Doc Math J DMV

Classical Motivic Polylogarithm

According to Beilinson and Deligne

Annette Huber Jorg Wildeshaus

Received February

Revised September

Communicated by Peter Schneider

Abstract The main purp ose of this pap er is the construction in motivic

cohomology of the cyclotomic or classical polylogarithm on the pro jective

line minus three p oints and the identication of its image under the regula

tor to absolute Deligne or l adic cohomology By sp ecialization to ro ots of

unity one obtains a compatibility statement on cyclotomic elements in mo

tivic and absolute cohomology of ab elian number elds As shown in BlK

this compatibility completes the pro of of the Tamagawa number conjecture

on sp ecial values of the Riemann zeta function

The main constructions and ideas are contained in Beilinsons and Delignes

unpublished preprint Motivic Polylogarithm and Zagier Conjecture

BD We work out the details of the pro of setting up the foundational

material which was missing from the original source the pap er contains

an app endix on absolute Ho dge cohomology with co ecients and its inter

pretation in terms of Saitos Ho dge mo dules The second app endix treats

K theory and regulators for simplicial schemes

Mathematics Sub ject Classication Primary F Secondary R

R R D F

Keywords Polylogarithm motivic and absolute cohomology regulators cy

clotomic elements

Introduction

The aim of this work is to present the construction of the class of the cyclotomic or

classical p olylogarithm in motivic cohomology It maps to the elements in Deligne and

l adic cohomology dened and studied in Beilinsons Polylogarithm and cyclotomic

elements B The latter elements can b e seen as b eing represented by a pro

variation of Ho dge structure or a prol adic on the pro jective line minus three

p oints

1

See correction on page of this volume

Documenta Mathematica

Annette Huber Jorg Wildeshaus

Our main interest lies in the sp ecialization of these sheaves to ro ots of unity

they represent the cyclotomic oneextensions of Tate twists already studied by

Soule Sou Deligne D and Beilinson B

Let us b e more precise denote by the set of primitive dth ro ots of unity in

d

Q Q T T d We get an alternative pro of of the following theorem of

d d

Beilinsons

Corollary Assume n and denote by r the regulator map

D

M

n

C i R H Sp ec Q Q n

d

M

Q C

d

There is a map of sets

H Sp ec Q Q n

n d

d M

such that

M

n

C i R r

D n

d

Q C

d

P

k

maps a ro ot of unity to Li

n

n

k

k

Now x a dth primitive ro ot of unity in Q This choice allows to identify

Sp ec Q Q n with a Q subspace of continuous etale cohomology H

d l l

cont

1

Gal Q Q

l

r

n

l

A

1 1

Q Q Q lim

r

Z l l l

l

l

r

Note that there is a distinguished ro ot of unity T in Q As was observed already

d

in B the study of the cyclotomic p olylogarithm gives a pro of of BlK Conjecture

cf Sou Theoreme for the case n Gr Theoreme IV for the lo cal

version if l d

Corollary Let b e the map constructed in Under the ab ove

n

inclusion the l adic regulator

r H Sp ec Q Q n H Sp ec Q Q n

l d d l

M cont

b

maps T to

n

X

d n

A

n

d n

r

l b

r

This result implies in particular that Soules cyclotomic elements in the group

K F Z for an ab elian number eld F and a prime l are induced by el

n Z l

ements in K theory itself Corollary Furthermore the case d of forms a

central ingredient of the pro of of the Tamagawa number conjecture mo dulo p owers of

for o dd Tate twists Q n n BlK x Finally as shown in KNF Theorem

the general case of implies the mo died version of the Lichtenbaum conjecture

for ab elian number elds

Documenta Mathematica

Classical Motivic Polylogarithm

The main ideas necessary for b oth the construction of the motivic p olylogarithm

and the identication of the realization classes together with a sketch of pro of are

contained in the unpublished preprint Motivic Polylogarithm and Zagier Conjecture

BD and its predecessors B BDp Our aim in this pap er is to work out the

details of the pro ofs To do this we have to set up a lot of foundational material which

was missing from the original sources K theory of simplicial schemes regulators to

absolute Ho dge and l adic cohomology of simplicial schemes and an interpretation

of the latter as Ext groups of Ho dge mo dules and l adic sheaves resp ectively This

material is contained in the two app endices which we regard as our main contribution

to the sub ject We hop e they prove to b e useful in other contexts than that treated

in the main text

Other parts of BD deal with the weak version of the Zagier conjecture We

do not treat this since a complete pro of has b een given by de Jeu Jeu although

by somewhat dierent means from those used in BD

We see two main groups of pap ers related to p olylogarithms

The rst deals with mixed sheaves ie variations of Ho dge structure or l adic

mixed lisse sheaves Maybe the nucleus of these pap ers is Delignes observation that

the analytic and top ological prop erties of the dilogarithm Li viewed as a multivalued

holomorphic function on P C nf g can b e co ded by saying that Li is an entry

of the p erio d matrix of a certain rank three variation of Q TateHodge structure on

P n f g

C

We refer to Rm section for a nice survey of the construction of a pro

variation on P n f g containing all Li The etale analogue is constructed in

k

C

Beilinsons Polylogarithm and Cyclotomic elements B where he dened pro

ob jects in the categories of l adic sheaves on P nf g In b oth settings the bres

at ro ots of unity dierent from coincide with the cyclotomic extensions mentioned

ab ove

The hop e and indeed the motivation underlying these pap ers is that once a

satisfactory formalism of motivic sheaves is developed the denition of p olylogarithms

should basically carry over We would thus obtain p olylogarithmic classes in Ext

groups of motives these groups b eing supp osedly closely connected to K theory of

which everything already dened on the level of realizations would turn out to b e the

resp ective regulator

Nowhere is this hop e do cumented more manifestly than in Beilinsons and

Delignes Interpretation motivique de la conjecture de Zagier reliant p olylogarithmes

et regulateurs BD if there is such a motivic formalism then the weak version of

Zagiers conjecture necessarily holds not only the values at ro ots of unity of higher

logarithms but also appropriate linear combinations of arbitrary values must lie in

the image of the regulator

For the time b eing and in each case separately honest work is needed to p erform

the K theoretic constructions and calculate their images under the regulators

The second class of pap ers is concerned with precisely that task In analogy with

the ab ove one should rst mention Blo chs Application of the dilogarithm function

in algebraic K theory and algebraic geometry Bl

Beilinsons Higher regulators and values of Lfunctions B provided the

K theoretic construction of cyclotomic elements together with the computation of

their images in Deligne cohomology lo c cit Theorem Neu E

Documenta Mathematica

Annette Huber Jorg Wildeshaus

As for Zagiers conjecture we mention Goncharovs Polylogarithms and Motivic

Galois Groups Go where Zagiers conjecture including the surjectivity statement

is proved for K of a number eld and de Jeus Zagiers Conjecture and Wedge

Complexes in Algebraic K theory Jeu which contains the pro of of the weak

version of Zagiers conjecture indep endently of motivic considerations for K of

n

a number eld and arbitrary n

Typically the ob jects of interest in this class of pap ers are complexes co cy

cles and symbols ie ob jects which do not constantly aord a geometric or sheaf

theoretic interpretation It is by no means easy to see say how a concrete element

in some Deligne cohomology group can b e interpreted as an extension of variations

of RHo dge structure These and similar diculties present themselves to the reader

willing to translate from one class to the other

The authors like to think of the present article as an attempt to bridge the gap

b etween the two disciplines

In a sense the coarse structure of the article follows the ab ove scheme sections

are entirely sheaftheoretic Anything we say there is therefore a priori restricted

to the level of realizations ie nonmotivic In sections K theory enters The

app endices provide the foundations necessary to connect the two p oints of view

Given that quite a lot has b een said ab out the l adic and Ho dge theoretic incar

nations of the classical p olylogarithm B BD WiIV the reader may wonder

why sheaf theoretic considerations still take up one third of this work

Indeed the construction of the motivic p olylog could b e achieved much more eas

ily if a satisfactory formalism of mixed motivic sheaves were available The necessity

to replace a simple geometric situation by a rather complicated one in order to replace

complicated co ecients like Log by Tate twists should b e seen as the main source of

diculty in any attempt to the construction of motivic versions of p olylogarithms

We now turn to the description of the ner structure of the main text sections

In section we normalize the sheaf theoretic notations used throughout the

whole article

Section gives a quick axiomatic description of the logarithmic sheaf Log and

the small p olylogarithmic extension pol The universal prop erty is needed

only to connect the general denition of the logarithmic sheaf as a solution of a

representability problem to the somewhat ad ho c but much more geometric denition

of section A reader prepared to accept the results on the shap e of the Ho dge

theoretic and l adic incarnation of the p olylogarithm may therefore take

the constructions in sections and as a denition of b oth Log and pol and view

section as an extended introduction providing background material

In section we establish the geometric situation used thereafter As section

it is mainly intended for easier reference

In section we construct a prounip otent sheaf G on U P n f g as

pro jective limit of relative cohomology ob jects of p owers of G over U relative to

m

certain singular subschemes The transition maps are given by the b oundary maps in

the relative residue sequence The universal prop erty then allows to identify

G with the restriction of Log to U

Section contains a geometric pro of of the splitting principle the bres

of Log at ro ots of unity have split weight ltration Since we need a pro of which

Documenta Mathematica

Classical Motivic Polylogarithm

translates easily to the motivic situation we return to Beilinsons original approach

to the splitting principle B which consists of an analysis of the action of the

multiplication by natural numbers on our absolute cohomology groups

The main ob jective of section is the description of pol in terms of geometric

data The Leray sp ectral sequence suggests that oneextensions of Q by Log

should b e describ ed as elements of the pro jective limit of cohomology groups with

Tate co ecients of p owers of G relative to certain subschemes The main result

m

allows to identify pol under this corresp ondence

In Section our main to ol the residue sequence is constructed in the setting of

motivic cohomology Prop osition and Lemma The arguments are very much

parallel to those used for absolute cohomology of realizations in section However

we have to replace the singular schemes by explicit simplicial schemes with regular

comp onents This is where the material of App endix B enters

Section is the K theoretic analogue of section We consider a certain pro

jective system of motivic cohomology groups In order to identify its pro jective limit

Corollary we use bijectivity or at least controlled injectivity of the regulator

to Deligne cohomology and the results of section We are then able to dene the

universal motivic p olylog

In the nal section the motivic version of the splitting principle is shown

Again we strongly use the known b ehaviour of the regulator to show that the action

of multiplication by natural numbers splits into eigenspaces Applied to the universal

motivic p olylogarithm this induces the cyclotomic elements in motivic cohomology In

the light of section it is clear from their very construction that they induce the right

elements not only in Deligne but also in continuous etale cohomology We conclude

by drawing the corollaries which are the main results announced at the b eginning

The App endices can b e read indep endently of the main text and of each other

They are meant to b e used as a reference but a careful reader might actually want

to read them rst We refer to the resp ective introductions for an account of their

content

The reader might nd it useful to consult HW for an overview of the strategy

of the pro of of the main results

Acknowledgments It is a pleasure for us to thank b oth Beilinson and Deligne

for not only letting us work with their ideas but explicitly encouraging us to bring

mathematics into a state in which the constructions of BD can b e p erformed

We thank C Deninger for suggesting to us that the metho ds developed and

results obtained in our resp ective PhD theses H Wi might form a sound basis

of a successful treatment of this theme

We are grateful to G Banaszak H Esnault W Ga jda T Geisser U Jannsen

M Joachim K K unnemann F Lecomte AJ Scholl and J Schurmann for useful

discussions We thank R de Jeu for his remarks on App endix B and C Soule for

making GSo available and explaining some critical p oints We are very grateful to

the referee for her or his helpful and detailed comments on App endix B

Part of this work was written up during a stay at the Fields Institute at Toronto

in Octob er We wish to thank M Kolster and V Snaith for the invitation

A

Finally we are obliged to Mrs Gabi Weckermann for L T Xing part of this pap er

E

Documenta Mathematica

Annette Huber Jorg Wildeshaus

Contents

Mixed sheaves

The Logarithmic Sheaf and the Polylogarithmic Extension

The Geometric SetUp

Geometric Origin of the Logarithmic Sheaf

The Splitting Principle Revisited

Polylogs in Absolute Cohomology Theories

Calculations in K theory

Universal Motivic Polylogarithm

The Cyclotomic Case

A Absolute Hodge Cohomology with Coefficients

A Algebraic Mixed Ho dge Mo dules

A Algebraic Mixed Ho dge Mo dules over R

B K Theory of Simplicial Schemes and Regulators

B Generalized Cohomology Theories

B K theory

B Cohomology of Ab elian Sheaves

B Continuous Etale Cohomology

B Absolute Ho dge Cohomology

B A Combinatorial Lemma

Mixed sheaves

We start by dening the sheaf categories which will b e relevant for us For our

purp oses it will b e necessary to work in the settings of mixed l adic p erverse sheaves

H and of algebraic mixed Ho dge mo dules over R A Since the pro cedures are

entirely analogous we introduce for economical reasons the following rules whenever

an area of pap er is divided by a vertical bar

the text on the left of it will concern the Ho dge theoretic setting while the text on

the right will deal with the l adic setting Of course we hop e that b efore long there

will b e a satisfactory formalism of mixed motivic sheaves providing a third setting to

Documenta Mathematica

Classical Motivic Polylogarithm

which our constructions can b e applied We let

l a xed prime number

A R

A Z

l

F Q

F Q

l

and set B Sp ecA

For any reduced separated and at scheme X of nite type over B we let

X X C as a top ol space

X X Q

top

top A

ShX PervX Q

ShX PervX Q

top top

top top l

the latter categories denoting the resp ective categories of p erverse sheaves on X

top

BBD

Next we dene the category ShX in the l adic setting we x a pair S L con

sisting of a horizontal stratication S of X H x and a collection L fLS j S

Sg where each LS is a set of irreducible lisse l adic sheaves on S For all S S

and F LS we require that for the inclusion j S X all higher direct images

n

R j F are S Lconstructible ie have lisse restrictions to all S S which are

extensions of ob jects of LS We assume that all F LS are pure

We can make this more explicit in our computations X will always b e a lo

n

cally closed subscheme of some A the stratication is by the number of vanishing

n

co ordinates in A LS is the set of all Tate sheaves on S

b b

Following H x we dene D X Q as the full sub category of D X Q

l l

c

SL

of complexes with S Lconstructible cohomology ob jects Note that all ob jects will

b

b e mixed By H x D X Q admits a p erverse tstructure whose heart we

l

SL

denote by Perv X Q

l

SL

Sh X Perv X Q ShX MHM XR

l Q

SL

see A

Because of the horizontality requirement in the l adic situation we have the full

b

U Q of the formalism of Grothendiecks functors only on the direct limit D

X l

m

b

D X Q for U op en in B and S L as ab ove see H x However for a

U l

SL

xed morphism

X Y

we have a notion of eg admissibility for a pair S L this is the case if



b b b

D U Q D X Q D U Q

Y l l X l

m m

SL

b

factors through some D Y Q Our computations will show at least a p osteriori

l

TK

that for our choice of S L all functors which app ear are admissible We will not

stress these technical problems and even suppress S L from our notation

Documenta Mathematica

Annette Huber Jorg Wildeshaus

As in BBD we denote by Hom etc the resp ective functors on the cate

gories

b b

b b

D ShX D X Q

D Sh X D MHM XR

l

Q

SL

q

and H for the p erverse cohomology functors

We refer to ob jects of ShX as sheaves and to ob jects of ShX as top ological

top

sheaves Let us denote by

V V

top

the forgetful functor from ShX to ShX If we use the symbol W it will always

top

refer to the weight ltration

If X is smo oth we let

s

lm

s

Sh X Var XR ShX

Sh X Et X ShX

Q

Q

l

see A

the category of lisse

s

mixed Q sheaves on X

Sh X the category of

l

top

s

Q lo cal systems on X

top

Sh X the category of

top

lisse Q sheaves on X

l top

s s

We refer to ob jects of Sh X as smo oth sheaves and to ob jects of Sh X as

top

s

smo oth top ological sheaves Denote by U Sh X the category of unip otent ob jects

s

of Sh X ie those smo oth sheaves admitting a ltration whose graded parts are

s

pullbacks of smo oth sheaves of Sh B via the structure morphism Similarly one

s

denes U Sh X

top

Remark Note that in the l adic situation the existence of a weight ltration ie

W

an ascending ltration W by subsheaves indexed by the integers such that Gr is

m

s

of weight m is not incorp orated in the denition of Sh compare the warnings in

H x In the Ho dge theoretic setting the existence of a weight ltration is part of

the data

Remark We have to deal with a shift of the index when viewing eg a variation as

a Ho dge mo dule which o ccurs either in the normalization of the embedding

b

Var XR D MHM XR

Q Q

or in the numbering of cohomology ob jects of functors induced by morphisms b etween

schemes of dierent dimension In order to conform with the conventions laid down

in app endix A and WiI chapter we chose the second p ossibility a variation is

a Ho dge mo dule not just a shift of one such Similarly a lisse mixed Q sheaf is a

l

p erverse mixed sheaf Therefore if X is of pure relative dimension d over B then the

embedding

lm

b

U Q Et X D

X l

m

Q

l

asso ciates to V the complex concentrated in degree d whose only nontrivial coho

mology ob ject is V

As a consequence the numbering of cohomology ob jects of the direct image say

will dier from what the reader might b e used to eg the cohomology of a curve

Documenta Mathematica

Classical Motivic Polylogarithm

is concentrated in degrees and instead of and Similarly one has to

s

distinguish b etween the naive pullback of a smo oth sheaf and the pullback

b s

on the level of D ShX lands in the category of smo oth sheaves while

of a smo oth sheaf yields only a smo oth sheaf up to a shift

In the sp ecial situation of pullbacks we allow ourselves one notational inconsis

tency if there is no danger of confusion eg in Theorem we use the notation

also for the naive pullback of smo oth sheaves Similar remarks apply for smo oth

top ological sheaves

For a scheme a X B we dene

b

F n a F n D ShX

X

where F n is the usual Tate twist on B

If X is smo oth we also have the naive Tate twist

s

F n Sh X ShX

on X If X is of pure dimension d then we have the equality

F n F n d

X

In order to keep our notation transparent we have the following

Definition For any morphism X S of reduced separated and at

B schemes we let

b b

R X D ShX D ShS

S

i i b

H X H D ShX ShS

S

Definition For a closed reduced subscheme Z of a separated reduced at B

b

scheme X of nite type with complement j U X and an ob ject M of D Sh X

dene

a R X M R Hom b F M

abs X

D ShX

i i

H X M H R X M

abs

abs

the absolute complex and absolute cohomology groups of X with co ecients in M

b R X n R X F n

abs abs X

i i

H X n H X F n

X

abs abs

c R X rel Z n R X j F n

abs abs U

i i

H X rel Z n H X j F n

U

abs abs

the relative absolute complex and relative absolute cohomology with Tate co ecients

