and assume
n | d (i) 1 (a j )
for alli
. This holdsforj = 1
. Thena j ∈ Z[[t]] ⊗ L j
,and we onsider the image of under
d (i) 1
:Z[[t]] ⊗ R L j d (i) 1
−→ Z[[t]] (i) ⊗ R M j ,
f j x j + . . . + f r x r 7→ f j (x j − y j ) + . . . + f r (x r − y r ).
Byassumption,
f j (x j − y j ) + . . . + f r (x r − y r ) = nb
forsomeb
. Nowwe usethe projetion
π j : M j −→ M j /M j+1
, whihindues amapZ[[t]] (i) ⊗ R M j π j
−→ Z[[t]] (i) ⊗ R M j /M j+1 , f j (x j − y j ) + . . . + f r (x r − y r ) = nb 7→ f j (x j − y j ) = n · π j (b).
We wish to map
M j /M j+1 −→ Z
. For now, assumer > 1
. Ifj > 1
, we usethe map
q
from (69), (70). Sineq(x j − y j ) = 1
forj > 1
, we getZ[[t]] (i) ⊗ R M j /M j+1
−→ q Z[[t]] (i) ⊗ R Z,
f j (x j − y j ) = n · π j (b) 7→ f j = n · qπ j (b).
(72)If
j = 1
,we use the well-dened mapq 1 (x) = 1
,q 1 (y) = 0
,and get the sameresult. The onlusion is that
f j (t) ∈ n · Z[[t]] (i) ⊗ R Z
for alli
. By Lemma6.7this implies
f j (t) ∈ nZ[[t]]
. Sinea j = f j (t)x j + a j+1
, indutively we getn | d (i) 1 (a j+1 )
for alli
. This nishestheindution step. Thisindution showsthat
n | f j (t)
for allj = 1, . . . , r
, soa ∈ nK hS 1 1 (Σ F 0 , ∗ )
.Now take
r = 1
. Thenj = 1
. We use the mapq : M 1 /M 2 −→ Z
from(69). Then in (72), we get instead
2f 1 (t) ∈ n · Z[[t]] ⊗ R Z
. By Lemma 6.7,2f 1 (t) ∈ nZ[[t]]
, and2a ∈ nK hS 1 1 (Σ F 0 , ∗ )
.Lemma 6.14. There is an
S 1
transfer mapτ
onK
-theory, whih ts intothe followingexat sequene,
−→ K 0 (X) −→ τ K hS 1 1 (X) −→ ϕ K hS 1 1 (X) −→ q K 1 (X) −→ τ K hS 0 1 (X) −→
where
K ∗ (BS 1 ) = Z[[t]]
, and the mapϕ
is multipliation by− t
.Proof. Let
T −→ BS 1
denote the standard omplex line bundle, as usual.Let
p : ES 1 × S 1 X −→ BS 1
,beprojetionontherstfator, andletξ = p ∗ T
denotethe pullbak. As in(48), we use the ober sequene,
S(ξ) −→ D(ξ) −→ T h(ξ).
As shown in (49),
S(ξ) ∼ = ES 1 × X ≃ X
. The long exat sequene onK
-theory beomes, using the Thom isomorphism,f. [Atiyah℄ Cor. 2.7.3,K ∗− 1 (X) −→ δ K ∗ (ES 1 × S 1 X) −→ ϕ K ∗ (ES 1 × S 1 X) −→ K ∗ (X) −→ δ
The map
ϕ
is given by multipliation withΛ − 1 (T ) = 1 − T = − t
, sineT
isalinebundle. Wedenethe
S 1
transfer mapτ
tobethe boundary mapδ
inthe long exat sequene.
