Usingourpreviousomputations(Theorems2.1and2.4)andSerre'sspetral
sequene,wewillbeabletoomputetheequivariantohomologyofthespae
of geodesis,
G(HP r ) (n)
.Werst onsider the ase
p ∤ n
, sine this is the easiest. We show:Proposition 2.11. For
p ∤ n
:H m (BC n ; F p ) = 0,
form > 0.
Proof. We are going to use that
EC n −→ BC n
is a overing, sineC n
isdisrete. In general, given a
k
-sheet overingπ : E −→ B
(assumeB
on-neted), one an onstrut a so-alled transfer map. By baryentri
sub-division one knows that it is enough to onsider very small simplies in
B
. Therefore, given a simplex inB
we an assume it is ontained in aneighborhood
U
suh thatπ − 1 (U )
is a disjoint union of open sets mappedhomeomorphially to
U
byπ
. Then we an pull the simplex inU
bakby
π
, yieldingk
opies of the simplex inE
, whih we formally add,giv-ing a hain map
τ : C m (B) −→ C m (E)
. This indues the transfer mapτ ∗ : H m (E) −→ H m (B)
on ohomology. From the denition,π ♯ ◦ τ
ismultipliation by
k
, and soτ ∗ π ∗
is also multipliation byk
. In our ase,EC n −→ BC n
isann
-sheet overing, and sothe ompositionH m (BC n ; F p ) τ
∗
−→ H m (EC n ; F p ) π
∗
−→ H m (BC n ; F p )
ismultipliationby
n
. Sine weare usingF p
-oeientsandp ∤ n
, this isanisomorphism. On the other hand, for
m > 0
, the middle term is zero, sineEC n
isontratible. ThusH m (BC n ; F p ) = 0
form > 0
.Withthis we an prove:
Theorem 2.12. For
p ∤ n
, the equivariant ohomology withF p
oeientsof the
n
-twisted spae of geodesis onHP r
isH ∗ ((G(HP r ) (n) ) hS 1 ; F p ) ∼ = F p [x, t]/ h Q r , Q r+1 i ,
where
x
has degree2
, andt
hasdegree4
, andx
istheimage of the generatoru ∈ H 2 (BS 1 )
under the map∆(HP r ) −→ BS 1
in (9).Proof. We use the Serre's spetral sequene of the brationfrom Prop. 1.6:
BC n −→ ES 1 × S 1 G(r) (n) −→ ∆(r).
Proposition2.11 above now immediatelyimplies that
H ∗ ((G(r) (n) ) hS 1 ; F p ) = H ∗ (ES 1 × S 1 G(r) (n) ; F p ) ∼ = H ∗ (∆(r); F p )
The ase
p | n
requires more work, and one needs to take into aountwhether or not
p | r + 1
. But rst we need a omputationofH ∗ (BC n ; F p )
:Proposition 2.13. For
p | n
,H ∗ (BC n ; F p ) ∼ = F p [u, e]/
e 2 .
Proof. Use Theorem 1.4
(i)
onthe brationS 1 −→ ES 1 −→ BS 1
to divideout the ation of
C n ⊆ S 1
, and obtaina brationS 1 −→ BC n −→ BS 1 .
(21)Here wehave identied the quotientgroup
S 1 /C n
withS 1
itself viathen
thpowermap
z 7→ z n
. Wewillapply Serre'sspetral sequene.First, though, we will nd
H 1 (BC n ; F p )
. SineC n
is disrete,EC n −→
BC n
is the universal overing. From overing spae theory,π 1 (BC n ) ∼ = C n
,and sine this is abelian, it follows that
H 1 (BC n ; Z) ∼ = Z/nZ
. Using theUniversal Coeient theorem, we an ompute
H 1 (BC n ; F p )
. Note thatH 0 (BC n ) = Z
, soExt(H 0 (BC n ), F p ) = 0
, and therefore,sinep | n
:H 1 (BC n ; F p ) ∼ =
Hom(H 1 (BC n ), F p ) ∼ =
Hom(Z/nZ, Z/pZ) ∼ = F p ,
Now we turn to Serre's spetral sequene for the bration (21), with
E 2 p,q = H p (BS 1 ; H q (S 1 ; F p )) = H p (BS 1 , F p ) ⊗ H q (S 1 ; F p )
. Note that theonly possible non-trivial dierential is
d 2
, sine theE 2
page has only twonon-zero rows. Knowing that
H 1 (BC n ; F p ) ∼ = F p
, we onlude that the rstdierential
d 0,1 2
must beamapF p −→ F p
withkernel isomorphitoF p
. Thisfores
d 2 (e) = 0
,wheree
generatesH(S 1 ; F p )
. Usingthederivationproperty:d(eu j ) = d(e)u j ± ed(u j ) = 0.