Documenta Mathematica

Annette Huber Jorg Wildeshaus

In the Ho dge setting absolute cohomology with Tate co ecients coincides with

Beilinsons absolute Ho dge cohomology over R Theorem A In the l adic set

ting it yields continuous etale cohomology see the remark following Denition B

Remark If X is a scheme over S then we have the formulae

R X R S R X

abs abs S

i i

H X H S R X

S

abs abs

The Logarithmic Sheaf and the Polylogarithmic Extension

We aim at a sheaf theoretic description of the small classical p olylogarithm on

P n f g The rst step is an axiomatic denition of the logarithmic prosheaf

We need the following result

Theorem Let X b e the complement in a smo oth prop er B scheme of an NC

divisor relative to B SGA Exp XI I I all of whose irreducible comp onents

are smo oth over B Let x X B and write a X B The functor

s s

x U Sh X Sh B

is representable in the following sense

a There is a proob ject

s

G en proU Sh X

x

the generic prounip otent sheaf with basep oint x on X which has a weight

ltration satisfying

s

G en W G en U Sh X for all n

x n x

Note that this implies that the direct system

R a Hom G en W G en V

x n x

nN

s

of smo oth sheaves on B b ecomes constant for any V U Sh X

This constant value is denoted by

a Hom R G en V

x

b There is a section

B x G en

x

s s

c The natural transformation of functors from U Sh X to Sh B

G en x ev R a Hom

x

x

Documenta Mathematica

Classical Motivic Polylogarithm

s

is an isomorphism Similarly for the transformation of functors from U Sh X

top

s

to Sh B

top

G en x ev R a Hom

x top

x

Consequently the pairs G en and G en are unique up to unique

x x top

isomorphism

d The natural transformations of functors

s s

Hom G en Hom F x and

x

U Sh X Sh B

s

Hom G en B x

x top top

U Sh X

top

s s

from U Sh X and U Sh X resp ectively are isomorphisms

top

Pro of For ac we refer to

WiI Remark d after Theorem WiI Theorem i

s

and B and lo c cit Theorem ii Apply the functors Hom F

top

Sh B

to the result in c in order to obtain d

Remark In the Ho dge setting and for the constant base B Theorem is

equivalent to the classication theorem for admissible unip otent variations of Ho dge

structure HZ Theorem In this case G en is the canonical variation with base

x

p oint x of lo c cit section

Now let

G G U P n f g

m mB B

B

j U G

m

p G B pe pj U B

m

We may form the generic prounip otent sheaf with basep oint on G

m

s

Definition Log G en pro U Sh G is called the logarithmic prosheaf

m

As we shall see b elow there is an isomorphism

Y

W

Gr Log F k

k

Assuming this for the moment we now describ e the higher direct images

H U j Log

B

q

Theorem a H U j Log for q

B

H U j Log is split b H U j Log has a weight ltration and W

B B

More precisely any isomorphism as ab ove induces an isomorphism

Y

W H U j Log F k

B

k

Documenta Mathematica

Annette Huber Jorg Wildeshaus

Remark By these statements on the higher direct images of the prosheaf j Log

we mean the following

a For q the pro jective system

q

H U j Log W Log

n

B n

is MLzero

b induces a morphism of pro jective systems

m

Y

H U j Log W Log F k

m

B

m

k

m

of sheaves with a weight ltration such that the weight parts of the

pro jective systems of kernels and cokernels are MLzero

Pro of One uses the exact triangle

id

G

m

j j

of it and the fact that H G Log is easily computable For or rather H G

m m

B B

the details see WiI I I Theorem Or use and whose pro of is indep endent

of

A xed choice of

Y

W

Gr Log F k

k

W

induces in particular an isomorphism of Gr Log and F The theorem then en

ables one to dene the small p olylogarithmic extension as the extension

W

Gr Log j Log j pol Ext

s

U U

U Sh U

Q

F k under the isomorphism mapping to the natural inclusion F

k

F Log j Hom F Log j Ext

b

U U U

D ShU

ShU

Hom b pe F j Log

D ShU

Y

A

Hom F F k

ShB

k

induced by the pro jective limit of the edge homomorphisms in the Leray sp ectral

sequence for pe and the isomorphism of b Note that the denition of pol is

indep endent of the choice of For the details we refer to WiI I I Theorem as

there we dene

Ext F Log W Log j Ext F Log j lim

n U U

ShU ShU

n

Documenta Mathematica

Classical Motivic Polylogarithm

A description of Log and pol in b oth incarnations was given by Beilinson and Deligne

see B and BD x for the Ho dge version and B for the l adic

setting The reader may nd it useful to also consult WiIV chapters and

setting N in the notation of lo c cit

We recall the values of pol at sp ectra of cyclotomic elds let d and

T T where T is the dth cyclotomic C Sp ecR where R A

d d

d

p olynomial

C is canonically a closed reduced subscheme of G A For any integer b

m A

d

prime to d there is an embedding

i C C G A

b m A

d

b

Since d is invertible on C the image of i is actually contained in U and hence we

b

may form the pullback of pol via i

b

pol Ext F Log

s

b b

Sh C

where Log denotes the pullback of Log

b

Now we have the following

Theorem Splitting Principle Log splits uniquely into a direct pro duct

b

Y

W

Log Gr Log

b b

k

k

W

and Gr Log is isomorphic to F k for any k

b

k

Pro of B or BD or WiIV Lemma Or use and whose

pro of is indep endent of

In order to identify pol with an element of

b

Y

Ext F F k

s

Sh C

k

we need to x an isomorphism

Y

W

Gr Log F k

b b

k

By denition is the pullback via i of the isomorphism

b b

Y

W

Gr Log F k

k

of prosheaves on G of WiIV chapters and which we briey describ e now

m

By d there is a canonical pro jection

Log F

Documenta Mathematica

Annette Huber Jorg Wildeshaus

Furthermore there is a canonical isomorphism

W

Gr Log p H G F

m

B

given by the fact that b oth sides are equal to p of the mixed structure on the

ab elianized G see WiI chapter

mtop

Observe that there is an isomorphism

res H G F F

m

B

given by the map residue at

Q

W

Finally b oth Gr Log and F k carry a canonical multiplicative structure

k

W

for Gr Log this is a formal consequence of

WiI Corollary ii WiI Corollary ii

see Remark b at the end of chapter of lo c cit

Our isomorphism

Y

W

Gr Log F k

k

is the unique isomorphism compatible with res and the multiplicative struc

ture of b oth sides

Using the framing of Log given by we may identify pol with an element of

b b b

Y

Ext s F F k

Sh C

k

or after twisting and forgetting the comp onent k as an element of

Y

F F k Ext

s

Sh C

k

Note that in the Ho dge setting we do not lose any information by forgetting the

s

comp onent k as there are no nontrivial extensions in Sh C of F by itself

This latter statement fails to hold in the l adic context It is however true that the

zerocomp onent of pol is trivial One way to see this is via WiI I I Corollary

b

where it is proved that there is in fact a mixed realization pol of which the ab ove

b

extensions are merely the Ho dge and l adic comp onents In the category of mixed

realizations there is a go o d concept of p olarization which ensures that there are

no nontrivial extensions of pure realizations of the same weight Alternatively one

uses Theorem where it is proved that our pol lie in the image of the resp ective

b

regulators The claim then follows from the vanishing of H C

M

Theorem Beilinson Under the isomorphism of A we have in the Ho dge

setting

Y M

k b k

A

pol Li C i Q

b k

k

k

C C

P

n

z

for jz j and z where Li z

k

k

n

n

Documenta Mathematica

Classical Motivic Polylogarithm

Pro of B or BD i or WiIV Theorem

Note that one may identify C C with f Q C g by asso ciating to the

d

unique embedding mapping T Q Q T T to

d d

Q It allows to identify In the l adic situation choose a geometric p oint C

C and

Sp ec Z

l d

and furthermore the category of continuous Q mo dules under the Galois group of

l

lm

s

Q that are mixed and unramied outside l d and the category Sh C Et C

Q

l

Given this we think of Ext Q Q k as sitting inside

s

l l

Sh C

H Q Q k

l

cont

Together with the natural map of Lemma B we thus have an inclusion of

Ext Q Q k into

s

l l

Sh C

1

GalQ Q

l

r

k

l

A

1 1

Q Q Q lim

r

Z l l l

l

l

r

Theorem Beilinson Under the ab ove inclusion we have in the

l adic setting

X

d k k

A

pol

b

k

d k

r

l b

rk

Pro of B or BD ii or WiIV Theorem

Remarks a Using the dening prop erty of pol one can show see B or

BD pro of of that it coincides with a sp ecic sub quotient of the generic

prounip otent sheaf on U The sp ecializations to sp ectra of cyclotomic elds of this

sub quotient were already studied in D section In particular Theorems and

are equivalent to the Ho dge and l adic versions of D Theoreme

b One of the main results of this work will b e Theorem that the elements in

and for xed b and d are the resp ective regulators of one and the same element

in motivic cohomology This implies that Soules construction of cyclotomic elements

in the K theory with Z co ecients of an ab elian number eld Sou Lemma

l

Sou actually factors over the image of K theory prop er Corollary As shown

in BlK x Theorem also implies that the Tamagawa number conjecture mo dulo

p owers of is also true for o dd Tate twists see our Corollary Finally is used

in KNF Theorem to prove the mo died version of the Lichtenbaum conjecture

for ab elian number elds

c There are relative versions of and for schemes over a base scheme S smo oth

over B They allow to directly dene the small p olylogarithmic extension pol on

S

U S which however turns out to b e the base change to S of pol

B

Documenta Mathematica

Annette Huber Jorg Wildeshaus

Remark In our denition of pol we chose not to follow BD The approach via

the universal prop erty of Log and the computation of its cohomology rather imitates

that of Beilinson and Levin in the elliptic case BL In fact one of the

predecessors of lo c cit contains a unied denition of Log and pol for relative curves

of arbitrary genus BLp

The Geometric SetUp

For easier reference we assemble the notation used in the next sections

As b efore we let

l a xed prime number

A R

A Z

l

B Sp ecA

G G U P n f g

m mB B

B

denote a smo oth separated scheme over B of pure relative Furthermore we let S

dimension dS

G S

m

the op en subscheme of S where and are disjoint We assume S to b e dense S S

in S

j S S

n S S i S

n S is equipp ed with the reduced scheme structure where S

Z S S

with the reduced scheme structure

G n Z V

mS

For n dene

n n

S G p

mS

n

n n

V v G

mS

n n

n n n

G G Z n V z

mS mS

n

v and Z where Z carries the reduced scheme structure So p id

S

Documenta Mathematica

Classical Motivic Polylogarithm

The base change of the ab ove ob jects and morphisms to S is denoted by the same

letters not underlined

S G

mS

Z S q S

V G n Z

mS

n n

p G S

mS

n n n

v V G

mS

n n n

z Z G

mS

n

Also we dene partial compactications of p

n n n

g G A

mS S

n n n n n

h H A n G A

S mS S

n

where again H has the reduced structure

n n

p A S

S

V A n Z

S

n

n n

v V A

S

n n

n n n

z Z A n V A

S S

n

Z is equipp ed with the reduced structure So Z Z Z where

Remarks a The under lined ob jects should remind the reader that the partial com

overlined ob jects pactication comes from the compactication j of the base S The

n

refer to compactication upstairs induced from g

b For xed n we have a natural action of the symmetric group S on our geometric

n

situation

For the purp oses of K theory in section we will have to replace the singular

n

scheme Z by some smo oth simplicial scheme Put

n

n n n

Z Z G q G Z G q q G Z

S mS S S

mS mS mS

n

n

is a prop er covering of Z This is the easiest case of a morphism Note that Z

of schemes with cohomological descent meaning that for any reasonable cohomology

n

theory the cohomology of Z will agree with the cohomology of the smo oth simplicial

scheme

n

n n

Z cosk Z G

mS

ie

n n n

n n

k fold pro duct Z Z Z

G G

k

mS mS

Documenta Mathematica

Annette Huber Jorg Wildeshaus

Put Z corresp onding to the empty scheme We will also use the simplicial

n n

n

Z Z sitting in A in the same way Finally let which is attached to scheme

S

n n n

G ConeZ G

mS mS

n

n n

A Cone Z A

S S

where the cone is taken in the category of p ointed simplicial sheaves on the big Zariski

site cf the discussion in app endix B

Geometric Origin of the Logarithmic Sheaf

In section we dened a prosheaf

s

Log pro U Sh G

m

and an element

F Log j pol Ext

U

ShU

F Log W Log j lim Ext

n U

ShU

n

The aim of this section is to identify Log j or rather its No etherian quotients

U

as relative cohomology ob jects with co ecients in Tate twists of certain schemes over

U Theorem

Recall that according to our conventions we have

F F

U

and hence we may view pol as an element of

Hom b F Log j H U Log j

U U U

D ShU

abs

where we have used the notation introduced in Denition

For the schemes of section we have the following

Definition For n

sgn sgn

ndS

n n n n n

n

G v F n H G v F n G H

V

mS mS S

S

where the sup erscript sgn refers to the signeigenspace under the natural action of

n n

the symmetric group S on G and V

n

mS

Observe in particular that G F

The following is an immediate consequence of the K unneth formula

Lemma There is a canonical isomorphism

n

n

Sym G G

Documenta Mathematica

Classical Motivic Polylogarithm

n

We want to compute G and simultaneously construct for each n a

pro jection

n n

G  G

via the residue at whose pro jective limit over n we shall then identify for sp ecial

and and S U with the restriction Log j of the logarithmic prosheaf to U

U

n n n

n n

the smo oth part b e the singular part of H and H H n H Let H

reg

sing sing

n

n

We the subscript reg will mean the complement of H For any subscheme of A

S

sing

work with the following geometric arrangement

n n

n

n

V H V V

reg

reg reg

n n

n

v v

v

y y y

Hreg reg

n n

h g

reg reg

n

n n

A H G

reg

Sreg mS

Both squares are cartesian All maps are either op en or closed immersions and each

line gives in fact a smo oth pair of S schemes

n

b n

v M is a shift of a Lemma For any complex M D ShA such that

reg

Sreg

n

V smo oth sheaf on there is an exact triangle

reg

n n

n n n n

h v v v v M h M

reg reg

reg reg

H reg H reg

n n n n

M v g v g

reg reg

n

v applied to the exact triangle obtained from purity for the closed Pro of This is

reg

immersion

n n

n

V H V

reg reg reg

of smo oth schemes

n

We apply this lemma to M F n and evaluate the cohomological functors

A

Sr eg

n sgn i

A on the triangle Following c we write everything as relative H

Sreg

abs

cohomology with Tate co ecients

n

n i n sgn n sgn i

rel A H rel Z n G Z n H

Sreg abs mS abs reg

n

i

n n sgn

H H rel Z H n

reg reg

abs

n

i

sgn n

Z n H A rel

reg Sreg

abs

We refer to this as the absolute residue sequence

i n sgn

Application of the cohomological functors H A to the same exact

S Sreg

triangle yields a long exact sequence of sheaves on S that we call the relative residue