By exatness, Im
(τ ) =
Ker(ϕ)
, and sowe willneed the kernel oft
:Lemma 6.15. The kernel of the map givenby multipliation by
t
,t : Z[[t]] (k) ⊗ R Z[x]/x 2 −→ Z[[t]] (k) ⊗ R Z[x]/x 2
is
Zp k − 1 (t)x
, where(t + 1) k − 1 = tp k − 1 (t)
.Proof. Firstwerelate thekernelof
t
tothe kernelofu : M −→ M
(this partholds for all
r
). ReallR = R(S 1 ) = Z[U, U − 1 ]
, and letu = U − 1
. ThenM
is anR
-module byu 7→ (x − y)/(1 + y)
, andZ[[t]] (k)
is anR
-module byu 7→ (t + 1) k − 1
. Consider the exat sequene0 // Z[[t]] t // Z[[t]] // Z // 0.
Tensoring with
M
overR
yieldsthe exat sequene0 //
TorR 1 (Z, M ) // Z[[t]] (k) ⊗ R M t // Z[[t]] (k) ⊗ R M
To ompute Ker
(t) ∼ =
TorR 1 (Z, M )
, we use the following free resolution ofZ
over
R
:0 // R u // R // Z // 0.
Again, we tensorover
R
withM
and nd0 //
TorR 1 (Z, M) // R ⊗ R M u // R ⊗ R M // Z[[t]] (k) ⊗ R Z // 0.
so Tor
R
1 (Z, M ) ∼ =
Ker(u)
. All we need to know is how to translate fromKer
(u)
toKer(t)
. The followingdiagram,0 //
Ker(u) //
R ⊗ R M u //
p k− 1 (t) ⊗
idR ⊗ R M //
Z[[t]] (k) ⊗ R Z
id
0 //
Ker(t) // Z[[t]] (k) ⊗ R M t // Z[[t]] (k) ⊗ R M // Z[[t]] (k) ⊗ R Z
isommutative,sine
tp k − 1 (t) = (t + 1) k − 1 = u
. Fromthis diagram,weseethat Ker
(t) = p k − 1 (t)
Ker(u)
.So all that remains is to determine Ker
(u)
. This an be done for anyr
,but it is espeially easy when
r = 1
, andM = Z[x]/x 2
, whereu1 = 2x
andux = 0
. Clearly Ker(u) = Zx
, and so Ker(t) = Zp k − 1 (t)x
.Wean now provethe MainTheorem inase
r = 1
:Proof of Theorem6.13. Bythe exat sequene
K hS 1 1 (Σ F 0 , ∗ ) −→ δ K ˜ hS 1 1 ( F ∞ / F 0 ) −→ K hS 1 1 ( F ∞ ) −→ 0,
we see that
K hS 1 1 (LCP 1 ) = K hS 1 1 ( F ∞ )
is isomorphi to the okernel Cok(δ)
of
δ
. Siner = 1
,K hS 1 1 (Σ F 0 , ∗ ) = Z[[t]] · h
, soletf(t) ∈ Z[[t]]
begiven, andassumethat
δ(f (t)h)
is divisibleby2
. We willshow thatthis impliesf(t)
isdivisible by
2
,meaning that there isno2
-torsion inCok(δ)
.For ontradition, assume that
f (t)
is not divisible by2
. Then, withoutloss of generality,
f(t)
has the formt l g(t)
, whereg(t) = 1 + tp(t)
for somep(t) ∈ Z[[t]]
. Herel
is the rst exponentinf(t)
with anodd oeient,andso
2 | δ(f(t)h)
ifandonlyif2 | δ(t l g(t)h)
. Theng (t)
isaunitinZ[[t]]
,sosineδ
is aZ[[t]]
-module homomorphism,2 | δ(t l g (t)h)
if and only if2 | δ(t l h)
.We haveshown that if
δ(f (t)h)
is divisibleby2
,butf(t)
is not divisible by2
,thenδ(t N − 1 h)
is alsodivisible by2
forallN > l
.We will now show that this leads to a ontradition if
N = 2 n > l
.Consider the omposite map, whih we all
d (N) 2
,K hS 1 1 (Σ F 0 , ∗ ) δ // K ˜ hS 1 1 ( F ∞ / F 0 ) // K ˜ hS 1 1 ( F 2N / F 0 ) P
∗
N // K ˜ hS 1 1 ( F 2 (N ) / F 0 )
Then
d (N) 2 (t N − 1 h)
is divisible by2
, sineδ(t N − 1 h)
is. We will investigated (N) 2 (t N − 1 h)
via the following diagram:Σ F 0 oo ( F 1 / F 0 ) (2N)
'' O O O O O O O O O O O
( F 1 / F 0 ) (N) //
OO
( F 2 / F 0 ) (N ) //
gg OOO
OOO OOO OOO O
( F 2 / F 1 ) (N)
The maps into
Σ F 0
are the ones induing the various dierentials in the Morse spetral sequenes. The map( F 1 / F 0 ) (2N) −→ ( F 2 / F 1 ) (N )
is simplythe omposite of the two other maps inthe triangle
( F 1 / F 0 ) (2N) P
(N)
2 // ( F 2 / F 0 ) (N) // ( F 2 / F 1 ) (N ) .
