Theorem 5.9. There is an isomorphism of spetral sequenes
E ∗ ( M )(LHP r hS 1 ) 1
u
∼ = E ∗ ( M )(LHP r ) ⊗ F p [u, u − 1 ].
when re-indexing the olumns: ltration
pm
goes to ltrationm
form ∈ N
.Note: Thisimpliesthattheloalizedspetralsequene
E ∗ ( M )(LHP r hS 1 ) 1
u
ollapsesfromthe
E p
page, sineE ∗ ( M )(LHP r )
ollapsesfromtheE 1
page.To prove SF(6), we rst reall that by Theorem 5.6, the indued map
i ∗ : E 1
odd( M )(LHP r hS 1 ) −→ E 1
odd( M )(LHP r )
is surjetive. Sine everydif-ferentialin
E s ( M )(LHP r hS 1 )
starting inodd total degree istrivial,the mapi ∗ : E ∞
odd( M )(LHP r hS 1 ) −→ E ∞
odd( M )(LHP r )
is also surjetive. It is agen-eral fat for spetral sequenes that the indued map ontheir limitsisthen
alsosurjetive, and this is easily seen by a ltrationargument. This means
that
i ∗ : H
odd(LHP r hS 1 ) −→ H
odd(LHP r )
is surjetive.We rst prove the Main Theorem for the odd part of the ohomology.
There are two kinds of
F p [u]
generators, torsion and free, and we need to use theS 1
transfer mapτ
tond the rst kind. Leti : LHP r −→ ES 1 × S 1
LHP r = LHP r hS 1
bethe inlusion. Thenit follows from [Bökstedt-Ottosen ℄ Thm. 14.1 thattheS 1
ationdierentiald
is omposed as followsH ∗ +1 (LHP r )
τ
((
Q Q Q Q Q Q Q Q Q Q Q Q Q
d // H ∗ (LHP r )
H ∗ (LHP r hS 1 )
i ∗ n n n n n n n 66 n n
n n n
(56)
In general, for a spae
X
with an ationµ : S 1 × X −→ X
, the mapd
isgiven by
H n+1 (X) −→ H n+1 (S 1 × X) −→ H n+1 (X) ⊕ H n (X)
a 7→ µ ∗ (a) 7→ (a, d(a))
where the last map is the Künneth formula. For ease of referene, in the
LemmabelowIhaveolletedallthefatsIneedabouttheationdierential.
Firstsome notation:
IF = IF (r, p) = { (4r + 2)i + 4j | δ ≤ j ≤ r, 0 ≤ i, p | (r + 1)i + j } \ { 0 } , IT = IT (r, p) = { (4r + 2)i + 4j | δ ≤ j ≤ r, 0 ≤ i, p ∤ (r + 1)i + j } ;
where
δ =
1, p ∤ r + 1
;0, p | r + 1
.Set
IA = IF ∪ IT
. Then denepowerseries byP I (t) =
X ∞ n=0
a n t n ,
wherea n =
1, n ∈ I (r, p)
;0, n / ∈ I (r, p)
. (57)for
I = IF , IT , IA
. By[Bökstedt-Ottosen ℄ Lemma 11.4,IF ∩ IT = ∅
, soweget
P IA = P IF + P IT
. Also notethat by (54),P H
odd(LHP r ) (t) = 1
t P IA (t).
(58)ThefollowingLemmaontheationdierentialisprovedin[Bökstedt-Ottosen℄
lemma11.6.
Lemma5.10(TheAtionDierential). Put
H ∗ = H ∗ (LHP r )
andletk ∈ N
.(i)
Ker(d : H 2k −→ H 2k − 1 )
is either a trivial or a 1-dimensional vetorspae. Itis non-trivial if and only if
2k ∈ IF (r, p)
.(ii)
Im(d : H 2k −→ H 2k − 1 )
is either a trivial or a 1-dimensional vetor spae. Itis non-trivial if and only if2k ∈ IT (r, p)
.(iii)
The okernel of the mapd : M
0 ≤ k ≤ (2r+1)mp − δ
H 2k+2 −→ M
0 ≤ k ≤ (2r+1)mp − δ
H 2k+1
has dimension
rm
ifp ∤ r + 1
, and dimension(r + 1)m
ifp | r + 1
.The next two Lemmasspeify the
F p [u]
generators forH ∗ (LHP r hS 1 ; F p )
:Lemma 5.11. There is a graded subgroup
T ∗ ⊆ H
odd(LHP r hS 1 )
suh that(i) u T ∗ = 0
.(ii)
Therestritedinlusionmapi ∗ | T ∗ : H ∗ (LHP r hS 1 ) | T ∗ −→ H ∗ (LHP r )
isinjetive.