Documenta Mathematica

Annette Huber Jorg Wildeshaus

sequence

sgn

sgn

n n i n i n

n

n

G v F n H v F n H A

V

mS S reg S Sreg

V

reg

sgn

i

n n

H H n v F n

n

reg H reg

S

V H

reg

reg

sgn

i

n n

n

A H v F n

Sreg reg

S

V

reg

sgn

ndS

n n n

n

o ccurs in this sequence F n v G Note that G H

V

mS

S

We are now going to further analyse and reshap e these sequences The nal

form will b e achieved in Prop osition and Theorem

First we need to identify the terms

n

i

n sgn n

Z H n n H H rel

reg reg

abs

sgn

i

n n

H H v F n n n

n

reg H reg

S

H V

reg

reg

n

n n

Z H in H is given by The complement of

reg reg

n

n n n

v V H H

H reg reg reg reg

n

n

n

n

V H V under the identication Since

reg

reg

k

n

a

n

n

H G

reg

mS

k

and these comp onents are p ermuted transitively by S we conclude

n

Lemma

n

a

n n

`

n

n

v F n a F n n v

n

V

H reg

V H

k

reg

reg

k

n

M

n

i n i

n n n

b rel Z n Z H n H G H H rel

reg reg

mS

abs abs

k

n

n n

i

sgn

and hence the signeigenspace H H rel Z H n is isomorphic to

reg reg

abs

i n

n sgn

H G rel Z n

mS

abs

where the last sgn refers to the action of S The isomorphism is given by pro jection

n

onto the comp onents unequal to k for some choice k f ng It is indep endent

of the choice of k

n

M

n n

n n

n

F n v c G v F n H R R n

n

S S

V

H reg reg

mS

V H

reg

reg

k

Documenta Mathematica

Classical Motivic Polylogarithm

sgn

n

i

n

v F n As in b the signeigenspace H H is canoni n

n

reg

H reg

S

V H

reg

reg

cally isomorphic to

sgn

i n n

n

H G v F n

V

S mS

n

the latter equals G For i n dS

Pro of The only p oint that remains to b e shown is the indep endence of the isomor

phisms in b and c of the choice of k Recall the identity

sgn

sgn

n n n n

n

G rel Z n G v F n S R R R

V S abs abs

mS mS

q

sgn n n

n

for q n dS We are going to prove in d that H G v F n

V

mS

S

So the asso ciated sp ectral sequence degenerates and shows that the indep endence of

the map in b follows from that of the map in c

ndS

n sgn n n

n

For c we only need to consider G H There our F n G v

V

mS

S

claim follows from Lemma and the gradedcompatibility of the cup pro duct with

b oundary morphisms GH Prop osition and Corollary

Remark The arguments of this section would b ecome simpler if we could use an

sgn

in c However we do not know whether it is p ossible to make a decom ob ject R

S

p osition into eigenspaces in our triangulated categories

By the identication of the lemma the residue sequences dene canonical residue

maps

i n

i n n sgn n sgn

res H G rel Z n rel Z n H G

abs mS

mS

abs

sgn

sgn

i n n

n n i

n

n

G v H F n res H v F n G

V

V

mS S

S mS

tting into the relative and absolute residue sequences In particular observe that

we have a residue map

n n

res G G

Now we concern ourselves with the identication of the remaining terms

n

sgn i n

Z n n H A rel

reg abs Sreg

sgn

i n n

n

n H v A F n

S reg Sreg

V

reg

of the residue sequences

n

We use the following ltration of A by op en subschemes

S

n n

j at most k co ordinates vanishg fx x A F A

n k

S S

n n n n

So we have F A A and F A G

n

S S S mS

The graded pieces of this ltration are

n n n

n F A F A G A

k k k

S S S

n

fx x A j precisely k co ordinates vanishg

n

S

Documenta Mathematica

Annette Huber Jorg Wildeshaus

n

is equipp ed with the reduced scheme structure Note that it splits into several G A

k

S

disjoint pieces For k and any such piece there is a transp osition of S acting

n

trivially By using triangles similar to for the inclusions

n n n

G A F A F A

k k k

S S S

n

is trivial we conclude inductively that the signeigenpart of the cohomology of H

sing

Lemma The adjunction morphism induces isomorphisms

n n

i n sgn i n sgn

H A rel Z n H A rel Z n

abs S abs Sreg reg

sgn

sgn

i n i n n n

n n

H A H A v F n v F n

S Sreg S S reg

V V

reg

By b and the absolute residue sequence takes the form

n

sgn i n n sgn i n

Z n H G rel Z n H A rel

abs mS abs S

res

i n

n sgn

H G rel Z n

mS

abs

n

i

n sgn

H A rel Z n

S

abs

Similarly the relative residue sequence lo oks as follows

sgn

sgn

i n n n i n

n

n

H v F n G v F n H A

V

S mS S S

V

sgn

res

n n i

n

G v F n H

V

mS S

sgn

i

n n

n

H A v F n

S

S V

For the computation of the term

sgn

n i n

n

v F n H A

S S

V

we use the K unneth formula

n n

n n

Lemma a R A v v F n H A F n and the K unneth formula

S

S S S

gives an isomorphism

sgn

n

n n n n

A v v Sym H v F n H F n F A H A

S S S S S S

b The choice of an ordering of the sections and gives an isomorphism

A v v F H F F R A

S

S S S

Up to sign it is canonical

c The isomorphisms of a and b induce an isomorphism

ndS

n n n n

n

A H v F n v F n F n H A

S S S

S

V

n

It dep ends on the choice made in b only up to the sign The group S acts

n

on these ob jects via the sign character

d For i we have

sgn sgn

indS

i n n n n

n

H G v F n G v F n H

V

S mS mS

S

Documenta Mathematica

Classical Motivic Polylogarithm

Pro of For b consider the long exact cohomology sequence asso ciated to the triangle

v F F

z F

We have

F i

i

H A F

S S

i

and

M

F i

i

A H z F

S S

i

The long exact cohomology sequence thus reads

M

H A A v v F F F F H

S S S

S

If we let f g fs s g then we identify the cokernel of

M M

F F F

i

L

f F to f f with F by mapping f

s s s s

i

N

n

n

a follows from b since F Sym F

c is a consequence of a and b

d follows from a and the relative residue sequence by induction on n

On the level of absolute cohomology the isomorphism of c induces an iso

morphism

sgn

n n

in in

n n i

H A rel Z n H A Z n rel H S n

S S abs

abs abs

This gives the nal shap e of the absolute residue sequence

in

i n n sgn

H S n H G rel Z n

abs mS

abs

res

in n

n sgn

n G rel Z H

mS

abs

i

H S n

abs

By d the relative residue sequence collapses into the short exact sequence

of sheaves on S

res

n n

F n G G

In order to identify the long exact absolute cohomology sequence asso ciated to

this sequence with the absolute residue sequence we need the following

Documenta Mathematica

Annette Huber Jorg Wildeshaus

b

Lemma Let K D ShX b e a complex of sheaves on a separated reduced and

at B scheme X Supp ose there is an action of a nite group G on K Let b e

the character of an absolutely irreducible representation of G over F For any ob ject

V with a Gaction of an F linear ab elian category denote by V the isotypical

comp onent of V ie the image under the pro jector

X

g g e

G

g G

i

Supp ose that H K vanishes for all i Then

Hom b F K i Hom b F H K i

D D

Pro of By applying e and e one checks the statement for a complex of the

H K For the general case consider the sp ectral sequence for sp ecial form K

Hom b F i induced by the truncation functors It degenerates after applying

n

D

e

Now that we know that formation of absolute cohomology commutes with for

mation of sign eigenspaces we have

Proposition The absolute residue sequence is the long exact sequence in abso

lute cohomology attached to the short exact sequence

res

n n

G F n G

We conclude the computational part of this section by collecting our results

Theorem a For n we have

sgn

n n n

G v F n G H

mS S

sgn

i n n

and H G v F n for i

S mS

b The residue at ie the b oundary map of gives an epimorphism

n n

res G  G

for n

c The K unneth formula gives an isomorphism

sgn

n n n n

H A v v kerres F n H F n A

S S S S

for n A choice of an ordering of the sections and induces an isomorphism

F n kerres

n

which dep ends on this choice only up to the sign

n

n

d Let G Sym G b e the canonical isomorphism of and

n

Sym F F

n

Sym F F n

Documenta Mathematica

Classical Motivic Polylogarithm

the isomorphisms given by multiplication Then the diagrams

n

G F

o o

n n

Sym G Sym F

and

n

F n G

o o

n n

Sym F Sym G

commute Here the horizontal maps are given by the successive residue maps and

by c resp ectively

n

e Let W G

n

n n n k

W G W G kerG G f or k n

k k

n n

and W G G The choice in c induces isomorphisms

n

M

W

n

F i Gr G

i

which by their construction t into commutative diagrams

L

n

W

n

F i Gr G

i

W

Gr res can

L

n

W

n

Gr F i G

i

n

The ltration W is therefore the weight ltration of G

Pro of a b and c follow from the previous results The commutativity of the rst

diagram in d follows from the denition of the residue map For the second diagram

we use the fact that the K unneth formula of is compatible with the K unneth

formula of the pro of of a For e apply induction on n

Recall that S is the op en subscheme of S where the sections and of G

mS

are disjoint For sp ecial S and the following is the main step towards the

n

identication of the pro jective limit of the G with the restriction Log j of the

U

logarithmic sheaf

n

n

Lemma a There is a unique smo oth sheaf G on S It has a extending G

weight ltration

b There is a canonical isomorphism

n n

Sym G G

and a unique isomorphism

n

M

W n

n

Gr G F i

i

Documenta Mathematica

Annette Huber Jorg Wildeshaus

which is compatible with the isomorphism of e

n

is split there is a canonical isomorphism c The weight ltration of i G

n

M

W n n

i G F i Gr i G

b

i

n S into S Here i denotes the inclusion of S

d There is an exact sequence

v F G G i F H

mS

S

of sheaves on S

Pro of If there is any smo oth sheaf as in a then it will automatically b e unique and

hence b follows from a and d e Also it will suce b ecause of d to show

the lemma for the case n

There we have the following diagram

y

i F

y

H F K H F

y

L

F

y

where

M

K H Cone dS i F dS F

S

S nS

with v v v v in terms of constructible sheaves this is just Ker shifted in

the appropriate degree to dene a p erverse sheaf The horizontal sequence is as in

the pro of of b the long exact cohomology sequence on S asso ciated to the short

exact sequence on G

mS

z F v F F

where we have set

i i

F G H H v

mS

S

We thus get the equality

v F H v F R G G

S mS mS

S

Documenta Mathematica

Classical Motivic Polylogarithm

and an exact sequence of sheaves on S

K F H v F F G

mS

S

whose restriction to S is isomorphic via the choice of an ordering of and to

F G F

Push out of the ab ove via the morphism

M

A

K F F F

whose kernel is i F recall again that we use p erverse indices gives the desired

extension G By construction b and d hold Applying i to the pushout diagram

is the pushout of F via and taking cohomology we see that the sheaf i G

F

and we get c

We now sp ecialize our geometric situation we let

S G

mB

G  B G

mB mB

id G G

mB mB

n S the closed subscheme of G given by the So we have S U and S

B mB

immersion of B into G

mB

After having made precise which choice of normalization we have and in how far

it aects our identications we now x it we let

s and s id in c

n

We thus get a pro jective system G of smo oth Tate sheaves on G with

n mB

n

M

n

G F i j

B

i

By the universal prop erty of Log Theorem d there is a unique morphism

n

Log G G lim

n

to such that j sends B Log j

B

B

Y

F i G F j

B

i

Documenta Mathematica

Annette Huber Jorg Wildeshaus

Theorem is an isomorphism

Pro of The claim can b e shown on the level of the underlying top ological sheaves The

l adic statement follows from the statement for the top ological spaces of C valued

p oints by comparison recall that we are dealing with lo cally constant sheaves

Over C the bre at of the prolo cal system Log equals the completion of the

top

Z with resp ect to the augmentation ideal a group ring Q of G C

m

The representation of is given by multiplication compare the general construction

in WiI In particular we have

n

Log lim Sym Log

top top

n

where Log Log a is of dimension two Now in the category of unip otent

top top

lo cal systems on G C the prosheaf Log has the universal prop erty of Theorem

m top

d

The resulting map factors over G We apply this universal prop erty to G

top top

Since G is twodimensional the representation of Q is necessarily

top

top

trivial on a and we get a morphism of lo cal systems

Log G

top top

top

giving rise to a morphism

n

Sym Log G lim

top top

top

n

Again b ecause of the universal prop erty of Log this morphism is identical to

top top

It therefore suces to show that is bijective which amounts to saying that

top

the coinvariants of G under the action of are onedimensional But taking

top

coinvariants under of a unip otent variation V amounts to computing singular

cohomology

G V H G C V H

mR m

Sp ecR

Firstly we claim that

F i

i

G G v H F

mR mR

Sp ecR

i

eg identify the left hand side with

i

H G C G C G C fg G C F

m m m m

i

H G C G C G C fg fg G C F

m m m m

and apply the K unneth formula From the pro of of we recall remember that

G we have S

m

G G v F H F G G v R

mR mR mR mR G

G

mR

mR

from which we conclude

F i

i

G H G G v H F

mR mR mR

G

Sp ecR

mR

i

Documenta Mathematica

Classical Motivic Polylogarithm

The long exact sequence obtained by applying R G to the exact se

mR

Sp ecR

quence of d

F G F H G G v

mR mR

G

mR

then shows that

F G G H

mR

Sp ecR

Remark The geometric situation used in this section is identical to the one of

BD see in particular lo c cit The comparison statement of our

Prop osition is implicit in lo c cit We mention that basically the same

geometric arrangement was used in Jeu More precisely writing down the iterated

cone construction of lo c cit one arrives at a simplicial ob ject which is

equivalent to Beilinsons and Delignes construction used here

The Splitting Principle Revisited

In order to b e able to translate easily to the motivic context we recall Beilinsons

original pro of B of the splitting of the logarithmic prosheaf over sp ectra of

cyclotomic elds Theorem

First we return to the general situation considered at the b eginning of section

For N we have the morphism of S schemes

G G

mS mS

N

x x

and for each n the induced morphism

n n n

G G

mS mS

We work under the additional assumption

A

n n n

If this is the case we have V V and hence get a morphism

n n n

v F n v F n

and hence a morphism

n n n

v F n F n v

n

which after application of p and pro jection onto the signeigenpart induces

n n n

G G

n n

We need to understand the action of on G and on absolute cohomology

n

First we establish in how far is compatible with the residue at

Documenta Mathematica

Annette Huber Jorg Wildeshaus

Lemma a Under any isomorphism

n

M

W

n

F i Gr G

i

W

n ni

the map Gr is multiplication by N on F i

b For any n the diagram

n

n n

G G

res res

n n

y y

n

N

n n

G G

commutes

Pro of Since the morphisms in b are strict with resp ect to the weight ltration it

suces to check that

W W W W

n n

Gr res Gr N Gr Gr res

n n

W

But if we choose the isomorphism of e then Gr res is simply the canonical

n

pro jection

n n

M M

F i F i 

i i

and therefore b follows from a For a we note rst that it suces to show the

statement for one choice of isomorphism

n

M

W

n

F i G Gr

i

This time we use the isomorphism on graded ob jects induced by thereby reducing

ourselves to the case n There we consider the long exact cohomology sequence

asso ciated to the exact sequence

z F v F F

i

and the cohomological functors H G We know the cohomology of G

mS m

S

F i

i

F i

H G F

mS

S

i f g

Of course we know the cohomology of two p oints

M

F i

i

H F G z

mS

S

i

Documenta Mathematica

Classical Motivic Polylogarithm

We get an exact sequence

M

G v F F F F G H

mS

S

n

and b ecause of assumption A it carries an action of But this action can b e

i i

G z F it is trivial on the F and G F and H identied on H

mS mS

S S

multiplication by N on F

Certainly A is only satised in very sp ecial situations namely if and are

supp orted in the schemes of N torsion of G

mS

Let again d C Sp ecR where R A T T as in section For

d

d

b prime to d consider

i C C G

b m

b

The pullback Log of the prosheaf Log j on U via i is identical to the pro jective

b U b

n

limit of the sheaves G obtained by setting

b

C S

C B G

m

i

b

Since A is satised with N d we may apply and conclude

n

Corollary G splits into a direct sum

b

n

M

n n

W

Gr G G

i

b b

i

Therefore there is a unique isomorphism

n

M

n n

F i G

b b

i

n

which is compatible with the isomorphism of b

n

ni n

Pro of F i G is the eigenspace of d under the morphism

b

We conclude with the implications of and for absolute cohomology with

co ecients For this recall the absolute residue sequence for n

res

n n n

sgn sgn n

H G n n C n H G H

mC abs

mC

abs abs

introduced after where we have set

n n

n sgn sgn n n

rel Z n n H G H G

mC mC

abs abs

thus saving enough space to get the ab ove sequence into a single line

Documenta Mathematica

Annette Huber Jorg Wildeshaus

Corollary a For n the absolute residue sequence splits into short exact

sequences

n n n

n sgn sgn

H C n H G n n H G

abs mC

mC

abs abs

n

b For N d the map acts on the short exact sequences of a there is a

commutative diagram

n n n

sgn sgn n

n H G n C n H G H

mC abs

mC

abs abs

n n

id d

y y y

n n n

n sgn sgn

H C n H G n H G n

abs mC

mC

abs abs

Pro of By the absolute residue sequence is the absolute cohomology sequence for

the exact sequence of sheaves on C

res

n n

b

G F n G

b b

Therefore a follows from while b follows from b and the fact that under

the identication of a

n

n n sgn

H C G H G n

abs mC

abs

the map induced by

n n

n

G G

b b

n

is the map of the absolute cohomology groups

n

n n sgn

It follows that the eigenvalues of on H G n are

mC

abs

n

d d The eigenspace decomp osition yields

n

M

n

n n

n sgn n n sgn

H C i H G n H G rel Z n

abs mC mC

b abs abs

i

which in sheaf theoretic terms corresp onds to the decomp osition

n

M

n n

Ext F F i Ext F G

ShC ShC

b b

i

given by Corollary

The pullback pol of the small p olylogarithmic extension pol on U is an element

b

of

n

n

n sgn n

rel Z n H G Ext F G lim lim

mC ShC

abs b

n n

n

sgn n

n H G lim

mC

abs

n

Documenta Mathematica

Classical Motivic Polylogarithm

n

We have shown that using the eigenspace decomp osition for the action of the

these groups are isomorphic to

Y Y

Ext F F k H C k

ShC abs

k k

and describ e pol as an element in this group

b

Actually in order to relate the ab ove decomp osition to the one used for and

we shall need to compare the isomorphism

Y

W

n

F k Gr G lim

k

n

of b to the isomorphism

Y

W W

Gr Log F k Gr G

k

of section

A priori we know that the isomorphisms

W

Gr G F k

k k

k

satisfy an identity of the type

q

k k k

for a constant q F

k

We remark that in order to prove the main results announced in the introduction

all one needs to know is that q is a rational number which is indep endent of

k

whether we work in the Ho dge or the l adic setting

In order to exhibit the precise relation of the motivic analogue of pol see section

to the cyclotomic elements in K theory see Corollary b we need to identify

q

k

Proposition We have the equality

k

k k

Pro of Because of the compatibility of with the canonical pro jection

G F

we have In order to show we compare the classes of G in

Ext F F

ShG

m

induced by and resp ectively Let

K C K Q

Documenta Mathematica

Annette Huber Jorg Wildeshaus

and choose any K valued p oint t of U Of course the value of q can still b e detected

from the extensions of

mixed Q Ho dge structures

Galois mo dules

of G via t In b oth settings there is a natural morphism given by the pullback t G

of K F into the resp ective Ext F F see eg

Z

WiIV Theorem WiIV Theorem

By

WiIV Prop osition a WiIV Prop osition a

the class of t G calculated in the framing given by equals the image of t K

under this morphism By Sch the same holds for the framing given by

note that here it is vital to choose the ordering of the sections and in the way we

did b efore For k let

k

k k

G Sym G

b e the isomorphism of b By d the diagram

k

F G

k

o o

k k

Sym F Sym G

commutes By WiIV Theorem a the commutativity of this diagram char

k

acterizes uniquely From lo c cit Theorem b and c we know that the

diagram

k

k

F k G

k

o o

k

Sym

k k

Sym F Sym G

commutes So our identity

k

k k

follows from d

Polylogs in Absolute Cohomology Theories

In section we showed that the logarithmic prosheaf is the pro jective limit of

relative cohomology ob jects with co ecients in Tate twists of certain schemes over