So all dierentials are zero, the spetral sequene ollapses, and
E ∞ = E 2
.There are no extension problems, sine eah diagonal
p + q = ∗
ontains atmost one non-zero group, so
H ∗ (BC n ; F p ) = E ∞
,as desired.Theorem 2.14. Let
p
be a prime number andn ∈ N
suh thatp | n
. AsF p [u]
-modules,the following holds:(i)
Supposep ∤ r + 1
. ThenH ∗ G(HP r ) (n)
hS 1 ; F p
∼ = F p [u]
1, y, y 2 , . . . , y r − 1 , τ, τ y, . . . , τ y r − 1 . (ii)
Supposep | r + 1
. ThenH ∗ G(HP r ) (n)
hS 1 ; F p ∼ = F p [u]
1, y, y 2 , . . . , y r , σ, σy, . . . , σy r .
where
y
has degree4
,τ
has degree4r + 3
andσ
has degree4r − 1
.Proof. In the beginning, the proofs of the two ases are the same. Consider
the spetral sequene forthe brationfromProp. 1.6
G(r) −→ ES 1 × S 1 G(r) (n) −→ BS 1 .
(22)Aording to our omputation of the ohomology of the bre in Theorem
2.1, neither the bre nor the base has anything in ohomology of degree 1.
This means that
H 1 ((G(r) (n) ) hS 1 ) = 0
. We an use this when onsideringthe spetral sequene forthe other brationfromProp. 1.6:
BC n −→ ES 1 × S 1 G(r) (n) −→ ∆(r).
Aording toProp. 2.13,
E 2 q,s = H q (∆(r); H s (BC n ; F p )
looksas follows:3 ue uex uex 2 , uet . . .
2 u ux ux 2 , ut
1 e ex ex 2 , et . . .
0 1 x x 2 , t
0 1 2 3 4 · · ·
(23)
Let us denote the two lower rows of the
E 2
page byF
. Then the next tworows (rows 2and 3)onsists of
uF
,the next two areu 2 F
, et. Consider thedierential
d 2
as a mapd 2 : eH ∗ (∆(r)) −→ H ∗ (∆(r))
from row 1 to row0. Then, using the derivation property of the dierentials we see that
d 2
ismultipliationwith
d 2 (e)
. When passingfrom theE 2
totheE 3
page,F
willbe replaed by two rows, Cok
d 2
and Kerd 2
,uF
willbereplaed byu
Cokd 2
and
u
Kerd 2
, et.So to determine the
E 3
page, we need to ndd 2 (e)
. As noted, the totalspaehas
H 1 = 0
,sod 0,1 2 : E 2 0,1 −→ E 2 2,0
must bean injetivemap, heneanisomorphism. Thisfores
d 2 (e) =
unit· x
;wemightaswellsayd 2 (e) = x
. Sod 2
ismultipliationbyx
,and we must determine Cok(x)
and Ker(x)
. UsingTheorem 2.4, we see that
Cok
(x) ∼ = F p [x, t]/ h x, Q r , Q r+1 i ∼ = F p [t]/ h Q r (0, t), Q r+1 (0, t) i .
(24)Now by Lemma 2.7,
Q r (0, t) = (r + 1)t r
andQ r+1 (0, t) = (r + 2)t r+1
. Thisiswhere we must distinguish between the two ases.
Butletusrst investigatethe kernel. I havetriedtodiagramthe
dimen-sions of
F p [x, t]/ h Q r , Q r+1 i
using Remark 2.10, with boldfae indiatingthedegrees where, for dimension reasons, the kernel must be non-trivial. The
degrees are in the top row:
0 2 4 6 8 · · · 4r 4r+2 4r+4 4r+6 4r+8 4r+10 4r+12 · · ·
1 1 2 2 3 · · · r r r − 1 r − 1 r − 2 r − 2 r − 3 · · ·
Thepatternis(hopefully)lear: Theremustbeapartofthekernelindegrees
4(r + i) − 2
fori = 1, ..., r
. In partiular, the dimension is at leastr
. Now,for the rest of the proof, we need tohandle the twoases separately.