(iii)
The imagei ∗ ( T ∗ ) ⊆ H ∗ (LHP r )
equals the imaged(H ∗ +1 (LHP r )) ⊆ H ∗ (LHP r )
.Proof. We use property
(iii)
toonstrutT ∗
. Wehoose agradedsubgroupT ∗ ⊆ H ∗ +1 (LHP r )
,suh thatd
mapsT ∗
isomorphially ontoImd
. This weandosimplybyliftingeahgeneratorofIm
d ⊆ H ∗ (LHP r )
toH ∗ +1 (LHP r )
.Now we put
T ∗ = τ( T ∗ )
. Then(iii)
follows by onstrution, sinei ∗ ( T ∗ ) = i ∗ ◦ τ ∗ ( T ∗ ) = d( T ∗ )
by the diagram (56). Also(ii)
holds, sinei ∗
restritedto
T ∗
orresponds toi ∗ ◦ τ = d
restrited toT ∗
, and we hoseT ∗
suh thatd
was an isomorphism ofT ∗
onto its image. As for property(i)
, this holdsbeause
uτ = 0
aording to[Bökstedt-Ottosen ℄ Thm. 14.1. This is beause thetransfer mapτ
appears rightaftermultipliationbyu
inthe Gysinexatsequene.
Remark 5.12. Bydenitionof
T ∗
itfollows fromLemma5.10(ii)
thatthenon-trivialpart of
T ∗
sitsindegree2k − 1
ifand onlyif2k ∈ IT (r, p)
. Usingthe notationin (57), we an write down the Poinaréseries of
T ∗
:P T ∗ (t) = 1
t P IT (t).
Lemma 5.13. There isa graded subgroup
U ∗ ⊆ H
odd(LHP r hS 1 )
suh that(i)
The ompositionT ∗ ⊕ U ∗ // H
odd(LHP hS r 1 ) i ∗ // H
odd(LHP r )
isan isomorphism.
(ii)
The restritionU 2i+1 // H 2i+1 (LHP r hS 1 ) j
∗
// H 2i+1 (( F pm ) hS 1 )
is trivial if either
p | r + 1
andi > (2r + 1)pm
, orp ∤ r + 1
andi > (2r + 1)pm − 2
.Proof. Againwe rst speify a subgroup
U ∗ ⊆ H
odd(LHP r )
, by demandingthatitmustbeaomplementarysubgroup of
i ∗ ( T ∗ )
, sothatwe have theF p
vetor spae isomorphism
H
odd(LHP r ) ∼ = i ∗ ( T ∗ ) ⊕ U ∗
. The idea is to ndU ∗ ⊆ H ∗ (LHP r hS 1 )
suh thati ∗
maps it isomorphially toU ∗
. This an bedone sine
i ∗
issurjetive by SF(6).We now use the Gysin sequene, see [Bökstedt-Ottosen℄ Thm. 14.1, to
makethe following diagramwith exat rows:
H 2i − 1 (LHP r hS 1 ) · u //
j ∗
H 2i+1 (LHP hS r 1 )
j ∗
i ∗
// // H 2i+1 (LHP r )
H 2i − 1 (( F pm ) hS 1 ) · u // H 2i+1 (( F pm ) hS 1 ) // H 2i+1 ( F pm )
(59)
Thevertial maps
j ∗
are surjetiveaording toSF(4). BySF(6),the upperhorizontalmap
i ∗
is surjetive.Under the assumptionin
(ii)
, weget fromSF(5)thatH 2i+1 ( F n , F n − 1 ) = E 1 n,2i+1 − n = 0
for0 ≤ n ≤ pm
. Using the long exat sequene for the pair( F n , F n − 1 )
forn = 0, 1, . . . , pm
gives aseries of injetivemaps,H 2i+1 ( F pm ) ֒ → H 2i+1 ( F pm − 1 ) ֒ → · · · H 2i+1 ( F 0 ) ֒ → H 2i+1 ( F − 1 ) = 0.
This means
H 2i+1 ( F pm ) = 0
. SoU 2i+1
is in the kernel of the right vertialmap. To ensure that
U 2i+1
is also in the kernel of the middle vertial mapj ∗
, we use diagram hase. The imagej ∗ ( U 2i+1 )
maps to zero, so it omesfrom
H 2i − 1 ( F pm )
. The leftj ∗
map isonto this, so we an lift it,map it intoH 2i+1 (LHP r )
, and subtratitfromthe originalU 2i+1
. This givesa hoieofU 2i+1
that satises both(i)
and(ii)
.Remark 5.14. By property
(i)
ofU ∗
, we an alulateits PoinaréseriesP U ∗ (t) = P H
odd(LHP r ) (t) − P T ∗ (t) = 1
t (P IA (t) − P IT (t)) = 1
t P IF (t),
where wehave used Remark 5.12 and (58).