U The Leray sp ectral sequence suggests that is should b e p ossible to recover pol as

a pro jective limit of elements in absolute cohomology with Tate co ecients of these

schemes and indeed this is what we do in Theorem That the co ecients are Tate

is of course the central p oint it allows us in section to imitate the construction

Documenta Mathematica

Classical Motivic Polylogarithm

of this section and thus to dene a motivic version of pol This detour is necessary

b ecause we know up to date of no satisfactory formalism of mixed motivic sheaves

whose absolute cohomology with Tate co ecients would give back motivic cohomology

dened via K theory

We return to the geometric situation set up b efore and start by computing

the higher direct images of the restriction of Log to U

and the pro jection Lemma a The inclusion F G

G  F induce natural isomorphisms

H F G G

B m

B

H G F G G H

m m

B B

and the latter group is isomorphic to F via the map residue at

B

n n

b The inclusion F n G and the pro jection G  F induce natural identi

cations

F n i

B

n

i

F i

H G G

B

m

B

i f g

Pro of The statements need only b e checked on the level of lo cal systems Part a

is shown in the pro of of From there we also recall that we have to compute

the invariants and coinvariants under the action of the group G C or

m

equivalently of a generator of Using b we may deduce b from a

Corollary

F n i

B

i n

H U G

B

i f g

n

L

n

F k and an ob ject For i the sheaf H U G is the direct sum of

B

B

k

which is an extension of F by itself

B

i n

Pro of By WiI Theorem there is a weight ltration on H U G Now use

B

the exact triangle

id

G

mB

j j

purity and c

Remark In the setting of Ho dge mo dules where a concept of p olarization is avail

able any extension of pure ob jects of the same weight is necessarily split

n

The map H U G F of the corollary yields in particular a map residue

B

at for n

B

n n

res H U G H B R U G H B

B

abs abs abs

Documenta Mathematica

Annette Huber Jorg Wildeshaus

Definition Let n The map

n

n n n sgn

res H U G H G rel Z n H B

abs mU abs

abs

is called the total residue map

For later reference we note

Corollary H G rel Z

abs mU

Pro of We have

H G rel Z H U G

abs mU

abs

which b ecause of equals H B F

abs

Next we have

Lemma i The transition morphism

n n

res G  G

satises

U res F n F n H

B B

B

n n

H U res H U G H U G

B B B

is surjective with kernel F n

B

In particular the total residue for n factors over the total residue for n

there is a commutative diagram

res

n

H B H U G

abs abs

res res

n

U G H

abs

ii The Leray sp ectral sequences for n give exact sequences

res

n

H B n H U G H B

abs abs abs

The map

n

H B n H U G

abs abs

is the comp osition of H B n H U n H U F n and the map induced

abs abs abs

n

by the inclusion of F n into G in other words the same noted map of the residue

sequence

The pro jective limit of the ab ove sequences identies

n

U G H U Log j lim H

U

abs abs

n

B and H

abs

Documenta Mathematica

Classical Motivic Polylogarithm

iii There are unique splittings

n

s H B H U G

n

abs abs

of the sequences in ii for any n such that for any n we have a commutative

diagram

s

n

n

H U G H B

abs abs

s res

n

n

H U G

abs

Pro of i The rst statement is clear For the second either go through the construc

tion or observe that the direct image of the morphism U B has cohomological

top top

U is right exact on smo oth sheaves dimension one hence H

B

ii We have the Leray sp ectral sequence

pq p q pq

n n

E H B H U G H U G

B

abs abs

whose lowterm sequence reads

n

d

n

H B n H U G H B H B n

abs abs abs abs

By i the MittagLeer condition is satised for the pro jective system

H B n and therefore

n

abs

n

kerd H B H U Log j lim

U

abs abs

n

n

B n is MLzero since the pro jective system imd H

n n

abs

But then any of the

n

H U G H B

abs abs

must b e surjective as well

iii Apply ii

n

Denote by pol the image of the small p olylogarithmic extension pol under

n

H U Log j H U G

U

abs abs

Theorem a Under the isomorphism

B U Log j H H

U

abs abs

of ii the small p olylogarithmic extension pol is mapp ed to

b For each n the map

n

s H B H U G

n

abs abs

n

maps to pol

Documenta Mathematica

Annette Huber Jorg Wildeshaus

Pro of This is the denition of pol and the s

n

Recall a that we may identify

n n n sgn

H U G H G v F n

abs abs mU

n

sgn n n

n

H G v F n

V

mU

abs

n

n n sgn

H G rel Z n

mU

abs

In section we are going to prove a motivic analogue of ii and then dene

pol as the element in

n

n n sgn

lim H G rel Z n

mU

M

n

B mapping to under the isomorphism to H

M

In order to prove a motivic version of ii we shall frequently use injectivity

of the Beilinson regulator on certain motivic cohomology groups and two technical

results on H that will o ccupy the rest of this section

abs

While this may app ear articial at rst sight we remind the reader that in the motivic

setting we cannot make use of any sheaf theoretic means like Leray sp ectral sequences

An imp ortant means will b e the lo calization sequence asso ciated to the geometric

situation

f g A U

B

B

It is the result of the degeneration of the Leray sp ectral sequence and reads

H A p H U p H f g p

B

abs B abs

abs

H A p

B

abs

Lemma a The structure morphism is an isomorphism

H B p H A p

abs abs B

b The b oundary map is trivial ie we have short exact sequences

M

H B p H B p H U p

abs abs

abs

i

F p A F p Pro of For a note that R

B B

A

B

B

b follows from the fact that there are B valued p oints of U

In particular for p we have the exact sequence

M

B H U B H H

abs abs abs

i

The last map equals the map of Ext groups

Ext F F Hom F H U F

ShB

ShU B

Documenta Mathematica

Classical Motivic Polylogarithm

obtained from the Leray sp ectral sequence observe that the residues at and

B B

provide an isomorphism

M

F H U F

B

i

We have a natural map

O U H U

abs

Its comp osition with

M

H B H U

abs abs

i

asso ciates to a function on U its orders at and resp ectively

We need to understand the comp osition

res H U Ext F F

abs ShU

U G Hom F H

ShB

B

Observe that due to the last group is equal to H B Furthermore we recall

abs

from the pro of of and the denition of res that the comp osition

M

res

F H U F H U G F

B B

i

L

F We have thus proved is given by pro jection onto the comp onent of

i

Lemma Consider the nonvanishing functions t and t on U We have

res t res t

In particular the map

G rel Z H U H U G H

mU abs abs abs

do es not map t O U to zero

Pro of Observe that res factorizes through

Remark The main technical result of this section ii corresp onds to BD

ii Observe that pol and the p olylogarithmic class of lo c cit do not quite

agree in our notation

H U Log j

U

abs

while pol H U Log j The connection is as follows there is a canonical

U

abs

monomorphism

Log Log

identifying Log with W Log and pol is the push out of via The present

denition of the p olylog seems more natural since it is an element of an H B

abs

mo dule of rank one which is canonically trivialized By contrast H U Log j

U

abs

is of rank two

Documenta Mathematica

Annette Huber Jorg Wildeshaus

Calculations in K theory

The next step is to do the constructions of section with K groups or more precisely

with relative K cohomology as introduced in app endix B For technical reasons we

will have to use simplicial schemes to replace the singular schemes that app eared b e

fore All constructions will b e compatible with the regulator maps to absolute Ho dge

cohomology app endix A and B and to continuous etale cohomology app endix

B

A priori these regulators have values in absolute cohomology groups for the same

simplicial ob ject cf B and B Using B and B these absolute cohomol

ogy groups are then identied with relative cohomology of singular schemes This

identication is made tacitly

Let B Sp ecZ and S a smo oth ane B scheme We will work in the category

of smo oth S schemes K cohomology is taken on the Zariski site over B

Before returning to the geometric situation introduced in section we have to

check a technical lemma Let us consider the following general construction Let X

b e a smo oth quasipro jective S scheme and Y a closed subscheme of X which is itself

also smo oth over S Put

n

n n n

Y Y X q X Y X q q X Y

S S S

n

Note that Y is a prop er covering of the singular scheme

n n n

Y X r X r Y

This is the easiest case of a morphism of schemes with cohomological descent meaning

n

that for any reasonable cohomology theory the cohomology of Y will agree with

the cohomology of the smo oth simplicial scheme

n

n n

X Y cosk Y

ie

n n n

n n

k fold pro duct Y Y Y

X X

k

For etale cohomology and absolute Ho dge cohomology the corresp onding results are

B and B resp ectively

We will work in the setting of spaces ie p ointed simplicial sheaves of sets

on the Zariski site of smo oth B schemes We refer to app endix B for details and

terminology We use the notation

n n n

X ConeY X

for the space that computes relative cohomology for the closed embedding cf B

n

The space Y do es not b ecome degenerate ab ove any simplicial degree How

ever we have

n

Lemma a Y is isomorphic in Ho sT to a simplicial scheme which is degen

erate ab ove degree n

Documenta Mathematica

Classical Motivic Polylogarithm

n n

b In particular Y and X are K coherent

n n

c X is a space constructed from schemes in a nite diagram over X in the sense

of B

d If T is another closed subscheme of X which is smo oth over S and disjoint of Y

then the inclusions

i ni n

T X X

S

are torindep endent of all morphisms in the diagram in c

Pro of By denition

n

Y q q Y Y

n

n

where Y is the reduced closed subscheme of X of those p oints whose ith co ordinate

i

n

lies in Y This induces a decomp osition of Y into disjoint subschemes of the form

k

n n

Y Actually this subscheme is canonically isomorphic to Y

i X X i

k

n

Y for j k g Y fx x X j x Y

i n i i

j

k

We get the following more familiar form of the simplicial scheme

a

n

Y Y

i

k

k

iI

I fng

Let n b e the simplicial set with

n fi i j i i ng

k k k

n

We dene the simplicial scheme Y by

a

n

Y Y

i

k

iI

I n

k

It is degenerate ab ove the simplicial degree n and from our previous considerations

n

We consider these simplicial schemes as we see that it is a natural subspace of Y

spaces in the sense of app endix B by adding a disjoint base p oint

For a scheme U in the big Zariski site over B we consider the morphism of simplicial

sets

n n

U U Y Y

By the combinatorial Lemma B it induces an isomorphism of homotopy sets

Hence the inclusion is a weak homotopy equivalence of spaces

b is an immediate consequence of a and Bb Recall that Y and X were assumed

n

smo oth over B We already have seen that all comp onents of X are disjoint unions

n

Y and a disjoint base p oint All morphisms of X schemes of the form Y

i i

k

b etween the scheme comp onents are given by the natural closed immersions b etween

Documenta Mathematica

Annette Huber Jorg Wildeshaus

them The condition on the tordimension required in B follows b ecause are

schemes are regular T Y and X are all at over S hence the maps in the diagram

X Y

S

y

T X X X

S S

are easily seen to b e torindep endent The inclusions of T and Y into X are trivially

torindep endent b ecause this is a lo cal condition

Basically this lemma tells us that all conditions hold that are needed to apply

the machinery of app endix B We have a wellbehaved relative motivic cohomology

theory cf B

Now we return to the geometric situation set up in section We consider

n n

Z G

mS

y y

n

n

Z A

S

where Z Z S q S with disjoint S rational p oints and of G There

mS

n

is a simplicial op eration of S on the situation which induces an op eration on relative

K cohomology and on motivic cohomology

Proposition There is a natural residue map

res

n

i n

n sgn i n n sgn

H G rel Z j H G rel Z j

M mS

M mS

where sgn means the sign eigenspace under the op eration of the resp ective symmetric

group

Moreover there is a long exact sequence

n

n i

sgn n sgn i n

j Z rel Z j H A rel G H

M S

mS M

i n n sgn

H G rel Z j

M mS

n i

n sgn

rel Z G j H

mS M

Under the regulators the long exact sequences are compatible with the ones in abso

lute cohomology after

k k

Remark Recall that Z and hence H G rel Z j H S j by

M mS M

denition

n n

Pro of We lter A by the op en subschemes F A dened just b efore Lemma

k

S S

n n n n n

In particular F A G Again G A F A r F A We use the notation

k k k

S mS S S S

n n n

F A and G A for the induced op en resp ectively lo cally closed subspaces of A

k k

Note that the situation is still symmetric under p ermutation of co ordinates Hence

there is a compatible op eration of the symmetric group on the space constructed from

n

schemes F A

k

Documenta Mathematica

Classical Motivic Polylogarithm

n n

The closed immersion G A F A satises the rst condition in TC in B

k k

The maps we have to consider for the rest of TC are lo cally of the form considered

in d Hence B applies ie we can use the lo calization sequences for motivic

n n n

cohomology induced by the triples F A F A G A We get

k k k

i i

i n n n

H G A j H F A j H F A j

k k k

M

M M

i

n

H G A j

k

M

The sequence remains exact when we take signeigenspaces Now let us compute one

of the groups involved

M

i n i n

n

G a a j H A H G A j

mS k A k

M M

fa a a ng

k

where

G a a fx x j x if i a for some j x else g

mS k n i j i

n

The decomp osition corresp onds to the decomp osition of G A into its connected

k

n

n

G a a means the op en subspace lying comp onents The notation A

mS k A

over the lo cally closed scheme Now consider the op eration of the symmetric group