On
S 1
-equivariantK
-theory this beomesK hS 1 1 (Σ F 0 ) d
(2N) 1 //
d (N) 1
d (N) 2
))
S S S S S S S S S S S S S S S
K ˜ hS 1 1 (( F 1 / F 0 ) (2N) )
K ˜ hS 1 1 (( F 1 / F 0 ) (N) ) oo i K ˜ hS 1 1 (( F 2 / F 0 ) (N) )
k
OO
K ˜ hS 1 1 (( F 2 / F 1 ) (N ) )
oo j E N
ii SSS
SSSS SSSS SSSS
(73)
with the lowerrow short exat (
i
surjetive andj
injetive). WhenN = 2 n
,we have
p N − 1 (t) = t − 1 ((t + 1) 2 n − 1) = t N − 1 + 2q(t),
for some polynomial
q(t)
, sine allbinomial oeients2 j n
are divisible by
2
forj 6 = 0, 2 n
. Sinewe havededued thatd (N 2 ) (t N − 1 h)
isdivisible by2
, wetherefore get
d (N) 2 (p N − 1 (t)h)
is also divisible by2
, sayd (N) 2 (p N − 1 (t)h) = 2a
forsome
a
inK ˜ hS 1 1 (( F 2 / F 0 ) (N) )
. ByLemma5.3weseethatd (N) 1 (p N − 1 (t)h) = 2p N − 1 (t)x
. Sine the diagram(73) is ommutative, we geti(a) = p N − 1 (t)x
,sine the group
K ˜ hS 1 1 (( F 1 / F 0 ) (N) )
istorsion-free, see Lemma6.4.We now use the
S 1
transfer, see Lemma 6.14. We an hoose a transferlass
e ∈ K 1 ( F 1 / F 0 )
, suh thatτ(e) = p N − 1 (t)x
by Lemma 6.15. We anliftthis transfer lassto
e ¯ ∈ K 1 ( F 2 / F 0 )
, soi(τ(¯ e)) = τ (e) = p N − 1 (t)x
. Thuswe have an element
w = a − τ(¯ e) ∈ K ˜ hS 1 1 (( F 2 / F 0 ) (N ) )
withi(w) = 0
. Byexatness of thelowerrowin(73), thereisanelement
z ∈ K ˜ hS 1 1 (( F 2 / F 1 ) (N) )
with
j(z) = w
. By ommutativity of (73), we getE N (z) = k(w) = k(a − τ(¯ e)) = k(a) − k(τ(¯ e)),
so let us ompute this. Sine
2a = d (N) 2 (p N − 1 h)
, we see thatk(2a) =
d (2N 1 ) (p N − 1 h) = 2p N − 1 x
, and sineK ˜ hS 1 1 (( F 1 / F 0 ) (2N) )
is torsion-free,k(a) =
p N − 1 x
. Butk(τ (¯ e))
isin the image of the transfer map, so by Lemma 6.15,k(τ (¯ e)) = mp 2N − 1 (t)x
for somem ∈ Z
. In onlusion,E N (z) = k(w) = (p N − 1 (t) − mp 2N − 1 (t))x.