Remark 5.15. We will need the dimension of parts of
U ∗
. AsT ∗ ⊕ U ∗ i
∗
∼ = H
odd(LHP r )
, andi ∗ ( T ∗ ) =
Imd ⊆ H
odd(LHP r )
, we an ompute thedi-mension of
U ∗
as the dimension of the okernel of the ation dierentiald
.Forthis wean use Lemma 5.10
(iii)
and(iv)
, and getp ∤ r + 1 : dim M
k ≤ (2r+1)mp − 1
U 2k − 1
= rm, p | r + 1 : dim M
k ≤ (2r+1)pm
U 2k − 1
= (r + 1)m.
Nowwe an prove the MainTheorem for the odd degree ohomology:
Theorem 5.16. The map of
F p [u]
-modules,h 1 ⊕ h 2 : (F p [u] ⊗ U ∗ ) ⊕ T ∗ −→ H
odd(LHP r hS 1 )
indued bythe inlusionsof
U ∗
andT ∗
, isan isomorphism ofF p [u]
-modules.Expressed in terms of generators,
H
odd(LHP r hS 1 )
is isomorphi as agraded
F p [u]
-module toM
2k ∈IF
F p [u]f 2k − 1 ⊕ M
2k ∈IT
(F p [u]/ h u i ) t 2k − 1 ,
where the lowerindex denotes the degree of the generators.
Proof. From Lemma 5.11
(i)
we see thatT ∗
is atually anF p [u]
-submoduleof
H
odd(LHP hS r 1 )
, and so the inlusionh 2 : T ∗ −→ H
odd(LHP hS r 1 )
is anF p [u]
-linear map. On the ontrary we just onsiderU ∗
as a subgroup, andmakethe
F p [u]
-moduleF p [u] ⊗ U ∗
. Thereis thenauniquewaytoextendtheinlusionof
U ∗
to anF p [u]
-linear maph 1 : F p [u] ⊗ U ∗ −→ H
odd(LHP hS r 1 )
.First we remarkthat
h 1 ⊕ h 2
issurjetive. Tosee this weuse partof theGysinexat sequene, see (59), wherethe rightmost zero is SF(6):
H 2i − 1 (LHP r hS 1 ) −→ u H 2i+1 (LHP hS r 1 ) i
∗
−→ H 2i+1 (LHP r ) −→ 0.
This is asequene of
F p
vetor spaes, so itsues to show that we an hitthe image
u(H 2i − 1 (LHP r hS 1 ))
and the okernelH 2i+1 (LHP r hS 1 )/ ker(i ∗ ) ∼ =
H 2i+1 (LHP r )
. The okernel an be hitaording to(i)
inLemma 5.13. Wenow use indution in the degree
2i + 1
. The indution start is trivial. Weget indutively that the image
u(H 2i − 1 (LHP r hS 1 ))
an be hit byu (F p [u] ⊗ U ∗ ) ⊕ T ∗
⊆ (F p [u] ⊗ U ∗ ) ⊕ T ∗
, wherethelastinlusionfollows fromLemma5.11. Soit remainstoshow that
h 1 ⊕ h 2
is injetive.The idea of the proof is now to show that map
h 1 ⊕ h 2
loalized awayfrom
u
, whih wedenote(h 1 ⊕ h 2 )[ 1 u ]
,is injetive. Againby Lemma 5.11(i)
wesee thatwhenloalizingaway from
u
,T ∗
vanishes. Sowelookath 1
,andby Lemma 5.13 there isa ommutative diagram,
F p [u] ⊗ L
i U 2i+1 h 1 //
id
⊗
projH
odd(LHP hS r 1 )
j ∗
F p [u] ⊗ L
i ≤ (2r+1)pm − δ U 2i+1 h 1 // H
odd(( F pm ) hS 1 )
(60)
where
δ =
1, p ∤ r + 1
;0, p | r + 1
.The map
j ∗
is surjetiveaording to SF(4).Loalizing away from
u
an be done by tensoring withF p [u, u − 1 ]
overF p [u]
. Sineh 1 ⊕ h 2
is surjetive, and loalization is exat,(h 1 ⊕ h 2 )[ 1 u ]
isalso surjetive. As noted,
h 2
vanishes when loalizing away fromu
, so weonlude that
h 1 [ 1
u ] : F p [u, u − 1 ] ⊗ U ∗ −→ H
odd(LHP hS r 1 )[ 1 u ]
issurjetive. When loalizing,weonlude from the diagram(60) that
h 1 [ 1
u ] : F p [u, u − 1 ] ⊗ M
0 ≤ i ≤ (2r+1)pm − δ
U 2i+1 −→ H
odd(( F pm ) hS 1 )[ 1 u ]
isalso surjetive.