If k then there is for each comp onent some transp osition which acts trivially

namely one that interchanges two vanishing co ordinates Hence the signeigenspace

vanishes altogether For k the decomp osition has the form

M

a na

i n i n

n

H G A j G fg G j H A

A

M M

mS mS

a n

The op eration of the symmetric group p ermutes the factors transitively The stabilizer

n

of one summand is the symmetric group S We get

n

sgn i i n sgn

j H G H G A j

M M

mS

where the sign eigenspace on the right hand side is taken with resp ect to the smaller

n

symmetric group S We have a choice of isomorphism here and use the one that

n n

identies G with G fg Putting these results in the long exact sequences

mS mS

we get iteratively

n

sgn i n sgn i n sgn i n

Z j H F A j H F A j H A rel

n

M M M

So the ab ove sequence for k gives the desired residue sequence We can do the

same construction for absolute cohomology Ho dge or ladic considered as general

ized cohomology theories By B B and B the long exact sequences for

motivic cohomology will b e compatible via the regulator with the ones in generalized

cohomology The next step is to pass from generalized cohomology to cohomology of

ab elian sheaves By B and B this can b e done In fact we get precisely the

residue sequence for absolute cohomology constructed in section

Documenta Mathematica

Annette Huber Jorg Wildeshaus

Remark a By B we have the same maps and long exact sequences for the

K cohomology groups themselves However note that there is a RiemannRo ch hid

den in the compatibility of the lo calization sequence in K cohomology and absolute

cohomology

b We shall show injectivity of the Beilinson regulator on

n

n n sgn

H G rel Z n

mS

M

in Prop osition Together with Lemma b it shows that the residue map on

i n n sgn

H G rel Z j

M mS

n

in do es not dep end on the choice of embedding of G

mS

a na

G fg G

mS mS

a n

of the ab ove pro of if i j n n Since we are only interested in these sp ecial

indices we chose to exclude from the statement of the dep endence of res in the

n

general case from the ab ove choice

Lemma Let j k Then

n

k n

k n

S j H Z j H A rel

M S

M

where the isomorphism is induced by a choice of ordering of the sections and It

is compatible with the identication in under the regulator map S op erates by

n

sign on the left hand side

i

Remark Here and in the sequel we put H S j if j i This makes sense

M

as S is regular and the corresp onding K group vanishes see B

Pro of Fix j We consider the skeletal sp ectral sequence B We have

pq q

n

A E H j

p

S

M

We will show that the only nontrivial E terms are concentrated in one vertical line

nq q

S j E H

M

This means that the sp ectral sequence converges in the strongest p ossible way This

yields isomorphisms as stated Before we can check this we need some preparation

If X is a space constructed from schemes we denote by C pX the simplicial set of

n n

has the same singular cohomology as C p Z Z its connected comp onents C p

cf pro of of which is the simplicial set attached to a CWcomplex dual to the

Z has two disjoint comp onents b oundary of the ndimensional hypercub e note that

n

This means that C p Z has a vertex for every n cell of the cub e etc In

n

particular we see that it has the homotopy type of an n sphere C pA is of

S

n

course contractible It follows that C pA has singular cohomology concentrated

in degree n where it is onedimensional

Documenta Mathematica

Classical Motivic Polylogarithm

Let us make this more explicit

In order to compute the cohomology of a cosimplicial group it suces to consider

n

Z is completely the subcomplex corresp onding to nondegenerate simplices C p

degenerate from cosimplicial degree n on In degree n there is one nondegenerate

n

simplex for each vertex of the hypercub e They are indexed by f g Hence any

n

n n n

element of H C pA H C p Z is represented by an element of

M

n

Q K

n

f g

n

Z do es not b ecome degenerate Let g b e a generator of the cohomology group C p

n

The nondegenerate part in degree n is given by one copy of f g for each p ossible

sgn

p ermutation of the numbers n It is easy to see that g is in

n

Z the kernel of the dierential It represents the generator of cohomology of C p

We see that S op erates by the sign of the p ermutation

n

We choose the generator g of cohomology given by the tuple

si si

n

Q

i i

n

where i f g and s s This choice of generator amounts to

k

picking the ordering and extending it by the K unnethformula Now let us

analyze our E term For xed q we have the complex attached to the cosimplicial

q

n n

are isomorphic All connected comp onents of A ab elian group H A j

p pN

S

M

to a copy of some p ower of A By the homotopy prop erty of K theory we have

S

q q

n n

S j C H j A H

Q pN S

p

M M

n

where C is the cosimplicial vector space computing singular cohomology of

n

C pA By the previous considerations we already know its cohomology It also

follows that the op eration of S on our motivic cohomology is by the sign

n

Now compare our isomorphism to the one constructed in the realization We have the

same sp ectral sequence there attached to the weight ltration The identication of

the E term also uses K unnethformula and choice of an ordering of the sections

Using this identication we obtain the motivic residue sequence

k n k n

k n sgn sgn

H S j H G j H G j

M mS

M M mS

k n

H S j

M

for j k By construction we have the following

Theorem Under the regulator the motivic residue sequence maps to the abso

lute residue sequence of section

Note that the residue sequences for all indices k and n organize into a sp ectral

sequence connecting the relative motivic cohomology of A and the relative motivic

cohomology of G In particular for each n there is the converging cohomological

m

sp ectral sequence

pq pq pq pq n

n n n

G rel Z n G n H S p H E H

mS mS

M M M

Documenta Mathematica

Annette Huber Jorg Wildeshaus

This is the motivic version of the weight sp ectral sequence in absolute cohomology

We refer to it as the motivic residue sp ectral sequence

Remark As in section the residue sequence or equivalently the residue sp ectral

sequence turns out to b e the central technical to ol in the construction of the motivic

p olylog see Denition The sp ectral sequence is identical to the one constructed

in BD The denition and basic prop erties of motivic cohomology of simplicial

schemes B B allow to justify the construction

At this p oint we should stress that the pro of of the inno cent lo oking Theorem

requires the whole of the theory covered in the app endices

Universal Motivic Polylogarithm

We now return to the sp ecial situation used in section Let B Sp ecZ We

consider now the case S U Let and the diagonal section of U G

B mB

First we compute the motivic cohomology of U We use the embedding of U into

A to do so The long exact lo calization sequence B reads

n

n n

H B q B j H A j H U j

M B M

M

n

H B q B j

M

By the homotopy prop erty of K theory we get

n n

n n

H B j H U j H B j H B j

M M

M M

n

B j H

M

The Gysin map for the inclusion of a p oint in the ane line vanishes by Q Thm

ii Hence we are actually dealing with a system of short exact sequences As all

motivic cohomology groups of B vanish for n this sequence only gives nontrivial

cohomology of U for n

Lemma For B Sp ec Z we have

Q if i

H U i

M

else

for j

H U j

Q Q for j

M

H B j for j

M

H U j H B j H B j

M M M

n

H U j if n

M

Pro of Clear from the ab ove using B

By Borels Theorem B the Beilinson regulator

i i

X j R H H X R j

p

Q R

H M

is injective for X Sp ecZ even an isomorphism but in the one case H B

M

where the co dimension is one We call Beilinson regulator what strictly sp eaking is

Documenta Mathematica

Classical Motivic Polylogarithm

i

U k its tensor pro duct with R This implies that it is also an isomorphism for H

M

with the exception of the indices and where the co dimension is resp

This means that many of the residue maps are actually isomorphisms The

following computations are carried out in the case B Sp ecZ With a little more

eort they generalize to the case of the ring of integers of a number eld

Consider the residue sequence for n j and S U

H U H G H U

M M mU M

H U H G H U

M M mU M

The Beilinson regulator induces a map b etween the ab ove sequence and the residue

sequence in section On H U R the regulator is an isomorphism and on

M

H U R it is injective of co dimension one By the absolute Ho dge coho

M

mology group H G R vanishes Hence the map from the rst to the second

p

H mU

R

line is injective and the regulator is injective of co dimension one on H G

M mU

Furthermore this last group is one dimensional

The image of under the Beilinson regulator is the map o ccurring in for

n

Definition Let s b e the comp osition of the maps

M

i

H B H U H G Q H B

M M M mU M

i

where i is the inclusion of the summand and is the map of the residue sequence

Lemma s is an isomorphism

Pro of Because of dimension reasons we only have to check that do es not vanish on

the image of i This follows from

Definition Let res b e the inverse of s We dene the total residue map

n

n sgn

res H G n Q

mU

M

by comp osition of the residue maps in our long exact sequence with res

We now have to check that the total residue map deserves its name By denition

and i it suces to consider res

Lemma The regulators map the motivic res to res in absolute cohomology

Pro of Let us consider the situation of The morphism

O U H U R

p

R

H

factors through H U K U There is a commutative diagram

Q

M

L

H H B R U R

p

R R

p

H H

H

i

x x

L

H B O U H U

M M

i

Documenta Mathematica

Annette Huber Jorg Wildeshaus

hence the functions t and t on U corresp ond to the canonical generators of the two

summands We consider the commutative diagram for absolute Ho dge cohomology

res

R R

H U R H G R H B R

p p p

R R

H H mU H

R

x x

r

L

G B H H

mU M M

i

By the comp osition from the b ottom left to the top right corner is given by the

pro jection to the comp onent tensored by R It follows that res r R is an

R

isomorphism In turn vanishes on the comp onent and is an isomorphism on the

comp onent But then by denition res is also the pro jection to the summand

As is surjective this suces The same argument works in the etale situation

Lemma There is a short exact sequence

r es

sgn

Q H B H G

M M mU

and the Beilinson regulator is an isomorphism on the middle term

Pro of This is nothing but the residue sequence using our computation of

H G The zero es on b oth sides come from vanishing cohomology groups

M mU

Comparison with the short exact sequence ii shows that the regulator is an

isomorphism

Proposition There are short exact sequences

res

n

n

n sgn

Q G n H H B n

mU M

M

n

sgn n

n It is even an isomorphism The Beilinson regulator is injective on all H G

mU

M

for n

Pro of The n and n cases are the previous lemmas By induction one checks

n n sgn

that all H G n vanish for n Hence the residue sequence reads

M mU

n

n n

n sgn n sgn

G H n H n H G U n H B n

mU M M M

M mU

By the ve lemma and inductive hypothesis we see that the regulator is an isomor

phism on the middle term for n We need the previous lemma to get started

Now consider the sequences of the prop osition All maps are welldened It follows

from ii that the sequence is exact

n

n sgn

Corollary There are canonical splittings s Q H G n such

n

mU

M

that the diagram

s

n

n

n sgn

H B H G n

M mU

M

s res

n

n

n sgn

H G n

M

mU

commutes They are compatible with the ones in iii Furthermore the group

n

sgn n

n is canonically isomorphic to Q G lim H

mU

M

Documenta Mathematica

Classical Motivic Polylogarithm

n

sgn n

n This Pro of Imr es is isomorphic to Q by the total residue on H G

n

mU

M

induces the same splitting as in

Definition For n N the system pol s denes the universal motivic

n n

p olylogarithm

By construction pol is mapp ed to the p olylogarithmic system in absolute Ho dge

n

cohomology and continuous etale cohomology

Remark The main result of this section is identical to BD Although

part of the argument involves only constructions within K theory the pro of of

relies heavily on a detailed analysis of the b ehaviour of the regulator b etween the

motivic and absolute residue sequences

The Cyclotomic Case

Let d As b efore let R Ad T T the ring of dintegers of the cyclotomic

d

eld of dth ro ots of unity Put C Sp ec R Let b e a primitive dth ro ot of unity

Q and b an integer prime to d We work in the situation S C G C in

m

and i G C as in section

b m

Lemma a For n we have

n n sgn n n n sgn

H G n H G rel Z n Q

M mC M mC

The Beilinson and the ladic regulators are isomorphisms

b For n the residue sequence induces short exact sequences

n n

n sgn n sgn

H C n H G n H G n

M mC M

M mC

n

n sgn

G n for n The ladic regulator is injective on the group H

mC

M

Pro of For n we have H G H C which is canonically isomorphic

M mC M

to Q by B In particular b oth regulator are isomorphisms

H G and its counterpart in absolute cohomology vanish

M mC

Consider the following bit of the residue sequence for n

n n

sgn n n sgn

n H G G H n H C n

M mC M

mC M

The rst map is injective since H C n The ladic regulator is always

M

injective on the last term by B By inductive hypothesis it is an isomorphism on

the middle term By Cor the last map vanishes in absolute cohomology This

implies a for n In the next bit of the long exact sequence

n n

n sgn n sgn

H C n H G n H G n

M mC M

M mC

the rst map is injective by a For n we have H C n while for

M

n the term

n

n

H G n H C

M M

mC

vanishes Hence in any case we end up with the short exact sequence in b The

regulator maps it to the short exact sequence By induction and B we can

control the injectivity of the l adic regulator

Documenta Mathematica

Annette Huber Jorg Wildeshaus

C b ecause d is inverted Remark The Beilinson regulator is not injective on H

M

in C

Consider the morphism G G that raises p oints to the d th p ower

mC mC

n n n

By contravariance A As in section it induces a morphism of spaces A

C C

it induces an op eration on motivic cohomology

Lemma BD Remark i i on page

n

op erates on the short exact sequence of the previous lemma as follows

n n

n sgn n sgn

H C n H G n H G n

M mC M

M mC

n 

n 

d

id

y y y

n n

n sgn n sgn

H C n H G n H G n

M mC M

M mC

Pro of This description follows immediately from the injectivity of the ladic regulator

and Cor b

n

Remark The op eration on H C n is given by the op eration on

M

n

n

H A n It is easy to check that it is trivial by considering the op eration

C

M

on the starting terms of the degenerating skeletal sp ectral sequence To understand

the compatibility with the residue map in terms of K theory is a lot harder The

factor d is induced by a pushforward from a nonreduced scheme to its reduction

The theory in App endix B is not even set up to handle such schemes

n

As in the case of absolute cohomology it follows that the eigenvalues of on

n

n sgn n

G rel Z n are d d H

mC

M

Lemma The eigenspace decomp osition yields a splitting

M

n

n

n sgn

H G n H C i

mC M

M

b

in

n

which is compatible with the splitting after Cor There is a canonical

b

isomorphism

Y

n

n sgn

lim H G n H C i

b

mC M

M

i

Pro of The rst assertion is clear by construction The second follows b ecause the

eigenspace decomp osition is compatible with the residue map

Definition Let i C U b e as b efore Let pol b e the pullback of the univer

b b

n

n sgn

sal p olylogarithm system pol dened in to the inverse limit lim H G n

mC

M

n

n n sgn

lim H G rel Z n Via the isomorphism of we have constructed

b

mC

M

Q

H C i an element in

M

i

Theorem Under the regulators the element

Y

n

n sgn

pol lim H G n H C i

b

mC M

M

i

Documenta Mathematica

Classical Motivic Polylogarithm

is mapp ed to the elements

Y

n

n sgn

pol lim H G n H C i

b

mC abs

abs

i

constructed at the end of section

Pro of This follows from the construction

We list the consequences of this result denote by the set of primitive dth ro ots

d

of unity in Q

d

Firstly the description of the regulator to absolute Ho dge cohomology yields an

alternative pro of of the following

Corollary Assume n

a B Neu I I E

There is a map of sets

H C n

n

d M

H Sp ec Q n for n

d

M

such that

r H Sp ec Q R n

p

D n d R

d H

M

n

A

C i R

A

Q C

d

maps a ro ot of unity to Li For n this prop erty characterizes the

n

map uniquely

n

b

b For a ro ot of unity T Q Q T T the element

d d

b

T H C n

n

M

is given by

b n

T n comp onent of pol

n b

n

Pro of Note that a really is Beilinsons formulation of the result his normalization

of the isomorphism

M

n

H Sp ec Q R n C i R

p

d R

H

diers from ours by the factor The unicity assertion is a direct consequence of

the injectivity of the regulator So our claim follows from and from

Documenta Mathematica

Annette Huber Jorg Wildeshaus

In B the ab ove compatibility statement is used to prove Grosss conjecture

ab out sp ecial values of Dirichlet Lfunctions An alternative pro of of this conjecture

using an entirely dierent geometric construction is given in section of Den

Recall that the l adic regulator r factorizes as follows

l

K C Q H C n H C n

n Z

l

M M

H C n

l

cont

H Sp ec Q n

d

cont

where we let C C Z

Z

l

l

For the rest of this section we x C Q As was observed already in B

the study of the cyclotomic p olylog yields a pro of of the following result

Corollary Assume n

a Sou Theoreme for the case n Gr Theoreme IV for the lo cal version

if l d

Let d and b e as in Let l b e a prime Under the embedding of the l adic

n

regulator

r H C n H Sp ec Q n

l d

M cont

b

maps T to

n

X

d n

A

n

d n

r

l b

r

b Conjecture of BlK holds

Pro of a is and As for b it remains to check the comparison statement of

BlK Conjecture for the ro ot of unity For this observe the relations

n

c c

n n

n

n

c c

n n

n

in the notation of lo c cit if n D Prop osition i

Soule has constructed maps

K C Z

l n Z l

l

d

for any prime l see end of App endix B for more details

The l adic regulator

r K C Q H Sp ec Q n Prop B

l n Z l d

l

cont

b

takes T to the cyclotomic element in continuous Galois cohomology

l

X

d n

A

r

l b

r

dened by Soule and Deligne cf Sou page D

Documenta Mathematica

Classical Motivic Polylogarithm

Corollary For each d and n there is a unique map

K Sp ec Q

n d

d

such that for each prime number l the map

K C Z

l n Z l

l

d

K Sp ec Q Z

n d Z l

equals the comp osition of and the natural map

K Sp ec Q K Sp ec Q Z

n d n d Z l

Furthermore the map Q agrees with

Z

H Sp ec Q n

d

n d M

n

given by d n

n

Pro of The uniqueness assertion is a formal consequence of the nite generation of

K Sp ec Q to give an element in a nitely generated ab elian group M is the

n d

same as giving elements in M Q and all M Z which coincide in M Q By

Z Z l Z l

the maps r and r agree for all l From Theorem B we conclude

l l l

n

that and agree as maps to K Q

l n Z l

n

As shown by Blo ch and Kato Corollary implies the validity of the following

also for even n

Corollary Let n

Then the Tamagawa number conjecture BlK Conjecture is true mo dulo a

p ower of for the motif Q n

Pro of BlK Theorem i gives the complete pro of for o dd n which is indep endent

of anything said in the present article In lo ccit Theorem ii it is shown that

the conjecture holds for even n if BlK holds But the latter is the content of

Finally the compatibility statement of forms a central ingredient in the

pro of of the mo died version of the Lichtenbaum conjecture for ab elian number elds

KNF Theorem

A Absolute Hodge Cohomology with Coefficients

The aim of this app endix is to provide a natural interpretation of absolute Ho dge

cohomology as extension groups in the category of algebraic Ho dge mo dules over R