(74)Toinvestigatethis equality,wewillneed touse
F 2
-oeients,and todeter-mine the map
E N
. This isdone in the followinglemmas:Lemma 6.16. As
K ∗ (BS 1 ) = Z[[t]]
-modules,K ˜ hS 1 1 (( F 1 / F 0 ) (2 k ) ; F 2 ) ∼ = (F 2 [t]/t 2 k )1 ⊕ (F 2 [t]/t 2 k )x.
Proof. As explainedin the beginning,
K ˜ hS 1 1 (( F 1 / F 0 ) (2 k ) ; F 2 ) ∼ = Z[[t]] (2 k ) ⊗ R M ⊗ Z F 2 ,
where
M = Z[x]/x 2
,andu1 = 2x
,ux = 0
. SoweseethatM ⊗ Z F 2 = F 2 ⊕ F 2
istrivial asan
R = Z[U, U − 1 ]
-module. SoZ[[t]] (2 k ) ⊗ R M ⊗ Z F 2 = (Z[[t]] (2 k ) ⊗ R F 2 )1 ⊕ (Z[[t]] (2 k ) ⊗ R F 2 )x.
On
Z[[t]] (2 k )
,u
atsas(t + 1) 2 k − 1 ≡ t 2 k (
mod2)
. Therefore,Z[[t]] (2 k ) ⊗ R F 2 = F 2 [t]/t 2 k
. This shows the Lemma.Lemma 6.17. The map
E N
is multipliation by1 − (t + 1) N
.Proof. Wemust determinethemap induedby
( F 1 / F 0 ) (2N ) −→ ( F 2 / F 1 ) (N )
,whih is the
(N )
-twistingof the omposite map( F 1 / F 0 ) (2) −→ F P 2 2 / F 0 −→ F 2 / F 1 .
We willrst study this untwistedase. The indued map, allit
E
, isgivenasfollows:
K ˜ hS 1 1 ( F 2 / F 1 ) ∼ = //
K ˜ hS 1 1 (T h(µ − 2 )) Φ 2 //
K hS 0 1 (G 2 (r))
E
˜
K hS 1 1 (( F 2 / F 1 ) (2) ) ∼ = // K ˜ hS 1 1 (T h((µ − 1 ) (2) )) Φ 1 // K hS 0 1 (G(r) (2) )
wherethe rstisomorphismsare Morse theory,and the
Φ j
denotethe Thomisomorphisms (the index indiates whih negative bundle). Also,
G 2 (r)
isthe geodesis of length2, whih asan
S 1
-spae isisomorphi toG(r) (2)
,the(2)
-twistedspae of simple losedgeodesis of length 1.This isaspeialase of thefollowinggeneralsituation: Forabundle and
a subbundle,
ξ ⊆ η
, overa spaeX
, the following diagramommutesK ˜ ∗ (T h(η)) // K ˜ ∗ (T h(ξ))
K ∗ (X)
Φ η ∼ =
OO
Λ // K ∗ (X)
Φ ξ ∼ =
OO
The vertial maps are the Thom isomorphisms. Then the indued map on
K
-theory of the base spae is given by multipliation by the Euler lassΛ = Λ − 1 (η − ξ)
of the bundleη − ξ
, i.e. the (orthogonal) omplement ofξ
inside
η
.Sowe need the negative bundle
µ − 2 = ε 2 ⊕ ν 2
overG 2 (r)
,see Proposition 4.2. I have givenε
andν
