• Tidak ada hasil yang ditemukan

with exact rows coming from (2.27) and where hi,nD :HDi,npXq –HD3´i,2´npXqis the isomor- phism induced by the perfect pairing (2.23) of Deligne cohomology groups. Specializing to i“2 yields the commutative square

H2,npXqR ρ2 //

hp2qBpX,nq

HD2,npXq

h2,nD

Hc2,2´npXq˚ //HD1,2´npXq˚

(3.17)

(2.23) simplifies for i “ 2 to the restriction of the Poincaré duality pairing of algebraic topology

H1pXpCq,Rpn´1qq ˆH1pXpCq,Rp2´nqq ÝÑ H2pXpCq,Rp1qq.

to itsGR-equivariant part. Since Poincaré duality also holds integrallyh2,nD does not contribute to the determinant of the upper right decompositionh2,nD ˝ρ2:H2,npXqRÑHD1,2´npXq˚ of the square. Finally, sinceHc2,2´npXq derives its integral structure from the bottom map of (3.17) the claim follows from taking determinants in (3.17).

It shows that the motivic decomposition of Hr2,1pXq is given by Hr2,1pXq “cokerpρ1q ‘

ˆPic0X ClF

˙

R

‘ R. (3.21)

Hr2n,npXq is isomorphic to then-th Arakelov Chow group CHnpXqR (cf. [8] Prop. 2.11) and the pairingB(X,n) translates into a perfect pairing

Hr2n,npXq ˆHr4´2n,2´npXq ÝÑ R. (3.22)

Remark 3.6. Flach’s and Morin’s identification Hr2n,npXq – CHnpXqR results from an application of the5-Lemma and thus depends on the choice of a splittingH2,1pXqRÑHr2,1pXq of (3.20). However, the identifications (3.19) and the decomposition (3.21) into motivic degree components provide one such splitting, i.e. we haveHr2n,npXq –CHnpXqR canonically.

From now on we will assume the following enhanced version ofB(X,n).

Conjecture 3.7 (B(X,n)). Conjecture B(X,n) holds and the perfect pairing CHnpXqRˆCH2´npXqR ÝÑ R

obtained from (3.22) via the canonical identifications Hr2n,npXq –CHnpXqR coincides with the Arakelov Intersection Pairing x´,´yAr.

For arithmetic surfaces x´,´yAr is only non-trivial ifn‰0,1,2. Moreover, it only involves information from motivic degree 1 ifn “ 1. We now make the Arakelov pairing explicit for n “ 1 and compare it to the intersection pairing x´,´yX from algebraic geometry, following [15].

Proposition 3.8. (Arakelov, Hriljac)

(i) (3.21) is an orthogonal decomposition of Hr2,1pXq, i.e. one has Hr2,1pXq “cokerpρ1q K

ˆPic0X ClF

˙

R

K R

andx´,´yAr is defined on each summand separately. x´,´yAr is negative definite on pPic0X{ClFqR (cf. [15] Thm. 3.4, Prop. 3.3).

(ii) Let D, D1 PPic0X be fibral divisor classes with support in the special fiber Xp. Then xD, D1yAr “logNp¨ xD, D1yX

(cf. [15] def. ofpD¨Eqv in Sec. 2).

(iii) There is a unique linear splitting

Pic0X ÝÑ pPic0XqQ, P ÞÑP

of the natural projection Pic0X ÑPic0X such that the image is orthogonal to all fibral divisor classes inPic0X (cf. [15] Thm. 1.3).

(iv) Fix an isomorphism φ :Pic0X ÝÑ JacX. X has a divisor such that its associated canonical height h satisfies for all P PPic0X (cf. [15] Thm. 3.1)

xP,PyAr “ ´hpφpPqq.

Definitions of RpXq, RpXq and cppXq. Let P be a basis of the image of 1H2,1pXq in

1H2,1pXqR– pPic0X{ClFqR. P also defines an integral basis onHc2,1pXq due to Hc2,1pXq – H2,1pXqR. LetP˚ be the basis of 1H2,1pXq˚R dual toP. We define theregulator RpXq of X to be

RpXq:“det`

xP,P1yAr

˘

P,P1PP “detMP,P˚

´

php2qBpX,1qq˚

¯ .

Next, fix a basis P “ tPiu1ďiďrk Pic0X of the image of Pic0X in pPic0XqR. The regulator RpXq of the generic fiberX equals

RpXq “det`

xP,P1yAr

˘

P,P1PP

since the Arakelov Intersection Pairing is by Proposition 3.8(iv) the same as the Neron-Tate height pairing on Pic0X.