A That such a sheaftheoretic interpretation should b e p ossible was already

anticipated by Beilinson B long b efore Ho dge mo dules were dened

The app endix is divided into two subsections The rst A starts with a

summary of those parts of Saitos theory relevant to us The central result is A

where we prove that for a smo oth scheme a U Sp ecC the p olarizable Ho dge

Documenta Mathematica

Annette Huber Jorg Wildeshaus

complex R U F of D and B x is a representative for a F the

U

ob ject in the derived category of p olarizable F Ho dge structures dened via Saitos

formalism S As a consequence we are able A to identify absolute

Ho dge cohomology of a smo oth scheme U over C as dened in B x it equals

the Ext groups of Tate twists in the category of algebraic Ho dge mo dules on U The

compatibility b etween the approaches of DeligneBeilinson and of Saito will come as

no surprise to the exp erts see eg S However we were unable to nd a

quotable reference

In A we turn to the variant of the theory we really need algebraic Ho dge

mo dules over R These live on the complexication of separated reduced schemes of

nite type over R and are basically the ob jects xed by the natural involution on

the category of mixed Ho dge mo dules given by complex conjugation The compari

son statement for absolute Ho dge cohomology over R Theorem A then follows

formally from A

A Algebraic Mixed Hodge Modules

In S x the category MHM X of algebraic mixed AHo dge mo dules is dened

A

where A is a eld contained in R and X a separated reduced scheme of nite type

over C

Saitos construction admits the full formalism of Grothendiecks functors

b

D on the level of b ounded derived categories D MHM S Hom

A

and a forgetful functor

rat MHM X Perv X

A A

to the category of p erverse sheaves on the top ological space X underlying X C which

have algebraic stratications such that the restrictions of their cohomology sheaves

to the strata are lo cal systems By the denition of MHM which we shall partly

A

sketch in a moment rat is faithful and exact The functor rat on the level of derived

categories is compatible with Grothendiecks functors S

For smo oth X one constructs MHM X as an ab elian sub category S Prop o

A

sition of the category MF W D A whose ob jects are

h X

M F W K W

where M F is an ob ject of the category MF D ie a regular holonomic alge

h X

X W is a braic D mo dule M together with a go o d ltration F and K Perv

X A

lo cally nite ascending ltration and is an isomorphism

DR M K C

A

resp ecting W Here DR denotes the de Rham functor from the category of D

X

mo dules to the category of p erverse sheaves

We note that by denition the weight graded ob jects of all algebraic Ho dge

mo dules satisfy a certain p olarizability condition see S

Call an algebraic Ho dge mo dule on a smo oth variety smo oth if the underlying

p erverse sheaf is a lo cal system up to a shift

Documenta Mathematica

Classical Motivic Polylogarithm

Theorem A Saito Let X b e smo oth and separated Then there is an equiv

alence

s

Var X MHM X

A A

b etween the category of admissible variations of mixed AHo dge structure Ks and

the category of smo oth algebraic AHo dge mo dules on X

Pro of This is the remark following S Theorem

In particular we see that MHM Sp ecC is the category MHS of p olarizable

A A

mixed AHo dge structures

If V is a variation on X with underlying lo cal system ForV then the p erverse

sheaf underlying the Ho dge mo dule V under the corresp ondence of A is

ForVd

if X is of pure dimension d

It turns out that the denition of Tate twists in MHM X is compatible with

A

the ab ove equivalence only up to shift

Definition A S Let n Z and An MHS the usual Tate

A

twist For a separated reduced scheme a X Sp ecC dene

b

An a An D MHM X

X A

If X is smo oth and of pure dimension d then An d is the variation of Ho dge

X

structure which one denotes An

For arbitrary X the complex An will not even b e the shift of a Ho dge mo dule

X

b p

but a prop er element of D MHM X whose cohomology ob jects H An are a

A X

priori trivial only for p dim X S

We note again that we follow Saitos convention and write eg for the functor

on derived categories

b b

D MHM X D MHM Y

A A

induced by a morphism X Y

In order to compare the Ho dge structures on Betti cohomology given by Saitos

and Delignes constructions we need to go into the details of S

Theorem A Saito Let j U X b e an op en immersion of smo oth sepa

rated schemes over C with Y X n U a divisor with normal crossings If X is of

pure dimension d then

d

j A d H j A MHM X MF W D A

U U A h X

equals the ob ject

w Y j A d

X top U

d

where w Y denotes the D mo dule log Y and j the direct image for

X X top

X

the derived category of p erverse sheaves

Documenta Mathematica

Annette Huber Jorg Wildeshaus

The de Rham complex with logarithmic singularities is quasiisomorphic to

L

O d DR w Y d hence w Y

X X X D

X

DR w Y log Y d

X

X

compare Bo VI I I and

log Y d j C d

top

X

is the usual quasiisomorphism

log Y j j C

top top

X U

compare D shifted by d

The Ho dge ltration F on w Y is induced from the stupid ltration while

X

the weight ltrations W on w Y and j C d are those induced from the

X top

canonical ltration on j shifted by d

top U

d

Pro of The equation j A d H j A follows from the faithfulness of rat

U U

and the fact that the corresp onding statement for j is true since j is ane In

top

our geometric situation the explicit construction of j of any admissible variation of

AHo dge structure is carried out in the pro of of S Theorem For A it

U

sp ecializes to our claim

In B Beilinson extends Delignes notion of Ho dge complexes D

to the p olarizable situation

Definition A Beilinson A mixed AHo dge complex

K K F W K W

C

is called p olarizable if the cohomology ob jects of the weight n Ho dge complexes

W

Gr K are p olarizable AHo dge structures

n

Remark The weight ltration W of a mixed Ho dge complex K induces mixed

W

i

Ho dge structures on its cohomology Observe however that Gr H K is of weight

n

n i

As in the nonp olarizable situation Beilinson proves

Theorem A B Lemma There is an equivalence of categories b e

b

tween D MHS and the derived category of p olarizable AHo dge complexes

A

Let X b e smo oth and separated over C Forgetting part of the structure of a

Ho dge mo dule yields a functor

b

For C MHM X T X

A

Here T X is the category of triples

M M F W K W

Documenta Mathematica

Classical Motivic Polylogarithm

b

where M F W is a class in the ltered derived category D W MF D of

h X

MF D and K W a class in the ltered derived category of sheaves of A

h X

b

vector spaces on X C denoted by D W X C A Furthermore the map is an

isomorphism

DR M K C

A

resp ecting W

b

Recall that in order to obtain a class in D W X C A from a complex of p erverse

sheaves one applies the realization functor of BBD

b

The global section functor can b e derived on D W X C A By S we

b

have a functor R on D W MF D if X is prop er and the two constructions are

h X

compatible with the comparison isomorphism of any ob ject in T X S

We end up with an ob ject

R M R M F W R K W R

of T Sp ecC The functor

b

R R For C MHM X T Sp ecC

A

b

factorizes through D MHM X

A

Our second comparison result is the following

Theorem A Let a X Sp ecC b e smo oth and prop er and M an ob ject of

b

D MHM X Write

A

For M M F W K W T X

a

M R M F W R K W R R

is a mixed p olarizable AHo dge complex

b The class of R M in the derived category of p olarizable Ho dge complexes is

canonically isomorphic under the identication of A to

b

a M D MHS

A

c Let f Y X b e a prop er morphism of smo oth and prop er schemes over C

and let b denote the structure morphism of Y such that

b af

b

For any N D MHM Y together with a morphism M f N in

A

b

D MHM X the morphism

A

M R N b N a f N a a M R

equals under the isomorphism of a the morphism

R R R

of AHo dge complexes

Documenta Mathematica

Annette Huber Jorg Wildeshaus

Pro of a We may assume that M is pure of some weight Using S we are

reduced to the case where M M is a Ho dge mo dule of weight n and we have to

M is a p olarizable Ho dge complex of the same weight Axiom CH show that R

of D follows from S Prop osition in particular applied to

pr M where

pr X A X

C

C

Furthermore by the remark following S and by lo c cit we have

isomorphisms in MF W D A

h

Sp ecC

i i i i i

M R M F W i R K W i R H a M R

Since the right hand side is a p olarizable Ho dge structure of weight i n S

we have CH and in addition p olarizability

b In the pro of of a we constructed a functor

b b

a R D MHM X D MHS

A A

such that

i i b

H a H a MHM X D MHM X MHS

A A A

b b

for all i Comp osition with j D MHM U D MHM X for op en immersions

A A

j U X denes

b b

aj a j D MHM U D MHS

A A

But for ane U aj is the left derived functor of

H aj MHM U MHS

A A

S pro of of Theorem If U is ane then so is j U X and hence j is

exact Therefore

j

H a





H aj H a j MHM U MHS MHM X

A A A

and we get a natural transformation coincides with H aj

aj aj

which is an isomorphism since this is true on the level of cohomology ob jects as

one checks on the level of vector spaces Observe that this natural transformation is

compatible with restriction to smaller ane subschemes of X Now recall S pro of

of that the functor a is constructed using the Cech complex asso ciated to an

ane covering of X for details see B In the same way the functor a is

recoverable from the aj We end up with an isomorphism of a and a which is

indep endent of the covering

c In the pro of of b we constructed a natural isomorphism

a a

Documenta Mathematica

Classical Motivic Polylogarithm

b b

of functors from D MHM X to D MHS For f id our claim is therefore

A A

proved For the general situation we use the same techniques as in the pro of of b to

rst construct a natural isomorphism

a f b

b b

of functors from D MHM Y to D MHS and then to see that the triangle

A A

b a f

y

a f

commutes

Corollary A cf S Let j U X b e a smo oth compactication

of a smo oth and separated scheme a U Sp ecC such that Y X n U is a divisor

with normal crossings

b

a a A D MHS is isomorphic under the identication of A to the

U A

class of the mixed p olarizable AHo dge complex

U A R DR log Y j A R

top U

X

of D and B x with the same notation

b If f X X is a morphism of compactications j U X and j U X

of U as in a then f induces an isomorphism

DR DR log Y j A log Y j A R R

0

top U U

X X top

D remark preceding so R U A dep ends only on U

The isomorphism in a also dep ends only on U

c In particular the Ho dge structures on

n n n

ratH a An H U C i A

U

B

given by Delignes and Saitos constructions coincide

Pro of a Combine A and Ab

b Use Ac

c follows from a and b

Actually the statement Ac implies the functoriality prop erty we were after

we have two functors

b

S mC D MHS

A

where S mC denotes the category of smo oth separated schemes over C

A U R U A R

A a U Sp ecC a A

U

Documenta Mathematica

Annette Huber Jorg Wildeshaus

Corollary A The isomorphism of Aa is functorial in U S mC In

other words there is a natural isomorphism

A R A

b

of functors from S mC to D MHS

A

Pro of Let

0

j

U X

f f

y y

j

U X

b e a commutative diagram of smo oth and separated schemes over C where X and

X are prop er and Y X n U and Y X n U are divisors with normal crossings

We have a morphism

0

j A f j A

U U

Application of a gives the morphism

X

0 0

A a A a

U U U U

b elonging to the functoriality requirement for A Our claim follows from Ac

applied to a shift of the morphism

Definition A Let XC b e separated reduced and of nite type and M an

b

ob ject of D MHM X

A

a The absolute Ho dge complex of X with co ecients in M is

p

R X M R Hom b A M

H X

D MHM X

A

b Its cohomology groups

i i

p

H X M X M H R

p

H

H

are called absolute Ho dge cohomology groups of X with co ecients in M

c We denote absolute Ho dge cohomology with co ecients in Tate twists by

i i

H X n H X An

p p

X

H H

d For a closed reduced subscheme Z of X with complement j U X we dene

relative absolute Ho dge cohomology with co ecients in Tate twists as

i

p

H X j An X rel Z n H

p

U H

H

Documenta Mathematica

Classical Motivic Polylogarithm

Note that if X is smo oth and of pure dimension d and if

M M MHM X

A

then the right hand side of Ab b eing equal to

Hom b A d M d i

X

D MHM X

A

admits an interpretation as the group of d iextensions of Ho dge mo dules mo dulo

Yoneda equivalence

Corollary A If X is smo oth and separated over C and n Z then

p p

X An and H X n H X n R X An R

p p

X H X H

H H

coincide functorially with the same noted ob jects of B x

Pro of This follows from A and the adjunction formula

A a M R Hom b A M R Hom b

X

D MHM X D MHS

A A

Remark The Leray sp ectral sequence for a X Sp ec C yields exact sequences

i i i

A H Ext A H H X An Hom

p

X MHS

MHS A H

A

k k n

with H H X C i A since MHS has cohomological dimension one B

A

B

i

Corollary Comparing them with the analogous sequences for H we see that

H

i i

H X An H X An

p

X X

H H

i

n

in the notation of B x if H X C i A has weights smaller than zero

B

which is the case if i n i n if X is prop er

Observe that this is the same range of indices where Deligne cohomology coincides

i

X Rn N we have natural morphisms with H

X

H

i i i

H X R n H X Rn H X Rn

p

X X X

H H D

b oth of which are isomorphisms if i n i n if X is prop er

A Algebraic Mixed Hodge Modules over R

Algebraic Ho dge mo dules over R are dened as the category of Ho dge mo dules xed

under a certain involution given by complex conjugation We start by constructing

this involution

Let XC b e smo oth and let X denote the complex conjugate scheme We have

an equivalence

Var X Var X

A A

Documenta Mathematica

Annette Huber Jorg Wildeshaus

of the categories of admissible variations induced by complex conjugation

X C X C

and dened as follows

The lo cal system and the weight ltration on X C are the pullbacks via of the

lo cal system and the weight ltration on X C and the Ho dge ltration on X C is

the pullback of the conjugate of the Ho dge ltration on X C

preserves admissibility and b ehaves in an obvious sense involutively

In particular if X is dened over R we get an involution on Var X C

A R

Definition A Let XR b e smo oth and separated

a The category Var XR consists of pairs V F where V is an ob ject of

A

Var X C and F is an isomorphism

A R

V V

of variations such that F F

In the category Var XR we may dene Tate twists An F acts via

A

n

multiplication by

b Var XR the category of admissible variations of mixed AHo dge structure

A

over R is the full sub category of Var XR of pairs V F which are graded

A

p olarizable for n Z there is a morphism

W W

Gr V F Gr V F An

A

n n

in Var XR such that the induced morphism

A

W W

Gr V Gr V An

A

n n

is a p olarization in the usual sense

Remark We note that implicit in our denition is a descent datum over R of the

biltered at vector bundle on X C underlying any admissible variation V F

R

of mixed AHo dge structure over R

For this claim to make sense recall rst D I I Theoreme that any at

analytic vector bundle on X C carries a canonical algebraic structure If the vector

bundle underlies an admissible variation then the Ho dge ltration is a ltration by

algebraic subbundles Ks Prop osition

Now the descent datum is given by the antilinear isomorphism

c F F c c F F F V F V

DR di di di di

of the C bundles underlying V and V Here c denotes the antilinear involutions

given by complex conjugation of co ecients and F is the forgetful functor to C

di

bundles

Lemma A The category Var Sp ec RR equals the category MHS of mixed

A

A

p olarizable AHo dge structures over R B x

Documenta Mathematica

Classical Motivic Polylogarithm

Pro of Straightforward

Our aim is to generalize our denition of sheaves over R to algebraic Ho dge

mo dules

For smo oth and separated XC recall that MHM X is an ab elian sub category

A

of MF W D A Ob jects of the latter are

h X

M F W K W

where M F is an ob ject of the category MF D of regular holonomic algebraic

h X

X W is a lo cally nite ascend D mo dules with a go o d ltration and K Perv

X A

ing ltration and is an isomorphism

DR M K C

A

resp ecting W

The equivalence

MF W D A MF W D A

h X h X

is constructed comp onentwise

The p erverse sheaf and the weight ltration on X C are the pullbacks via

X C X C of the p erverse sheaf and the weight ltration on X C

The equivalence

Mo d Mo d

D D

X

X

which by construction will resp ect holonomicity comes ab out as follows

Given a D mo dule N we may form the inverse image in the sense of sheaves

X

of ab elian groups N which is a D mo dule All we therefore need is an

X

D of sheaves of rings extending the isomorphism isomorphism c D

X X

O given by complex conjugation of co ecients we then dene c O

X X

D N N

X

D

X

Of course the map c is itself given by conjugation of co ecients in lo cal co ordinates

x x we have

n

X X

c f f c

x x

Altogether we get

MF W D A MF W D A

h X h X

which again b ehaves involutively

Going through the denition one checks that induces

MHM X MHM X

A A

Using lo cal embeddings as in S we can dene for any scheme X which is

separated reduced and of nite type over C Furthermore if X is dened over R we

get an involution on MHM X C

A R

Documenta Mathematica

Annette Huber Jorg Wildeshaus

Theorem A Let X and Y b e separated and reduced schemes of nite type over

C

b

a is compatible with Hom and D eg for M N D MHM X we

A

have

M N M N Hom Hom

X X

b If X Y is a morphism then is compatible with eg for

b

M D MHM X we have

A

b

M M D MHM Y

A

Pro of This follows from the denitions

Definition A a Let a X Sp ecR b e smo oth and separated The

category MHM XR consists of pairs M F where M is an ob ject of

A

MHM X C and F is an isomorphism

A R

M M

such that F F

By Ab we have a An MHM XR

A

b Let a X Sp ec R b e smo oth and separated MHM XR the category

A

of algebraic mixed AHo dge mo dules over R on X is the full sub category of

MHM XR of pairs M F which are gradedp olarizable for any n Z

A

there is a morphism

W W

Gr M F Gr M F a An

A

n n

in MHM XR such that the induced morphism

A

W W

Gr M Gr M a An

A

n n

is a p olarization in the sense of S

As in A we identify the category of smo oth ob jects in MHM XR with

A

Var XR

A

c For an arbitrary separated and reduced scheme X of nite type over R one

denes the category MHM XR using lo cal embeddings as in S

A

Remark a As in the case of variations over R we get a descent datum over R for

the biltered D mo dule underlying any Ho dge mo dule over R on a smo oth and

X C

R

separated scheme X over R

b As in S the category MHM Z R for any closed reduced subscheme

A

Z of X is equivalent to the category of Ho dge mo dules over R on X with supp ort in