To show
h 1 [ u 1 ]
as injetive, we will prove that the domain and targetspaes are isomorphi as abstrat modules. So we rst study the domain
of
h 1 [ 1 u ]
. The dimension of theU ∗
part is alulated in Remark 5.15, andtensoring with
F p [u, u − 1 ]
we obtain the rank:rank
F p [u, u − 1 ] ⊗ M
0 ≤ i ≤ (2r+1)pm − δ
U 2i+1
=
(r + 1)m, p | r + 1
;rm, p ∤ r + 1
.Turningtothetargetspaeof
h 1 [ 1 u ]
,H
odd(( F pm ) hS 1 )[ u 1 ]
,weuseTheorem5.9:H
odd(( F pm ) hS 1 )[ 1
u ] ∼ = H
odd( F m ) ⊗ F p [u, u − 1 ]
(61)Consequently,byCorollary5.8weanalulatetherankasan
F p [u]
-module:rank
H
odd(( F pm ) hS 1 )[ 1 u ] =
m(r + 1), p | r + 1
;mr, p ∤ r + 1
.So
h 1 [ 1 u ]
is a surjetive map between two freeF p [u, u − 1 ]
-modules of thesame rank. Then
h 1 [ 1 u ]
must also beinjetive.All that remains is to show that
h 1 ⊕ h 2
is injetive. Atually it will beenoughtoshowthat
h 1 ⊕ h 2
isinjetiveforeahm
,sineagiven elementwillbe in the domain of
h 1 ⊕ h 2
for a large enoughm
. So onsider an element(a, t) ∈ F p [u] ⊗ L
i ≤ (2r+1)pm − δ U 2i+1 ⊕ T ∗
in the kernel ofh 1 ⊕ h 2
. Whenloalizing,
t
vanishes, soc
loalized must be in the kernel ofh 1
loalized,whih we have shown is injetive. This means
c
loalized is zero. But theloalizationmap on
F p [u] ⊗ U ∗
,F p [u] ⊗ U ∗
loalization−→ F p [u, u − 1 ] ⊗ F p [u] (F p [u] ⊗ U ∗ ) ∼ = F p [u, u − 1 ] ⊗ U ∗
is injetive, so
c
is zero itself. This meanst
is in the kernel ofh 1
. And byLemma 5.11,
h 1
isinjetive, sot
iszero.The expression with generators follows diretly from the isomorphism
H
odd(LHP r hS 1 ) ∼ = (F p [u] ⊗ U ∗ ) ⊕ T ∗
together with the omputation of thePoinaré series inRemarks 5.12 and 5.14.
Wean nowprove thegeneralMainTheorem, givinga omplete
desrip-tion of
H ∗ (LHP r hS 1 ; F p )
:Theorem 5.17. As a graded
F p [u]
-module,H ∗ (LHP r hS 1 ; F p )
is isomorphito
F p [u] ⊕ M
2k ∈IF
F p [u]f 2k ⊕ M
2k ∈IF
F p [u]f 2k − 1 ⊕ M
2k ∈IT
(F p [u]/ h u i ) t 2k − 1 .
Here the lower index denotes the degree of the generator, and their names
are meant to suggest free and torsion generators.
Proof. First, note that when taking the odd part, we have already proved
this in Theorem 5.16. So it remains to show that
H
even(LHP r ; F p )
is a freeF p [u]
-module with generators inthe stated degrees.FirstI arguewhy
H
even(LHP r hS 1 ; F p )
is free,usingtheMorse spetralse-quene,
E s ∗ , ∗ = E s ∗ , ∗ ( M )(LHP r hS 1 )
. BySF(1)andSF(2),E 1
evenisafreeF p [u]
-module,whihisonentratedin
E 1 pm, ∗
. Sine bySF(3)allnon-trivial dier-entialsstartinevendegrees,E ∞
evenisasubmoduleofE 1
even. NotethatE 1 pm, ∗
isanitelygenerated
F p [u]
-module. SineF p [u]
isaprinipalidealdomain,thesubmodule
E ∞ (pm, ∗ )
even of the freeF p [u]
-moduleE 1 (pm, ∗ )
even is also free. Sinethe spetral sequene
E s
onverges toH ∗ (LHP r hS 1 ; F p )
,H
even(LHP r hS 1 ; F p )
is ltered by free
F p [u]
modules and is thus free itself. The generators arethe generators of
E ∞
even.Now we must nd the degrees of the generators. We will ompute
E ∞
evenintermsof Poinaréseries, anddedue the generatordegrees fromthis. The
Morse spetral sequene alone does not provide enough information, so we
ompare with Serre's spetralsequene for the bration
LHP r −→ LHP r hS 1 −→ BS 1 ,
that is,
H ∗ (BS 1 ; H ∗ (LHP r , F p )) ⇒ H ∗ (LHP r hS 1 ; F p ).