Case
(i)
:p ∤ r + 1
. In this ase,r + 1
is a unit inF p
, so (24) beomesCok
(x) ∼ = F p [t]/ h t r i
. In partiular, the dimension of Cok(x)
isr
, generatedby
1, t, . . . , t r − 1
.Sine
dim
Ker(x) = dim
Cok(x) = r
,we have determined above that thekernel is indegrees
4(r + i) − 2
fori = 1, ..., r
. In eah degree, the kernel isone-dimensional,say generated by
ϕ i
indegree4(r + i) − 2
. Sowean writedown the
E 3
page:3 uϕ 1 uϕ 2 · · · uϕ r
2 u ut ut 2 · · · ut r − 1
1 ϕ 1 ϕ 2 · · · ϕ r
0 1 t t 2 · · · t r − 1
0 2 4 6 8 . . . 4r − 4 4r − 2 4r 4r+2 4r+4 4r+6 . . . 8r − 2
Beause there are no further dierentials on
t
andu
, and the dierentials satisfy the derivation property, we see that the spetral sequene ollapsesfrom the
E 3
page. Now let usompare this to the spetral sequene for thebration
G(r) −→ ES 1 × S 1 G(r) (n) −→ BS 1
from (22) onsidered in thebeginning, whihalso onverges to
H ∗ ((G(r) (n) ) hS 1 ; F p )
. SineH ∗ (G(r); F p ) ∼ = F p [y, τ ]/
y r = 0, τ 2 = 0 ,
where
y
has degree4
andτ
has degree4r + 3
,we get theE 2
page,E 2 ∗ , ∗ ∼ = F p [y, τ]/
y r = 0, τ 2 = 0
⊗ F p [u]
Comparing this to the
E 3
page above, we see that we have in eah ase2r
generators whih are multiplied by1
,u
,u 2
, et. This means, sine therst spetral sequene ollapses, that this seond one must also ollapse.
Consequently we an read o that
H ∗ (G(r) (n) hS 1 ; F p )
as anF p [u]
-module isgenerated by
1, y, y 2 , . . . , y r − 1 , τ, τ y, . . . , τ y r − 1
Case
(ii)
:p | r + 1
. In this ase,r + 1
is zero inF p
, butr + 2
is aunit, so(24) beomes:
Cok
(x) ∼ = F p [t]/t r+1 .
In partiular, the dimension of Cok
(x)
isr + 1
,generated by1, t, . . . , t r
.Consequently,
dim
Ker(x) = r+1
,soweneedtondanadditionalelementinthe kernel. By Lemma2.7,
Q r
is the polynomialQ r = (r + 1)t r −
r + 2 r − 1
t r − 1 x 2 + · · · ± x 2r ,
so
x
dividesQ r
inF p [x, t]
. This means we have an elementϕ 0 = Q r /x
indegree
4r − 2
whih is in the kernel ofx
. So together with the elementsϕ 1 , . . . , ϕ r
frombefore, we have found generators of the kernel.As in Case
(i)
, we see that the spetral sequene ollapses from theE 3
page. Comparingwiththe
E 2
pageofthe bration(22),andusingthatsinep | r + 1
,H ∗ (G(r); F p ) ∼ = F p [y, σ]/
y r+1 = 0, σ 2 = 0 ,
weonlude asabovethat
H ∗ (G(r) (n) hS 1 ; F p )
asanF p [u]
moduleisgeneratedby
1, y, y 2 , . . . , y r , σ, σy, . . . , σy r .