Z

Documenta Mathematica

Classical Motivic Polylogarithm

Theorem A There is a formalism of Grothendiecks functors

b

Hom D on D MHM R It is compatible with the forgetful functor

A

b b

D MHM R D MHM C

A A R

Pro of By A we may eg dene

M F M F

Definition A Let XR b e separated reduced and of nite type and M an

b

ob ject of D MHM XR

A

a The absolute Ho dge complex of XR with co ecients in M is

p

XR M R Hom b A M R

X H

D MHM XR

A

b Its cohomology groups

i i

p

H XR M XR M H R

p

H

H

are called absolute Ho dge cohomology groups of XR with co ecients in M

c We denote absolute Ho dge cohomology with co ecients in Tate twists by

i i

H XR n H XR An

p p

X

H H

d For a closed reduced subscheme Z of X with complement j U X we dene

relative absolute Ho dge cohomology with co ecients in Tate twists as

i

p

H XR j An X rel Z R n H

p

U H

H

Again if X is smo oth and of pure dimension d and M M MHM X we have

A

di

i

H XR M Ext A d M

p

X

H

MHM XR

A

We have statements analogous to AA for the situation over R For

reference we note explicitly

Theorem A If X is smo oth and separated over R and n Z then

p

XR n and H R XR n

p

H

H

coincide functorially with the absolute Ho dge complex and cohomology groups of B

x

Next we have

Documenta Mathematica

Annette Huber Jorg Wildeshaus

Lemma A Let XR b e separated reduced and of nite type and M an ob ject

b

of D MHM XR Then the forgetful functor

A

b b

D MHM XR D MHM X C

A A R

induces functorial isomorphisms

p p

X C M XR M R R

R H H

H XR M H X C M

p p

R

H H

Here the sup erscript denotes the xed part of the action of the involution on

R Hom b A M

X C

D MHM X C

R

A

R

In particular the category MHS has cohomological dimension one since this is

A

p

X true for MHS Furthermore observe that the ab ove action of ZZ on R

R A H

C An is precisely that of B x

X C

R

Corollary A Let XR b e separated reduced and of nite type The forgetful

functor

X C rat MHM XR Perv

R A A

is faithful and exact

Remark Again we have

i i

H XR An H XR An

p

X X

H H

if i n i n if X is prop er We have natural morphisms

i i i

H XR R n H XR R n H XR R n

p

X X X

H H D

which are isomorphisms in the same range of indices

We conclude with an explicit formula for Ext in MHM XR of a nite scheme

A

XR

there is a canonical isomorphism Theorem A For any H MHS

A

A H Ext W H W H W F H

C A C

MHS

A

H Sp ecR R H

p

H

where the sup erscript on the left hand side denotes the xed part of the de Rham

conjugation

c

1

F H W H W H W W H W H W F H

C C A C A C

W H W H W F H

C A C

F

1

W H W H W F H

C A C

Documenta Mathematica

Classical Motivic Polylogarithm

The isomorphism is given by sending the class of h W H to the extension describ ed

C

by the matrix

h id

H

This means that we equip C H with the diagonal weight and Ho dge ltrations and

C

the Arational structure extending the Arational structure H of H by the vector

A C

h C H

C

thereby obtaining an extension E of A by H in the category MHS

A

A H is given with the same The conjugate extension E Ext

MHS

A

notation by the matrix



F h id

H

and the extension of F to an isomorphism

F E E

sends h to F h Thus

F id F C H C H

C C C C

Pro of Using B x or Jn Lemma and Remark a we see that there is

an isomorphism

A H W H W H W F H Ext

C A C

MHS

A

Note that our normalization follows that of Jannsen and therefore diers from that

of Beilinson by the factor

In general if h W H corresp onds to an extension E in MHS then c h W H

C A C

corresp onds to E and its pullback via

F H H

A H therefore is describ ed by F c h The action of the involution on Ext

MHS

A

corresp onds to F c on the left hand side of the ab ove isomorphism

Corollary A Let XR b e nite and reduced and M MHM XR Then

A

there is a canonical isomorphism

M

A

W M W M W F M

xC xA xC

xX C

M

A

M A Ext

x

MHS

A

A

xX C

H XR M

p

H

Documenta Mathematica

Annette Huber Jorg Wildeshaus

Pro of The last isomorphism is given by the observation that we have

M

MHS MHM X

A A

xX C

Corollary A For XR nite and reduced and n we have

M

n

A

C i A Ext A An

X X

MHM XR

A

xX C

H XR n

p

H

Here the sup erscript denotes the xed part with resp ect to the conjugation on b oth

n

X C and C i A and the isomorphism asso ciates to z the extension

x

xX C

whose stalk at x X C is given by the matrix

z

n x

i

n n

if e and e are the base vectors F C and i i A C then the

n

Ho dge structure is sp ecied by

F he i W C he i

C n A n C

and the Arational structure is generated by e and

n

z e e

x n

n

i

Pro of This is A and A using the basis e of An

n

B K Theory of Simplicial Schemes and Regulators

We start with a presentation of K theory B for simplicial schemes in terms

of generalized cohomology Applied to a regular scheme we get back its K groups

cf Ba Next we dene op erations on K cohomology cf B Motivic

cohomology of simplicial schemes in particular relative motivic cohomology B

is introduced as graded pieces of the ltration with resp ect to these op erations

This discussion is based on the extremely useful unfortunately unpublished pap er

GSo by Gillet and Soule More often than not the results in B and B will b e due

to them The wish for a complete published reference made us go over the material

again Meanwhile an alternative approach to K theory of simplicial schemes and

op erations was also worked out by Levine Le De Jeu was the rst to use the setting

of GSo to dene motivic cohomology of simplicial ob jects In his article Jeu he

proves RiemannRo ch in this setting We give a more general version in B

We then construct regulators ie Chern classes from K cohomology to con

tinuous etale cohomology B and to absolute Ho dge cohomology B in this sit

uation Our main interest is the construction of a long exact sequence for relative

Documenta Mathematica

Classical Motivic Polylogarithm

K cohomology of simplicial schemes as well as for their motivic cohomology which is

mapp ed to the corresp onding long exact sequences in sheaf cohomology B

We would like to thank the referee for her or his comp etent and detailed comments

and corrections

B Generalized Cohomology Theories

We need a framework which is general enough to treat K theory and the usual coho

mology theories in parallel It turns out such a framework is given by homotopical

algebra as axiomatized by Quillen in Q

We dene cohomology of spaces simplicial sheaves of sets with co ecients in

another space B We then construct a long exact sequence for relative cohomol

ogy in this context B Finally we deduce the sp ectral sequence relating general

ized cohomology of a space to generalized cohomology of its comp onents B

A systematic investigation of generalized cohomology for Grothendieck top ologies

was carried out by Jardine in particular Jr We recapitulate the denitions for the

convenience of the reader A rst introduction to the necessary simplicial metho ds is

M

We x a regular ane irreducible base scheme B of nite Krull dimension In

our applications B is either a eld or an op en subscheme of the ring of integers of

a number eld We x a small category of no etherian nite dimensional B schemes

which is closed under nite disjoint unions and contains all op en subschemes of all

its ob jects We turn it into a site using the Zariski top ology Typically this will b e a

sub category of all smo oth schemes over the base B

Let T b e the top os of sheaves of sets on our Zariski site over B Let sT b e

the category of p ointed simplicial Tob jects Its ob jects will b e called spaces in the

sequel We denote the nal and initial ob ject of sT by

Remark A space is given by a simplicial sheaf of sets X and a simplicial map from

the constant simplicial sheaf all of whose comp onents are given by the constant

sheaf attached to the set with one element to X Equivalently we can consider it

as a simplicial ob ject in the category of sheaves p ointed by

Let X b e a scheme We can also see it as an ob ject of T The corresp onding

constant simplicial ob ject p ointed by a disjoint base p oint

U Mor U X fg for connected U T

B

will also b e denoted X

Definition B A space is said to b e constructed from schemes if all comp onents

are representable by a scheme in the site plus a disjoint base p oint

Note that any simplicial scheme whose comp onents are schemes in the site

gives rise to a space constructed from schemes but there are many spaces constructed

from schemes which do not come from simplicial schemes The main example is the

mapping cone of a map of schemes taken in sT cf B b elow

If P is a prop erty of schemes and if the space X is constructed from schemes we

say X has P if the scheme parts of the comp onents have P

The easiest way to dene the homotopy sets X x of a simplicial set X with

n

basep oint x X is to take the homotopy sets of its geometric realization X x

n

Documenta Mathematica

Annette Huber Jorg Wildeshaus

is a group for n even ab elian for n If X is a space and K a nite simplicial

set ie all K are nite then we dene the space X K comp onentwise as the sum

n

of p ointed sheaves

n X

n

K

n

Definition B Brown Gersten Gillet Soule Let X b e a space and f

X Y b e a map of spaces

a f is called a weak equivalence if all stalks f X Y are weak equivalences of

P P P

simplicial sets ie if f induces an isomorphism on all homotopy sets for all

P

choices of base p oint

b f is called a cobration if for all schemes U in T the induced map f U X U

Y U is injective

c f is called a bration if it has the following lifting prop erty given a commutative

diagram

A X

f

i

y y

B Y

where i is a cobration and a weak equivalence there exists a map B X that

makes the diagram commute

d For two spaces X and Y let Hom X Y b e the p ointed simplicial set

n Hom X n Y

sT

where n is the standard simplicial nsimplex eg M p ointed by zero

This is the p ointed version of the global theory discussed in Jr x

Quillens notion of a closed mo del category axiomatizes the prop erties which

are needed in order to pass to a homotopy category which b ehaves similar to the

homotopy category of CWspaces

Proposition B Brown Gersten Joyal sT is a p ointed closed simplicial

mo del category in the sense of Quillen Q

Pro of For a mo del category we need brations cobrations and weak equivalences

satisfying a set of axioms Q I Def This is GSo Theorem Gillet and

Soule attribute this theorem to Joyal letter to Grothendieck For simplicial sheaves

a published pro of of all prop erties can b e found in Jr Cor It is an abstract

nonsense fact that with the category of simplicial sheaves the category of p ointed

simplicial sheaves is also a mo del category It is p ointed by The simplicial structure

Q II Def is given by Bd

Documenta Mathematica

Classical Motivic Polylogarithm

Technical Remark Note that the unique map X is always a cobration

ie all spaces are cobrant A space will b e called brant if the unique map X

is a bration If a space is brant than its sections X U over a scheme U form

a simplicial set satisfying Kans extension condition cf M However this

prop erty do es not suce to make X brant Part of the pro of of the prop osition is

the existence of brant resolutions In fact the construction in Jr Lemma is

even functorial

Let HosT b e the homotopy category asso ciated to the mo del category sT by

lo calizing at the class of weak equivalences As usual we will write X Y for the

morphisms from X to Y in the homotopy category If Y is brant then this set is

given by the set of morphisms from X to Y in sT up to simplicial homotopy For

general Y we compute X Y by X Y where Y is a brant resolution of Y

Remark The category of p ointed presheaves with the same notions as in B is

also a p ointed mo del category By Jr Lemma the map from a presheaf to its

sheacation is a weak equivalence and we get the same homotopy category from

presheaves or sheaves

If X is a space then its susp ension SX is given by X where is

the usual equivalence relation generated by x x By Q Ch I the lo op

space functor is right adjoint to S on the homotopy category

There are two natural ways of thinking ab out HosT From the p oint of view of

algebraic top ology it corresp onds to the category of CWcomplexes with morphisms

up to homotopy From the p oint of view of theory it corresp onds to the

category of homological complexes which are concentrated in p ositive degrees with

morphisms up to homotopy S and shift the complexes This second p oint of view

is not quite precise note that in general morphisms in HosT form p ointed sets

rather than groups

Definition B For any space A we dene cohomology of spaces with co ecients

in A by setting

m

m

H X A S X A for m

sT

This is a p ointed set for m a group for m and even an ab elian group for

m If A b elongs to an innite lo op sp ectrum ie if there are spaces A for i

i

with A A and weak equivalences A A then we also dene cohomology

i i

groups with p ositive indices by setting

nm

m

H X A S X A for m n

n

sT

Note that the set only dep ends on n m b ecause the susp ension S and the lo op

functor are adjoint

Definition B Let f X Y b e a map of spaces Then the mapping cone of f

is the space

C f X q Y

where is the usual equivalence relation of the mapping cone ie x f x

x For any map of spaces f X Y we dene relative cohomology by

m m

H Y rel X A H C f A

sT sT

Documenta Mathematica

Annette Huber Jorg Wildeshaus

C f is the standard construction of the homotopy cobre of a map

Proposition B For any morphism f X Y of spaces there is a long exact

cohomology sequence

m m m m

H Y A H X A H Y rel X A H Y A

sT sT sT sT

Pro of By Q Ch I we have the ab ove long exact sequence attached to the triple

of spaces

i

X Y Y

X

if i is a cobration The mapping cylinder of f is dened as X Y It is weakly

X

equivalent to Y and the induced mapping X X Y is a cobration The

X

mapping cone of f is nothing but the cobre of this inclusion Hence the long exact

sequence of the lemma is a sp ecial case of Quillens with Y X Y

X

If A is only a space then the sequence will end at the index zero There is no

reason for the last arrow to b e right exact The H are only p ointed sets The

sT

H are groups all others are even ab elian groups However if A is an innite lo op

sT

sp ectrum then all cohomology groups will b e ab elian groups and the sequence is

unbounded in b oth directions

We will consider a couple of sp ectral sequences which are constructed by means

of homotopical algebra Their dierentials are

pq prq r

d E E

r

r r

We refer to this b ehaviour as homological sp ectral sequence as opp osed to a cohomo

logical sp ectral sequences with dierentials

pq prq r

d E E

r

r r

In the same way as with the long exact sequences which involve p ointed sets we

also have to b e careful ab out our sp ectral sequences They will b e constructed by the

metho d of BouseldKan cf BouK Ch IX xx We refer to them as sp ectral

sequences of BouseldKan type We give an overview over their prop erties They

lo ok like this

pq q p

E L q p r

r

with homological dierentials

are ab elian groups if q p

q p pq

L E

are groups if q p

r

are p ointed sets if q p

pq

pq prq r pq

We have E Ker d im d Treat nonexisting E as zero for this

r r

r

formation By BouK IX iv this makes also sense for p q Let

pq pq pq

lim E E E

r

r r

r p

Documenta Mathematica

Classical Motivic Polylogarithm

n n

There is a descending coltration Q on the limit term L ie Q L is a quotient of

i

n

L Let

pq q p q p

e Ker Q L Q L

p p

pq pq

In general there will b e an injection e E Convergence is a more complicated

pq

question The sp ectral sequence stabilizes if all pro jective systems E b ecome

r p

r

eventually stable Then we have complete convergence BouK IX Hence the

n n

coltration on the limit term is exhaustive lim Q L L and we have isomor

s

phisms

pq pq

e E for p q

Note that even then the case p q has to b e discussed separately We refer to this

problem and more generally the fact that p ointed sets rather then groups app ear as

the fringe eect

Proposition B a Let X and A b e spaces The ltration of X by its skeletons

sq X induces a sp ectral sequence of BouseldKan type for its Acohomology

n

q p

pq q

E H X A H X A for q p

p

sT sT

It converges completely if X is degenerate ab ove some degree ie if there is N

such that for n N X is covered by the image of the degeneracy maps

n

b If A is an innite lo op sp ectrum and X as in a then we have a converging

homological sp ectral sequence

q p

pq q

E H X A H X A for p

p

sT sT

Pro of This is the hypercohomology sp ectral sequence of GSo We sketch

their pro of We can assume A to b e brant We can construct a weak equivalence

p

S X The Hom sk X A form a tower of X X such that sk X sk X

p p p p

brations of simplicial sets converging to Hom X A The attached BouseldKan

sp ectral sequence BouK x x has starting terms

pq

E Hom sk X sk X A

q n p p

q

p

Hom S X A H X A

q p p p

sT

This nishes the construction of the sp ectral sequence In order to discuss convergence

we consider the same sp ectral sequence attached to X itself It stabilizes by the

assumption on degeneracy see BouK x Both sp ectral sequences agree from r

on

For b we consider the sp ectral sequence in a for each space in the sp ectrum By

shifting q accordingly we get a direct system of sp ectral sequences whose limit is the

one we are interested in

Remark It would b e much nicer to work with sp ectra and their homotopy category

throughout It would b e a triangulated category It would help to get rid of the fringe

eects However the question of convergence of the sp ectral sequences do es not get

Documenta Mathematica

Annette Huber Jorg Wildeshaus

easier the reason b ehind this b eing that all these sp ectral sequences are constructed

for some kind of homotopy limit and pro jective limits are not exact However the

literature we want to use is in the setting of spaces The reason is that we want to

use the ring structure in order to dene motivic cohomology and the op erators

do not delo op

B K theory

We now introduce higher algebraic K theory of spaces as a generalized cohomology

theory It gives back usual K theory in the case of regular schemes B We then

dene op erators on these K cohomology groups B This allows denition of

motivic cohomology of spaces as graded parts of the ltration B We then

prove a GrothendieckRiemannRoch type theorem B As a consequence we

get a long exact lo calization sequence for motivic cohomology B

Recall that all schemes in the site underlying T are assumed to b e no etherian

and nite dimensional

Let K b e the space Z Z B Gl where Z B Gl is the simplicial sheaf asso ciated

to the simplicial presheaf U Z B Gl U lim Z B Gl U K is p ointed by

n

lim B Gl E It is in fact part of an innite lo op sp ectrum We also need the

n n

N N

unstable spaces K Z Z B Gl There are natural transition maps K

N

N

K K As K groups commute with direct limits the stalk of K in a p oint P

on U T is weakly equivalent to

Z Z B Gl O K

P P

where O is the stalk of the structural sheaf

P

Remark Even though it is wellknown that K theory is dened by a sp ectrum it

is not completely trivial to dene it as a functor from schemes to sp ectra rather

than just a functor up to homotopy We refer to GSo for the details of this

construction For a dierent account of K theory as a presheaf and its prop erties

including the pro duct structure we also refer to Jardines b o ok Jr

Definition B Gillet Soule For any space X in sT we dene its K

cohomology

m

m

X K S X K for m Z H

sT

m

N

X K for m Following GSo we call a space and the unstable K groups H

sT

m m

N

X K for m is an isomorphism X K H K coherent if lim H

sT sT

Proposition B Brown Let K b e the sheacation of the presheaf Y

q

q

H Y K Let X b e a scheme in T There is a homological sp ectral sequence

sT

q p

pq

E H X K

sT

with

p pq

H X K E

q

ZAR

It converges completely

Documenta Mathematica

Classical Motivic Polylogarithm

Pro of For q p this is the sp ectral sequence GSo Prop The basic version

for the small Zariski site was constructed in BrG Theorem Our generalization

follows from the pro of of Jr and which deals with the etale top ology The

key is to construct a Postnikovtower for K This is done as in in the pro of of BrG