Denote this spetral sequene by
E s ∗ , ∗ ( S )
. ThenE 2 ∗ , ∗ ( S ) = H ∗ (LHP r ; F p ) ⊗ F p [u]
. Aordingto(54)and(53),H ∗ (LHP r ; F p )
has thefollowingform: thenon-trivial part is one-dimensional in eah degree, and, apart from degree
zero,sitsindegrees that omeinpairsof odd-even, withatleast 2zero-rows
between the pairs. I have tried to diagramwhat this might look like below,
astar indiatinganon-trivialgroup.
E 2 ( S ) 8 ∗ ∗ ∗ · · ·
7 ∗ ∗ ∗ · · ·
4 ∗ ∗ ∗ · · ·
3 ∗ ∗ ∗ · · ·
0 ∗ ∗ ∗ · · ·
0 1 2 3 4 5 ...
E 3 ( S ) 8 ∗ ∗ ∗ · · ·
7 ∗ ∗ ∗ · · ·
4 · · ·
3 ∗ · · ·
0 ∗ ∗ ∗ · · ·
0 1 2 3 4 5 ...
We also see the only non-trivial
d 2
dierentials must be from the even to the odd row in the odd-even pairs. What happens when we pass toE 3 ( S )
dependsonwhether
d 2
iszerooranisomorphism(theonlypossibilities). Ifd 2
iszero,theodd-even rowpairwillsurviveto
E 3
,andifd 2
isanisomorphism, onlythe odd group inltration 0willsurvive toE 3
, asindiated above.Here we an use a shortut: The dierential
d 2
an be determinedgeo-metrially;itis atually given by the ationdierential. ByLemma 5.10
(i)
we then see that
d 0,2k 2 = 0
if and only if2k ∈ IF
. Then we an write downthe Poinaré series of the
E 3
page:P (E 3 ( S ))(t) = 1
1 − t 2 + P (H
odd(LHP r ))(t) + P IF (t)
1 − t 2 + tP IF (t)
1 − t 2 .
(62)This mightnot lookvery helpful, but if we use (52) to alulate
P (E 3
even( S ))(t) − 1
t P (E 3
odd( S ))(t) = 1
1 − t 2 + 1
t P (H
odd(LHP r ))(t) = 1
1 − t 2 − t 2 (1 − t 4r )
(1 − t 4 )(1 − t 4r+2 ) = 1 − t 4r+4
(1 − t 4 )(1 − t 4r+2 )
(63)we get a quantity that does not depend on
P IF (t)
.LetusreturntotheMorsespetralsequene. UsingRemark2.10,wean
ompute the same quantity forthe
E 1 ( M )
page. Forp ∤ r + 1
this yieldsP (E 1
even( M ))(t) − 1
t P (E 1
odd( M ))(t)
= 1 − t 4r+4 + K(t)(1 − t 4r )t 4r+3 (1 − t 2 )(1 − t 4 )
− 1 − t 4r (1 − t 2 )(1 − t 4 )
(1 − t 4r+4 )t 2
1 − t 4r+2 + K(t)t 4r+3
= 1 − t 4r+4 (1 − t 2 )(1 − t 4 )
1 − (1 − t 4r )t 2 1 − t 4r+2
= 1 − t 4r+4
(1 − t 4 )(1 − t 4r+2 ) .
(64)Usingthe formulas for
p | r + 1
,thoughslightly dierent, alsogive the samequantity. As we wanted to ompute
E ∞ ( M )
, we really want to know thisquantity for
E ∞ ( M )
. Sine by SF(3), all non-trivialdierentialsinE ∗ ( M )
goes fromeven toodd total degree, wehave
dim E ∞ 2n+1 + dim M
k ≥ 1;i+j=2n+1
Im
(d k : E k i − k,j − k+1 −→ E k i,j )
!
= dim E 1 2n+1 .
From this we dedue
dim E ∞ 2n = dim E 1 2n − dim M
k ≥ 1;i+j=2n+1
Im
(d k : E k i − k,j − k+1 −→ E k i,j )
!
= dim E 1 2n − dim E 1 2n+1 + dim E ∞ 2n+1 .
Expressing this by Poinaréseries yields
P (E ∞
even)( M ) − 1
t P (E ∞
odd)( M ) = P (E 1
even)( M ) − 1
t P (E 1
odd)( M )
Now by (63) and (64) we an onlude
P (E ∞
even)( M ) − 1
t P (E ∞
odd)( M ) = P (E 3
even)( S ) − 1
t P (E 3
odd)( S )
To onlude
P (E ∞
even)( M ) = P (E 3
even)( S )
, we must showP (E ∞
odd)( M ) = P (E 3
odd)( S )
. We an omputeP (E ∞
odd)( M )
by Theorem 5.16:P (E ∞
odd)( M ) = P (H
odd(LHP r hS 1 )) = P ((F p [u] ⊗ U ∗ ) ⊕ T ∗ )
= 1
1 − t 2 P U ∗ (t) + P T ∗ (t) = 1
t(1 − t 2 ) P IF (t) + 1
t P IT (t),
whereI have used Remarks 5.12 and 5.14. Nowby Lemma5.10
(i)
,P (E 3
odd( S ))(t) = P (H
odd(LHP r ))(t) + tP IF (t)
1 − t 2
= 1
t P IA (t) + t
1 − t 2 P IF (t) = 1
t(1 − t 2 ) P IF (t) + 1
t P IT (t).