anindex, so one an distinguish between them forµ − 2
andµ − 1
. Now(µ − 1 ) (2)
isnotapriori asubbundleofµ − 2
, butsineµ − 1 = ε 1
wherethe
S 1
ationistrivialonthebers,weseethat(µ − 1 ) (2) = ε 2
asbundlesover
G(r) (2) ∼ = G 2 (r)
, sothatµ − 2 − (µ − 1 ) (2) = ν 2 = ν
, whereν
is the omplexbundlefound inthe proof ofProposition4.2. From here, weknowthatfor a
geodesi
f
of length 2, parametrizedasf(t)
fort ∈ [0, 1]
, the berofν
overf
is given byg(t)if ′ (t)
fort ∈ [0, 1]
, whereg ∈
spanR { cos(2πt), sin(2πt) }
.The rotationation of
S 1
isgiven by, forθ ∈ [0, 1]
:θ ∗ (f (t), cos(2πt)if ′ (t)) = (f (t − θ), cos(2πt − 2πθ)if ′ (t − θ))
and similarly for
sin(2πt)
. The omplex strutureJ
found in the proof ofProposition4.2 is
J (cos(2πt)) = sin(2πt)
.Nowletusomparethis tothebundle
T
,i.e. thebundleomingfromthestandard representation of
S 1
. Ignoring theS 1
ation,T
is just a produtbundle
G 2 (r) × C
. TheS 1
ation ofθ ∈ [0, 1]
is given byθ ∗ (f (t), c) = (f(t − θ), e 2πiθ c),
fort ∈ [0, 1].
Wewillnow onstrut a map
ϕ : T −→ ν
, given byϕ(f, c)(t) = (f(t), c cos(2πt)if ′ (t)).
We hek that this is
S 1
-equivariant, i.e. that the following diagramom-mutes (it sues to hek
c = 1
):T ϕ //
θ ∗
ν
θ ∗
(f (t), 1) ϕ //
_
θ ∗
(f (t), cos(2πt)if ′ (t))
_
θ ∗ ?
T ϕ // ν (f (t − θ), e 2πiθ ) ϕ // (f(t − θ), e 2πiθ cos(2πt)if ′ (t − θ))
This ommutes, sine
e 2πiθ = cos(2πθ) + i sin(2πθ)
is multipliedoncos(2πt)
as
e 2πiθ cos(2πt) = cos(2πθ) cos(2πt) + sin(2πθ)J(cos(2πt))
= cos(2πθ) cos(2πt) + sin(2πθ) sin(2πt)
= cos(2π(t − θ))
by the trigonometri formula. So
ϕ
isS 1
-equivariant. Thenϕ
denes anisomorphism of
S 1
bundles, sine it is learly an isomorphism on the bers.We haveshown
µ − 2 − (µ − 1 ) (2) = ν ∼ = T
.Now let us look at the
(N)
-twisted ase. We get again(µ − 1 ) (2N ) = ε 2N
,and so
(µ − 2 ) (N) − (µ − 1 ) (2N) = ν (N ) ∼ = T (N)
, by the above isomorphism. Now,T (N )
is the bundle withS 1
ationofθ ∈ [0, 1]
given byθ ∗ (f (t), c) = (f (t − θ), (e 2πit ) N c),
fort ∈ [0, 1].