Now, fix a prime p. Let J ÑSp denote the Neron model of the Jacobian J “JacXFp of the generic fiber of the local surface XOp overSp“SpecOp. Let J˜“Jp denote the special fiber ofJ and let J˜0 be its identity component. We also write J0 ĂJ for the subgroup scheme with generic fiberJ and special fiber equal toJ˜0. We define

cppXq:“

#

JpFpq

J0pFpq “ # ˜Jpkppqq

# ˜J0pkppqq.

Decomposition of RpXq. Fix a prime pof O. Recall the notationsdppqand njppq from Lemma 2.2. Also, let tCjpu1ďjďdppq be the reduced irreducible components of Xp and let mjppq be the multiplicity of Cjp inXp. The sections:SÑX provides a rational point on one component – sayCdppqp – ofXp. ThusCdppqp must be simple and cannot decompose further over any algebraic extension of kppq, i.e. mdppqppq “ndppqppq “1. We conclude that the set of classesDp :“ trCjps P pΛpqRu1ďjădppq is a basis of the image ofΛp insidepΛpqR.

Lemma 3.9. (Raynaud;Bosch,Liu) Fix a prime p of O. The sequence CH0pXpq

α

ÝÝÝÝÑ CH0pXpq

β

ÝÝÝÝÑ Z

Cjp ÞÑ

dppq

ÿ

i“1

xCip, CjpyX¨Cip

Cjp ÞÑ mjppq

(3.23)

is a chain complex and one has

#

Kerβ

Imα “cppXq ¨

dppq

ź

j“1

njppq.

Proof. This is [5] Theorem 1.11 applied to the abelian varietyA“J. Indeed, the right-most term in Thm 1.11qdZ{d1Zvanishes sinceXphas a component satisfyingmdppqppq “ndppqppq “ 1. Moreover, the geometric multiplicitiesej of Cjp inCjp (cf. [6] Def. 9.1.3) occurring in [5]

Rmk. 1.12 equal 1 since the base change of any reduced curve over a perfect field to its algebraic closure remains reduced (see e.g. [26] example (6.1.7)).

Remark 3.10. Raynaud has shown the analogue of Lemma 3.9 for algebraically closed residue fields (cf. [26] Prop. 8.12); Bosch and Liu extended it to more general residue fields. As part of his proof Raynaud has shown that, in the casekppq “kppq, the Picard-scheme Pic0X

p

is isomorphic to the group of components J˜{J˜0 of J – a finite étale group scheme that only depends on the generic fiberX. This should serve as intuition for why#KerImαβ does not depend on the special fiber Xp beyond the values of the njppq.

Proposition 3.11. One has

RpXq “ ˘ 1 p#Tor Pic0Xq2

ˆ

#

Tor Pic0X

ClF

˙2

¨RpXq ¨ź

p

¨

˝plogNpqdppq´1

dppq

ź

j“1

njppq

˛

‚cppXq

“ ˘ 1

p#Tor Pic0Xq2 ˆ

#

Tor Pic0X

ClF

˙2

¨RpXq ¨Π˚pX,1q ¨ź

p

cppXq.

Proof. (3.3) gives a short exact sequence of real vectorspaces 0 ÝÑ

ˆ à

p

Λp

˙

R

ÝÑ

ˆPic0X ClF

˙

R

ÝÑ `

Pic0

R ÝÑ 0. (3.24)

Due to the splitting provided by Proposition 3.8(iii) the above sequence yields an orthogonal decomposition

ˆPic0X ClF

˙

R

– ˆ

à

p

Λp

˙

R

K `

Pic0

R (3.25)

and we may regardP1 :“PYŤ

pDp as a further (R-)basis of 1H2,1pXqR. (3.25) gives RpXq “`

detMP,P1pidq˘2

¨det`

xP,P1yAr

˘

P,P1PP1

“`

detMP,P1pidq˘2

¨det`

xP,P1yAr

˘

P,P1PP¨ ź

p

det

´

xrCips,rCjpsy

¯

1ďi,jădppq

“RpXq ¨`

detMP,P1pidq˘2

¨ ź

p

plogNpqdppq´1det

´

xCip, CjpyX

¯

1ďi,jădppq

(3.26)

where the last equation uses Proposition 3.8(ii). We evaluate the remaining factors separately.

First, since (3.3) is an integral exact sequence,detMP,P1pidqmeasures the discrepancy in torsion between Pic0X{ClF and its surrounding terms in (3.3), i.e. one has

detMP,P1

´ id1

H2,1pXqR

¯

“ 1

#Tor Pic0X ¨

#

Tor Pic0X

ClF .

Finally, recall the sequence (3.23) of the previous Lemma. Sincemdppq“1the set Dp may also be viewed as a basis of Kerβ. Besides,α is represented by the full intersection matrix

´

xCip, CjpyX

¯

1ďi,jďdppq. It follows that det

´

xCip, CjpyX

¯

1ďi,jădppq

#

Kerβ

Imα. Lemma 3.9 completes the proof.