Corollary 2.15. For the Serre spetral sequene of the bration
G(HP r ) −→ G(HP r ) (n) hS 1 −→ BS 1
the following holds: If
p | n
, it ollapses from theE 2
page. Ifp ∤ n
theinlusion of the bre indues a surjetive map on even degree ohomology
H 2 ∗ (G(HP r ) (n) hS 1 ; F p ) −→ H 2 ∗ (G(HP r ); F p )
Proof. Thease
p | n
follows diretly fromtheproof ofTheorem 2.14 above.Forthe ase
p ∤ n
, wemust hek thatthe lassesy j
fromTheorem 2.1are inthe image of the inlusion of the bre. To do this, we onsider the
E 2
pageof the spetral sequene, and must show that the lasses
y j
survive toE ∞
.Sine the dierentialsare derivations,
d s (y j ) = jy j − 1 d s (y)
, and so it suesto show
y
survives. Clearly it does, sine any dierentialstarting aty
endsintotal degree
5
,and there are nonon-triviallasses in total degree5
.3
K
-theory of spaes of geodesis inCP r
Let
G(r) = G(CP r )
be the spae of simple, losed, parametrized geodesis inCP r
, and let∆(r) = S 1 \ G(r)
be the quotient spae under the rotationation. In this hapter we obtain
K
-theoreti analogues of the results forohomologyfromthe previous hapter.
By
K
-theory we mean omplexK
-theory, i.e.K 0 (X)
for aCW-omplexX
isthe groupompletionof the semi-groupofomplexvetor bundleswithbase spae
X
. DeneK ∗ (X)
for a general spaeX
as follows: Chose anyCW omplex
Y
weakly equivalent toX
, putK(X) = K(Y )
. This is welldened, sine two hoies of
Y
will be homotopy equivalent, andK
-theoryis homotopy invariant. We most often employ the
Z/2Z
-grading fromBott-periodiity,writing
K ∗ (X) = K 0 (X) ⊕ K 1 (X)
.3.1 The unparametrized geodesis
Reall the model for
∆(r)
from the end of setion 1.2. We hadγ 2
, thestandard 2-dimensional bundle over the Grassmannian Gr
2 (C r+1 )
andp : P(γ 2 ) −→
Gr2 (C r+1 )
the assoiated projetive bundle. Thenwe had aom-posite map (6),whih isan
S 1
-equivariant dieomorphismϕ : ∆(r) −→ P(γ 2 )
Take the standard line bundle
γ 1
overP(γ 2 )
. The pullbakϕ ∗ (γ 1 )
ofγ 1
under
ϕ
is a line bundle we willdenoteX
. We onsider also the onjugateline bundle
γ 1 ⊥
toγ 1
overP(γ 2 )
, i.e.γ 1 ⊕ γ 1 ⊥ = p ∗ γ 2
. The pullbakϕ ∗ (γ 1 ⊥ )
of this bundle to
∆(r)
we will denoteY
. InK 0 (∆(r))
we dene the lassesx = [X] − 1
andy = [Y ] − 1
.Theorem 3.1. Let
x, y ∈ K 0 (∆(r))
be the lasses dened above. ThenK 0 (∆(r)) ∼ = Z[x, y ]/ h Q r , Q r+1 i ,
K 1 (∆(r)) = 0,
where
Q s
fors ∈ N
is the homogeneouspolynomial inx, y
given byQ s (x, y ) =
X s j=0
x j y s − j .
Note that thesepolynomialsare not the same isin the ohomologyase,
Proof. We apply the Atiyah-Hirzebruh spetralsequene, Theorem 1.3
H ∗ (∆(r); K ∗ ( ∗ )) ⇒ K ∗ (∆(r)).
(25)Sineweknowthe ohomologyof
∆(r)
from [Bökstedt-Ottosen ℄,H ∗ (∆(r)) ∼ = Z[x 1 , x 2 ]/ h Q r , Q r+1 i ,
and
x 1 , x 2
have degree 2, we see that all dierentials in (25) are trivial, so thatE ∞ = E 2 ∼ = Z[x 1 , x 2 ]/ h Q r , Q r+1 i ⊗ Z[β, β − 1 ],
where
β
denotes the Bott element. This shows thatK 1 (∆(r)) = 0
, andK 0 (∆(r))
is free abelianof the same rankasH ∗ (∆(r))
.Weuse the Chern harater,
h
: K 0 (X) −→ H ∗ (X; Q),
whihisaring homomorphism. Byonstrution,
x 1 = c 1 (X)
andx 2 = c 1 (Y )
are the rst Chern lasses of
X
andY
, f. [Bökstedt-Ottosen℄ Thm. 3.2, so sineX, Y
are linebundles, we geth
(x) =
h(X) − 1 = exp(c 1 (X)) − 1 = exp(x 1 ) − 1,
h
(y) =
h(Y ) − 1 = exp(c 1 (Y )) − 1 = exp(x 2 ) − 1.
