Addendum to the article “On the torsion in K∗ (Z)” by R. Lee and R. Szczarba,
Duke Math. J. 45 (1978), 131–132.
A. Suslin, On the K-theory of local fields, J. Pure Applied Algebra 34 (1984), 301–318.
A. Suslin, Higher Chow groups and étale cohomology, preprint (1994).
A. Suslin and V. Voevodsky, The Bloch-Kato conjecture and motivic cohomology with
finite coefficients, preprint (1995).
J. Tate, Duality Theorems in Galois cohomology over number fields, Proc. ICM 1962,
Stockholm, 1963, pp. 288–295.
V. Voevodsky, The Milnor Conjectures, In preparation.
C. Weibel, Étale Chern classes at the prime 2, Algebraic K-theory and Algebraic Topology
(P. Goerss and J.F. Jardine, eds.), NATO ASI Series C, vol. 407, Kluwer, 1993, pp. 249–
286.
Department of Mathematics, Rutgers University, New Brunswick, N.J. 08903 U.S.A.
E-mail address: [email protected]