This allows us to onlude that
P (E ∞
even)( M ) = P (E 3
even)( S )
, and we anompute by (62),
P (E ∞
even)( M ) = P (E 3
even)( S ) = 1
1 − t 2 + P IF (t) 1 − t 2 ,
asstated inthe Theorem.
6
S 1
-equivariantK
-theory ofLCP r
ReallthattheMorsespetralsequeneomesfromthe
S 1
-equivariantenergy ltrationCP r = F 0 ⊆ F 1 ⊆ · · · ⊆ F n ⊆ · · · ⊆ F ∞ = LCP r ,
(65)whih onsequently gives a ltration
{ ( F n ) hS 1 } n
ofLCP r hS 1
. The Morsespetralsequene
E ∗ ( M )(LCP r hS 1 )
inK
-theoryhas the following struture,Theorem 6.1. The Morse spetral sequene
E r ∗ , ∗ ( M )(LCP r hS 1 )
onvergingto
K ∗ (LCP r hS 1 )
is a spetral sequene ofK ∗ (BS 1 ) = Z[[t]]
-modules, and ithas the following
E 1
page, using theZ/2Z
grading ofK
-theory:E 1 0,j =
Z[[t]] ⊗ Z Z[h]/ h h r i ,
j even;0,
j odd.E 1 n,j =
0,
j even;Z[[t]] (n) ⊗ R Z[x, y ]/ h Q r , Q r+1 i ,
j odd. forn ≥ 1.
Here,
R = R(S 1 ) = Z[U, U − 1 ]
, andZ[[t]] (n)
denotes theR
-module strutureU 7→ (t + 1) n
onZ[[t]]
. TheR
-module struture onZ[x, y]/ h Q r , Q r+1 i
isU 7→ (x − y)/(1 + y) + 1
.Proof. The method is exatly as in Theorem 5.1. The Morse spetral
se-quene is Theorem 4.4, and we use Theorem 3.7whih gives
K hS ∗ 1 (G(r) (n) )
,with the module strutures stated just below the Theorem. Finally, using
the
Z/2Z
-gradingfromBott-periodiity,wesuppresstheThomisomorphism, and simplyget a shift fromeven to odd degree whenn ≥ 1
.Remark 6.2. Note that when
n = 1
, theS 1
-ation is free onG(r)
, soG(r) hS 1 ≃ ∆(r)
. SoE 1,
odd∼ = K 0 (∆(r)) ∼ = Z[x, y]/ h Q r , Q r+1 i
, withZ[[t]]
-module struture
t 7→ (x − y)/(1 + y)
.WeandepittheMorsespetralsequeneshematiallyasfollows,where
anempty spae denotes zero, and a
∗
denotes a non-trivialmodule:3 ∗ ∗
2 ∗ ∗ ∗ ∗
1 ∗ ∗
0 ∗ ∗ ∗ ∗
− 1 ∗ ∗
− 2 ∗ ∗ ∗ ∗
− 3 ∗ ∗
− 4 ∗ ∗ ∗ ∗
From the ongurationof this spetral sequene, we an immediately
estab-lishanumberof strutural fats. Reallthe notation
K hS ∗ 1 (X) = K ∗ (X hS 1 )
,when
X
is anS 1
-spae.Proposition 6.3. The Morse spetral sequene onverging to
K hS ∗ 1 (LCP r )
has the following properties:
(i)
The onlypossiblenon-trivial dierentials start from olumn0
.(ii) K hS 0 1 (LCP r )
isasubmoduleofK hS 0 1 ( F 0 ) = K 0 (BS 1 ) ⊗ Z K 0 (CP r )
, andin partiular it is a free abeliangroup.
(iii)
The spetral sequene for the ltration{F i / F 0 } i
hasK ∗ (
point)
inol-umn0, andthus itollapses. So
K ˜ hS 0 1 ( F ∞ / F 0 ) = 0
, andK hS 1 1 ( F ∞ / F 0 )
isfree abelian.
Wewillalsoneedthetwistedase,i.etheMorse spetralsequeneforthe
(n)
-twisted ltrationF 0 = F 0 (n) ⊆ F 1 (n) ⊆ · · · ⊆ (LCP r ) (n)
,where we haveLemma 6.4. For the
(n)
-twisted ltrationF 0 (n) ⊆ F 1 (n) ⊆ · · · ⊆ (LCP r ) (n)
,the followingholds:
K ˜ hS 0 1 ( F 1 (n) / F 0 ) = 0
, andK ˜ hS 1 1 ( F 1 (n) / F 0 ) ∼ = Z[[t]] (n) ⊗ R Z[x, y]/ h Q r , Q r+1 i .