This shows that this is the same bundle as
T N
, so the mapE N
ismulti-pliation by the Euler lass of
T N
, and sine this is a line bundle, we getΛ − 1 (T N ) = 1 − T N = 1 − (t + 1) N
.Using the previous two lemmas, we an now investigate equation (74)
in
K ˜ hS 1 1 (( F 1 / F 0 ) (2N ) ; F 2 )
, whereN = 2 n
. As already noted,p N − 1 (t) ≡ t N − 1 (
mod2)
, and so the left-hand side of (74) is(t N − 1 − mt 2N − 1 )x
mod-ulo 2. The right-hand side is
E N (z) = (1 − (t + 1) 2 n )z ≡ − t 2 n z(
mod2)
.So
(t N − 1 − mt 2N − 1 )x = − t N z ∈ (Z[t]/t 2N )1 ⊕ (Z[t]/t 2N )x
Clearly,thisisimpossible,sinetheterm
t N − 1 x
annotbeanelledby− t N z
in
(Z[t]/t 2N )1 ⊕ (Z[t]/t 2N )x
. This gives a ontradition, so the givenf
westartedwith must be divisible by
2
. This proves the Theorem.Notation
Inthistableanbefoundsomeofthefrequentlyusednotationinthis paper:
≃
(between topologialspaes): homotopy equivalent.F C
orH
.G(r)
The spae of simple parametrizedlosed geodesis onFP r
.Sometimes written
G(HP r )
orG(CP r )
tobe spei.G n (r)
The spae of parametrizedlosed geodesis of lengthn
,an be obtained by iterating
n
times the elements ofG(r)
.∆(r)
The quotientS 1 \ G(r)
under the rotationation ofS 1
.EG
A ontratiblespae with a freeation of the groupG
;unique up to homotopy.
BG EG/G
,the lassifying spaeofG
.X hS 1 ES 1 × S 1 X
, whereX
is anS 1
-spae.K hS ∗ 1 (X) K ∗ (X hS 1 )
.K hS ∗ 1 (X, ∗ )
The relativegroupK ∗ (X hS 1 , BS 1 )
.T
The standard omplex linebundle overBS 1 = CP ∞
, or itspullbak to
X hS 1
underthe map pr1 : ES 1 × S 1 X −→ BS 1
.Also used forthe lass of this bundle inK-theory.
t
the lassT − 1
, seeT
.F n E − 1 (] − ∞ , n 2 ])
,then
th term inthe Morse ltration.µ − n
the negativebundle for the ritial manifoldG n (r)
.Referenes
[Atiyah℄ M. Atiyah, K-theory,Benjamin, New York,1967.
[Atiyah2℄ M. Atiyah, Charaters and ohomology of nite groups, Publ.
Math. InstitutdesHautes ÉtudesSientiques,Paris(1961),247-289.
[Atiyah-Hirzebruh℄ M. Atiyah, F. Hirzebruh,Vetor bundles and
homoge-neousspaes,Proeedings ofSymposiuminPureMathematis,Vol3,
Amer. Math. So., 1961.
[Atiyah-MaDonald℄ M.Atiyah, I.MaDonald,Introdution toommutative
Algebra, Addison-Wesley, 1969.
[Bökstedt-Ottosen℄ M. Bökstedt, I. Ottosen, String Cohomology Groups of
Complex Projetive Spaes, University of Aarhus, Preprint Series no
8, 2006.
[Bökstedt-Ottosen2℄ M. Bökstedt, I. Ottosen, The Suspended Free Loop
Spae of a Symmetri Spae, University of Aarhus, Preprint Series
no 18,2004.
[Borel℄ A. Borel, Sur la ohomologie des espaes brés prinipaux et des
es-paes homogènes de groupes de Lie ompats. Ann. Math. 57, 1953,
p. 115-207.
[Bredon℄ G.Bredon, Introdutionto Compat Transformation Groups,
Aa-demi Press, 1972.
[Freed-Hopkins-Teleman℄ D.Freed, M.Hopkins, C. Teleman, Twisted
equiv-ariant
K
-theory with omplex oeients, Journal of Topology 20081(1), 16-44.
[Hather1℄ A. Hather, Algebrai Topology, Cambridge University Press,
2002.
[Hather2℄ A. Hather, Spetral Sequenes in Algebrai Topology, A.
Hather's homepage (www.math.ornell.edu/
∼
hather/).[Klilngenberg1℄ W. Klingenberg, Letures on losed geodesis, Grundlehren
der Math. Wiss.230, Springer Verlag, 1978.
[Klingenberg2℄ W. Klingenberg, The spae of losed urves on a projetive