ThereisarelationbetweentheChernharaterhandtheAtiyah-Hirzebruh
spetral sequene, by [Atiyah-Hirzebruh℄ Cor. 2.5. We see that h
(x i y j ) =
h
(x) i
h(y) j = x i 1 x j 2 +
higher terms, where higher terms means terms inhigher ltration, whih in this ase is equivalent to higher total degree in
x 1
,x 2
. By(iii)
in the orollary, this shows that the ring homomorphismZ[x, y ] −→ K ∗ (∆(r))
issurjetive.This means we an use
x, y
as polynomial generators forK ∗ (X)
, and itremains todetermine the relations. Againwe use the Chern harater, this
time aftertensoring with
Q
:h
: K 0 (X) ⊗ Q −→ H ∗ (X, Q)
whih is then a ring isomorphism. We now want to prove that h
(x)
andh
(y)
satisfy the relationsQ r , Q r+1
. Ifwe anprovethis, we are done: SinetheChernharaterisanisomorphismaftertensoringwith
Q
,andthegroupsare torsion-free, there an beno furtherrelationsin
K 0 (S(τ)/S 1 )
,sine thishas the same rank as
H ∗ (S(τ)/S 1 ) ∼ = Z[x 1 , x 2 ]/ h Q r , Q r+1 i
.Soweneedtoprovethat
Q s (exp(x 1 ) − 1, exp(x 2 ) − 1) = 0
ifQ s (x 1 , x 2 ) = 0
for
s = r, r + 1
. Realling that the idealsh Q r , Q r+1 i
andQ r , x r+1 1 , x r+1 2
oinide, we rst get that
(exp(x i ) − 1) r+1 = x r+1 i (1 +
higherterms) = 0
.Consider the quotient map
R = Q[x 1 , x 2 ]/
x r+1 1 , x r+1 2
−→ Q[x 1 , x 2 ]/ h Q r , Q r+1 i = S,
whih has kernel
I = h Q r i
. Given a power series without onstant term,g(z) = a 1 z + a 2 z 2 + · · ·
, we an deneg ∗ : R −→ R
byx i 7→ g(x i )
fori = 1, 2
. In our ase,g(z) = exp(z) − 1
. If we an prove thatg ∗ I ⊆ I
, themap
g ∗
will bewell-dened as a mapS −→ S
, as shown below:0 // I //
g ∗
R //
g ∗
S //
0 0 // I // R // S // 0
Wewillshow
I =
Ker(x 1 − x 2 )
. Consider a homogeneouspolynomialf ∈ R
of degree
m
. It sues totakem ≥ r
,foriff
had lowerdegree, itouldnotbe in
I = h Q r i
, sineQ r
has degreer
. Then, usingx r+1 1 = x r+1 2 = 0
, weanwrite
f = X r i=m − r
c i x i 1 x m 2 − i ⇒ (x 1 − x 2 )f = X r i=m − r+1
(c i − 1 − c i )x i 1 x m 2 − i .
By[Bökstedt-Ottosen ℄ Lemma 3.4,
f ∈ I
if and only ifc m − r = . . . = c r
, andwe onlude
I =
Ker(x 1 − x 2 )
. This impliesg ∗ I ⊆
Ker(g ∗ x 1 − g ∗ x 2 )
. So wealulate
g ∗ x 1 − g ∗ x 2 = X
i ≥ 1
a i (x i 1 − x i 2 ) = (x 1 − x 2 ) X
i
a i i − 1
X
k=0
x k 1 x i 2 − k − 1
! .
This shows
g ∗ I ⊆
Ker(g ∗ x 1 − g ∗ x 2 ) ⊆
Ker(x 1 − x 2 ) = I
,as desired.Remark 3.2. Let
M = K ∗ (∆(r)) = Z[x, y ]/ h Q r , Q r+1 i
. We often useltration arguments, so let us x the notation now. Let
M j ⊆ M
bethe group generated by monomials in
x, y
of total degree at leastj
, i.e.M j = Z[x, y ] ≥ j / h Q r , Q r+1 i
. This makes sense sineQ r
,Q r+1
arehomoge-neous. Then