Proof. Morse theory says that
F 1 / F 0 ≃ T h(µ − 1 )
asS 1
-spaes, sine thel-trationis
S 1
-equivariant. As aonsequene,F 1 (n) / F 0 = ( F 1 / F 0 ) (n) ≃ (T h(µ − 1 )) (n) = T h((µ − 1 ) (n) ),
where the last equality is lear from the denition
T h(ξ) = D(ξ)/S(ξ)
. Soby Thom isomorphism,
K ˜ hS 1 1 ( F 1 (n) / F 0 ) ∼ = K hS 0 1 (G(r) (n) )
, whih by Theorem3.7is isomorphi to
Z[[t]] (n) ⊗ R Z[x, y]/ h Q r , Q r+1 i
. Likewise forK ˜ hS 0 1
.6.1 The rst dierential
We want to determine the rst dierential
d 1 : E 1 0, ∗ −→ E 1 1, ∗
in the Morsespetral sequene onverging to
K hS ∗ 1 (LCP r )
. UsingRemark 6.2, we have aonretedesriptionofthe
E 1
term,andwegetthefollowingexpliitformulafor
d 1
:Theorem 6.5. The rst dierential
d 1
inE ∗ ( M )(LCP r hS 1 )
is theZ[[t]]
-module homomorphism
d 1 : Z[[t]] ⊗ Z[h]/h r+1 −→ Z[x, y]/ h Q r , Q r+1 i
givenby
d 1 (h j ) = x j − y j
forj = 0, 1, . . . , r
.Proof. The rst dierentialis indued by the boundary map
δ
below:( F 0 ) hS 1 // ( F 1 ) hS 1 // ( F 1 ) hS 1 /( F 0 ) hS 1 δ // Σ(( F 0 ) hS 1 )
where
Σ
denotes the (redued) suspension. FromMorse theory (40) wehave( F 1 ) hS 1 /( F 0 ) hS 1 ≃ T h((µ − 1 ) hS 1 )
, whereµ − 1
is the negative bundle overX = G(r)
, and wehave the diagramS((µ − 1 ) hS 1 ) //
D((µ − 1 ) hS 1 ) //
T h((µ − 1 ) hS 1 ) //
∼ =
ΣS((µ − 1 ) hS 1 )
( F 0 ) hS 1 // ( F 1 ) hS 1 // ( F 1 ) hS 1 /( F 0 ) hS 1 δ
// Σ(( F 0 ) hS 1 )
Thevertialmaps fromthesphere-and disbundlesare given by the owof
theenergyfuntional;wereturntothemlater. First,sine
µ − 1
isanS 1
-vetorbundle,we an assumethat the Riemmanian metrionit is
S 1
-invariant,so thatS((µ − 1 ) hS 1 ) = ES 1 × S 1 S(µ − 1 )
,andD((µ − 1 ) hS 1 ) = ES 1 × S 1 D(µ − 1 )
. ThenT h((µ − 1 ) hS 1 ) ∼ = ES + 1 ∧ S 1 T h(µ − 1 )
,see [Bökstedt-Ottosen ℄ Lemma5.2, and we get the diagramES 1 × S 1 S(µ − 1 ) //
id
× (f + ⊔ f − )
ES 1 × S 1 D(µ − 1 ) //
ES + 1 ∧ S 1 T h(µ − 1 )
∼ =
ES 1 × S 1 F 0 // ES 1 × S 1 F 1 // ES + 1 ∧ S 1 F 1 / F 0
This means we an simplyignore the
ES 1
-fator, and onsider the diagramS(µ − 1 ) //
f + ⊔ f −
D(µ − 1 ) //
T h(µ − 1 ) //
∼ =
ΣS(µ − 1 )
Σf + ∨ Σf −
F 0 // F 1 // F 1 / F 0 δ
// Σ F 0
BytheproofofProp. 4.2,
µ − 1
isatrivialreallinebundle,and overageodesiγ ∈ X
,weanparametrizeµ − 1
asRiγ ′
. Therefore thespherebundleS(µ − 1 ) = X + ⊔ X −
isadisjointunionoftwoopiesofthe basespaeX
,wheretheberis
(X + ) γ = +iγ ′
and(X − ) γ = − iγ ′
. The mapf ± : X ± −→ F 0
isgiven bytheowoftheenergyfuntional: Forageodesi
γ ∈ X
,f ± (γ)
givestheendpointin
F 0 = CP r
for the owlines in diretion± iγ ′
. Sineµ − 1
is 1-dimensional, the ThomspaeT h(µ − 1 )
isjustthe suspensionΣX
of the base spaeX
, andΣS(µ − 1 ) = ΣX + ∨ ΣX −
. Themapδ : F 1 / F 0 −→ Σ F 0
isnowtheompositionδ : F 1 / F 0 ∼ = // Σ(X) // ΣX + ∨ ΣX − Σf + ∨ Σf − // Σ F 0 .
(66)Here, the last mapfolds the twosummands in the wedge.
We now investigate the maps
f ± : G(r) −→ CP r
. Reall from (4) thatthe simple losed geodesi
γ
inCP r
determined by[v, w] ∈ P V 2
is given bythe map
P V 2 −→ G(r), [v, w] 7→ q ◦ c(x, v),
where
c(x, v)(t) = cos(πt)x + sin(πt)v
fort ∈ [0, 1]
, andq : S 2r+1 −→ CP r
is the projetion. Suh a
γ
is a geodesi on aCP 1 = P { v, w } ⊆ CP r
, andwean give
P { v, w }
homogeneousoordinates,[a v , a w ] = q(a v v + a w w)
,andmap
P { v, w } −→ C ∪ {∞} , [a v , a w ] 7→ a v
a w
.
We see that
γ
under this map is the urvet 7→ cos(πt) sin(πt) = tan(πt) 1 ∈ C ∪ {∞}
for
t ∈ [0, 1]
, i.e. the real line traversed in the negative diretion, from+ ∞
to−∞
. It is now lear that the ow in diretion+iγ ′
will end in− i ∈ C ∪ {∞}
, or homogeneous oordinates√ 1
2 [1, i] ∈ P { v, w }
, sof + (γ) =
√ 1
2 [1, i] ∈ P { v, w }
. The ow in diretion− iγ ′
ends ini ∈ C ∪ {∞}
, sof − (γ) = √ 1 2 [1, − i] ∈ P { v, w }
.Having determined
f ±
, we an now alulate the indued mapf ± ∗
onK 0 (CP r ) ∼ = Z[h]/ h h r i
,so we need onlydeterminef ± ∗ (h)
, whereh = [H] − 1
and
H ց CP r
is the standard line bundle. We do this by determining the pullbakf ± ∗ (H)
. From the preeding paragraph we see that the beroff + ∗ (H)
over asimple losedgeodesiγ
determined by[v, w] ∈ P V 2
is exatlyall the points on the line given by
√ 1
2 (v + iw)
. Reall that the line bundleX
was dened as the pullbak of the standard bundleγ 1 ց P(γ 2 )
under theomposite
G(r) −→ P V 2 −→ P V g 2 −→ P(γ 2 ),
γ 7→ [v, w] 7→ √ 1 2 [v + iw, v − iw] 7→ C(v + iv) ⊆ Cv ⊕ Cw
It follows that
f + ∗ (H) = X
, sof + ∗ (h) = x
. Likewise we getf − ∗ (h) = y
,beause
Y
is the pullbak of the omplement ofγ 1
inγ 2
. Sinef ± ∗
is a ringhomomorphism,weget
f + ∗ (h j ) = x j
,andf − ∗ (h j ) = y j
. From(66),weannowompute
d 1 (h j )
. Whenfoldingthe maps,theseondsuspensioninthewedgeΣX + ∨ ΣX −
has the orientationreversed,so weobtaind 1 (h j ) = x j − y j
.In the Morse spetral sequene
E ∗ ( M )((LCP r ) (n) hS 1 )
for the(n)
-twistedltration,the rst dierentialisamap
d (n) 1 : K hS ∗ 1 (Σ F 0 ) −→ K ˜ hS ∗ 1 ( F 1 (n) / F 0 )
,f. Lemma6.4.
Lemma 6.6. The rst dierential in
E ∗ ( M )((LCP r ) (n) hS 1 )
is the map ofZ[[t]]
-modules given byd (n) 1 : Z[[t]] ⊗ Z[h]/
h r+1
−→ Z[[t]] (n) ⊗ R Z[x, y]/ h Q r , Q r+1 i , d (n) 1 (h j ) = x j − y j ,
forj = 0, 1, . . . , r.
Proof. Usingthesame diagramasinthe proofof Theorem6.5above,wesee
that the geometry of this situation is exatly the same, so the ow map is
idential tothe one omputed before.
From(44),thepowermap
P j
givesamapofthefollowingexatsequenes,givinga ommutative diagram:
F 0 //
id
F 1 (j) //
P j
F 1 (j) / F 0 δ (j) 1
//
P j
Σ F 0
id
F 0 // F ∞ // F ∞ / F 0 δ
// Σ F 0
where
δ 1 (j)
denotes the boundary map whih induesthe rst dierentiald (j) 1
inthe Morse spetralsequene
E ∗ ( M )((LCP r ) (j) hS 1 )
. So the dierentiald (j) 1
determined inLemma 6.6an alsobewritten asthe omposite map