Thm We then have to check that the homotopy sheaves of K are isomorphic to

the homotopy sheaves of the limit of its Postnikovtower It suces to check this

for the small Zariski site Zar Y for all schemes Y in T Hence we are reduced to

the situation considered in lo c cit Note that Y was assumed to b e no etherian and

nite dimensional We extend to arbitrary p q using the full K theory sp ectrum

Convergence follows b ecause X has nite cohomological dimension

Remark We could generalize the sp ectral sequence to arbitrary spaces X

p

H X K would have to b e understo o d as in B Convergence would not b e

q

ZAR

guaranteed anymore

The most imp ortant application of this prop osition is that it allows to transp ort

prop erties which are wellknown for cohomology with co ecients in an ab elian sheaf

to cohomology with co ecients in a space One such prop erty is the comparison

b etween dierent Zariski sites

Proposition B Gillet Soule de Jeu a Let X b e a no etherian regular

m

nite dimensional scheme in the site Then one has the equality H X K

sT

K X where the right hand side means Quillen K theory of the scheme X

m

m

X K for m In particular H

sT

b Let X b e a space constructed from schemes Assume that all comp onents are

regular No etherian nite dimensional schemes and that X is degenerate ab ove

some simplicial degree Then X is K coherent

Pro of The constant case is proved in GSo Prop We sketch a slightly

dierent argument We use the converging Brown sp ectral sequence and comparison

theorems for sheaf cohomology to show that it suces to prove the prop osition in

the case of T Zar X Note that the existence of the whole sp ectrum means we

do not have to worry ab out fringe eects In this case we have a MayerVietoris se

quence for K theory Q Rem and hence the presheaf dening K cohomology

is pseudoasque in the sense of Brown and Gersten BrG p By lo c cit

Thm this implies a for the site Zar X

The vanishing follows b ecause the K theory sp ectrum is connective The general

ization to spaces constructed from schemes using the skeletal sp ectral sequence was

carried out in Jeu and Lemma

Corollary B If X is a space meeting the conditions of part b of the prop osi

tion then its K cohomology do es not dep end on the category of schemes underlying

the top os

Pro of If X is constant then we always get its K theory For more general X we

have to use the converging skeletal sp ectral sequence There are no fringe problems

b ecause K is an innite lo op sp ectrum

Documenta Mathematica

Annette Huber Jorg Wildeshaus

The direct sum of matrices cf Lo together with addition on Z induces

a compatible system of maps

N N

K K K

Our aim is to show that its direct limit denes an H group structure on K It will

b e used to dene addition on K cohomology

Lemma B Let G G b e algebraic groups over Z E a subgroup of G with E

E E Let f f G G b e homomorphisms which dier by conjugation by a

global section of E Then the induced maps

f f

Z BG Z BG

agree in the homotopy category of spaces

Pro of The construction in Lo A is functorial Hence it yields a free homotopy

b etween B f and B f By construction we get a commutative diagram

j



Z BE

i

y y

Z BG Z BG

The comp osition of with d Z BG C i is a homotopy b etween df and df

Now it suces to show that d is a weak equivalence ie that Z BE is contractible

This can b e checked on stalks As homotopy groups commute with direct limits it is

enough to show that Z BE U is contractible for all ane schemes U We consider

the diagram

BE U BE U

y y

Z

1

Z BE U Z BE U

By denition of Quillens construction see Lo ch induces an isomorphism

on homology Hence Z is a weak equivalence BlK Ch I BE U is con

tractible b ecause E U E U E U Lo Prop osition Hence Z BE U

is also contractible

The standard application of this lemma is with G Gl G Gl and E the

n

subgroup generated by elementary matrices which contains all even p ermutation

matrices see Lo

Proposition B The direct sum of matrices induces an H group structure on

K

Pro of The same pro of as in Lo Theorem allows to check the identities of an

H space On nite level they hold up to conjugation with a p ermutation matrix

Documenta Mathematica

Classical Motivic Polylogarithm

By the previous lemma this implies that they hold in the homotopy category We

N N

use that the transition maps K K are cobrations in order to show that the

maps on nite level dene one on K For the existence of a homotopy inverse we

argue dierently An H space is an H group if and only if the shear map

K K K K k k k k k

is a weak equivalence This can b e checked on stalks But the stalks of K are the

simplicial sets computing K theory of lo cal rings They are H groups with the same

addition by the ane case Lo

Remark We now have two H group structures on K the explicit one we just have

constructed and one b ecause K is a lo op space as part of a sp ectrum We exp ect

them to b e equal but have not b een able to prove it They certainly induce the same

X K they agree at least if X is addition on higher K cohomology groups On H

sT

represented by a scheme b ecause they do for K theory of schemes This is enough for

our needs In the sequel the addition on K cohomology is the one of the prop osition

The next aim is the denition of a multiplicative structure on K We start with

the op eration of Z on K The H group structure on K allows to dene a map of

spaces

Z K K

Z

It vanishes on Z K and hence factors over Z K

The construction of the Lo day pro duct Lo

Z B Gl U Z B Gl U Z B Gl U

N N

is functorial in U Together with the pro duct on the factor Z it denes a system

Z

of maps

N N

K K K

K

compatible up to homotopy which denes a pro duct

Y K Y K Y K

n

for all K coherent spaces Y It turns all H Y K for n into a ring p ossibly

sT

without unity

n

Remark Note that this pro duct on Y K is zero on H Y K for n cf Kr

sT

Ex p The same map of spaces also induces a nontrivial pro duct

K

n m nm

S Y K S Y K S Y K

This is the one which is usually called Lo day pro duct We do not need it in the sequel

Let S b e the simplicial version of the sphere ie the constant simplicial sheaf

asso ciated to f g p ointed by We will use the notation K sT for H S K

sT

It is a ring with unity where the ring structure is induced by the ring structure on Z

Documenta Mathematica

Annette Huber Jorg Wildeshaus

Lemma B If the site underlying T has a nal ob ject X then

K X K sT

Pro of If X is the nal ob ject of the site then the space we denote by X is equal to

S

The following lemma generalizes an op eration of K X which was explained to

us by de Jeu in the case where Y is constructed from X schemes

n

Lemma B Let Y b e a space in sT Then the ring K sT op erates on H Y K

sT

for n and makes it into an K sTalgebra

Pro of If Y is a space in sT then there is canonical isomorphism Y S Y The

n

pro duct K sT with H Y K is dened by the comp osition

sT

K

Y S Y K K K

Lemma B Gillet Soule Let G b e a group over Z Let R G b e the

Z

Grothendieck group of representations of G on free Zmo dules of nite type

a Let A b e an N dimensional representation of G There is a canonical class in

Z Z B G K which dep ends only on the equivalence class of A The direct

sum of representations is mapp ed to the sum of classes

b The map in a induces an algebra homomorphism

r R G Z Z B G K

Z

Pro of We follow GSo or the ane case Kr By choice of a basis of an

N dimensional representation A induces a map of sheaves

A G Gl

N

and hence by functoriality a map

N

r A Z BG fN g Z B Gl K

N

For dierent choices of basis the maps dier by conjugation with an element of

Gl The matrix is in the p erfect subgroup E Gl Gl hence by Lemma

N

N

B the image of r A in Z B G K do es not dep end on the choice of matrix

Viewed as map to K this r A extends to the factor Z using the ab ove pro duct

Z

The last statement of a follows by denition of the H group structure on K

For b we have to check that the relations of the Grothendieckgroup are mapp ed to

zero and that the multiplicative structure is wellbehaved We rst prove the analogue

of Kr Theorem The canonical maps

Gl Gl

induce weak equivalences of simplicial sheaves after applying Z B This can b e

checked on stalks and is hence reduced to the ane case From now the pro of works

precisely as in the ane case see Kr Cor

Documenta Mathematica

Classical Motivic Polylogarithm

K sT is a ring ie the axioms in Kr Def are satised If R is a

K sTalgebra then it is called a K sTalgebra if it is equipp ed with op erators

i

for i such that K sT R is a ring cf Kr Note that has to have

the constant value If R itself do es not have a unity then it cannot b e a ring

Theorem B Gillet Soule Let Y b e a K coherent space For k and

m there are maps

m m

k

H Y K H Y K

sT sT

m

They turn H Y K into a K sTalgebra

sT

Pro of This is essentially GSo Prop Put G Gl in the previous lemma Let

n

n n n

is the canonical representation of Gl on n R Gl where Z Z Z

n Z n

id id

n k k n

Z and is the trivial representation We dene r Z By comp osition it

n

m m

k n

induces a map H Y K H Y K These form a pro jective system and

n

sT sT

hence dene an op eration on K cohomology of a K coherent space Welldenedness

n

and all prop erties of a ring are checked on the universal level ie on K for varying

n and hence as in the ane case Kr Thm For example we want to show

k

X

i j k

x y x y

i

n

Assume that x y are represented by elements in Y K On R Gl Gl we have

Z n n

the ring identity

k

X M

i j k

i

n

We evaluate this identity in Z and get an equality of elements in R Gl Gl

Z n n

n n

By the previous lemma it induces the same equality of elements in K K K

Comp osed with x y this is the required equality

Remark A more conceptual pro of was suggested to us by Soule and the referee

One should use the integral completion functor constructed by Go erss and Jardine

Go eJr It has a universal prop erty similar to the one of the construction and

hence allows to copy directly Kratzers arguments

Technical Remark When we try to dene in the same way then we still get a

map

Z B Gl Z Z B Gl

N

It do es not extend to the factor Z b ecause Z Z do es not resp ect the base

p oint in fact it maps to This reects the fact that the ring K Y do es not

have a unity for a general space Y The most striking example is Y C i where

i Z X is a morphism b etween regular schemes cf Sou Then K Y

Ker K X K Z do es not contain

Gillet and Soule GSo Prop consider the structure as a H Y Kalgebra

sT

Y K happ ens to have a unity However we can check This only makes sense if H

sT

Documenta Mathematica

Annette Huber Jorg Wildeshaus

m

Y K on H Y K is compatible with the in general that the op eration of H

sT

sT

K sTalgebra structure of b oth groups

Note that the structure is compatible with the contravariant functoriality of

K cohomology This means that the long exact sequences for relative K theory are

compatible with the op eration where it is dened

Once we have op erations we get as usual a ltration and Adamsop erators

n

Y K for n If the ltration is lo cally nite then we on the mo dule H

sT

have in particular the Chern character

M

j

n n

ch H Y K Gr H Y K for n

Q Q

sT sT

j N

which is an isomorphism For a quick survey cf T pp

Definition B Let Y b e a K coherent space Supp ose that the ltration is

lo cally nite and hence that rationally K cohomology splits into Adamseigenspaces

Then we put for j n

nj

j

n

H Y K H Y j Gr

Q

M

sT

the motivic cohomology of the space Y If i X Y is a morphism of spaces then

we dene relative motivic cohomology by

n n

H Y rel X j H Conei j

M M

Remark We restrict to this range of indices b ecause we did not dene Adams

eigenspaces for K cohomology with p ositive indices K theory with negative in

dices However if these K groups vanish we can simply dene the corresp onding

motivic cohomology groups to b e zero This is the case if X is a regular scheme

The long exact sequence for relative cohomology B together with the ab ove

remarks on the op eration give a long exact sequence for relative motivic cohomology

m m m m

H Y A H X A H Y rel X A H Y A

M M M M

Lemma B Let X b e a space degenerate ab ove some simplicial degree We

assume the conditions of the previous denition Fix an integer j There is a coho

mological sp ectral sequence with starting terms

t

H X j for s j t

s

st

M

E

else

st

It converges to H X j for j s t

M

Pro of Consider the skeletal sp ectral sequence Ba with co ecients in the space

K It reads

q p

pq q

X K H X K E H

p

sT sT

for p By carefully checking the construction of the sp ectral sequence we see that

pq

are induced by functoriality in the rst argument Hence they all dierentials d

r

Documenta Mathematica

Classical Motivic Polylogarithm

are morphisms of mo dules For q p the limit terms are also mo dules and

pq pq

by construction the morphisms e E are compatible with this structure They

are isomorphisms for q p Note however that we do not get enough information

pp

on the limit terms on the p q line Convergence only implies that e injects into

pp

E We want to show that it is even a bijection In order to see this we consider the

skeletal sp ectral sequence with co ecients in the sp ectrum K The sp ectral sequences

agree where the rst is dened in particular convergence of the second sp ectral se

quence implies our isomorphism There is an issue here with the H group structure

A priori the two sp ectral sequences use dierent group laws But on all initial terms

they give the same addition and hence also on all higher terms

Now we take Adamseigenspaces By reindexing s p t q j we get a coho

mological sp ectral sequence as stated Note that we use the terms b elow the p q

diagonal to compute the terms on it but we do not consider their limit terms

The same sp ectral sequence also shows that the conditions in the denition of

motivic cohomology hold if X is a space constructed from schemes and degenerate

ab ove some degree

The next thing we need is pushout at least for certain closed immersions and a

RiemannRo ch theorem Over a eld pushforward was dened by de Jeu in Jeu

We adapt his metho d to more general bases and formalize the geometric situation

Definition B Let S b e a regular irreducible No etherian ane scheme Let X

b e smo oth and quasipro jective over S A nite diagram D over X is a category

X

of nitely many smo oth quasipro jective S schemes with nal ob ject X such that all

Y Y are nite sets and such that all morphisms in D are of nite Tor Mor

X D

X

dimension

we mean the category of all nite disjoint unions of By the small Zariski site Zar

D

X

op en subschemes of ob jects in D with the induced morphisms b etween them It is

X

equipp ed with the Zariskitop ology The corresp onding top os will b e denoted T

X

An easy case of such a diagram is a single morphism Y X that meets the

conditions

We consider the following situation Let i Z X b e a closed immersion of

smo oth quasipro jective S schemes and D a nite diagram over X We assume the

X

following conditions corresp onding to the ones formulated by de Jeu in Jeu

TC For all X in D the pullback X Z is S smo oth If f X X is a

X X

morphism in D then in the cartesian diagram

X

Z X Z X

X

f f Z

X

y y

i

Z X Z X

X

the maps f and i are torindep endent ie

k

O O Tor

X Z

O

X

k

for k Tor denotes the sheaf of torgroups

Documenta Mathematica

Annette Huber Jorg Wildeshaus

Lemma B The pullback D of D by Z satises the conditions for a nite

Z X

diagram over Z

Pro of Finite Tordimension in D follows from Torindependence and the same prop

Z

erty in D

X

Let Y b e a space in sT Let j U X b e the op en complement of Z in X

X

Let Y U b e the p ointed version of j j Y ie the sheaf asso ciated to the presheaf

X

Y V if V U X

V

else

It is a space in sT Let Y Z i Y a space in sT If Y is constructed from

X X Z

schemes then so are Y U and Y Z The scheme comp onents are given by

X X

the base change with U or Z resp ectively Note that i Y U is empty ie only

X

consists of the base p oint

Proposition B de Jeu Let i Z X b e a closed immersion with op en

complement U Let D b e a nite diagram over X such that TC holds with resp ect

X

to i Then for Y sT

a There is a natural pushout map

k k

Y K Y Z K H H

X

sT sT

X Z

b Let Y b e a space in sT which is constructed from schemes We assume that it

X

is degenerate ab ove some simplicial degree Then

Y Z C Y U Y Z

X X X

and the pushout

k k

Y rel Y U K Y Z K H H

X X

sT sT

X Z

is an isomorphism

let M V b e the category of all coherent Pro of For an ob ject V of the site Zar

D

X

sheaves on V In it let P V D b e the sub category of those sheaves F satisfying

X

j

0

F O Tor

V

O

V

for all j and all V V in D Note that there are only nitely many conditions

X

as our diagram is nite The nice thing ab out P V D is that it is contravariantly

X

functorial Hence Quillens B QP D lo op space of the classifying space of the

X

Qconstruction denes a presheaf of simplicial sets on the site by Q x It is

here where we use the fact that all schemes are quasipro jective Let B QP b e the

X

space in sT dened by its sheacation By Quillens Resolution Theorem Q

X

Thm Cor p there is a weak equivalence of spaces B QP K Basically

X

X

this is the fact that K theory and K theory agree for regular schemes

We also have the space B QP in sT For the closed immersion i V Z V

Z X

Z

the pushout i is exact on the category of coherent sheaves Because of TC it maps

Documenta Mathematica

Classical Motivic Polylogarithm

the the sub category P V Z D to P V D In fact we get a morphism of spaces

Z X

in sT

X

i



B QP i B QP

X Z

Using the weak equivalences to K this denes a map in the homotopy category

i



K i K

X Z

If Y is a space in sT then we get the map in a as

X

k k k

Y K i i Y i K H i Y K H H

X Z Z

sT sT sT

X X Z

In the sp ecial case of a scheme Y part b is nothing but Quillens pushout isomorphism

K i Y K Y rel Y U

n n X

for regular schemes Q x Prop recall that all schemes in the site are regular

This generalizes to the case of spaces constructed from schemes by the skeletal sp ectral

sequence

Lemma B Consider a cartesian diagram of smo oth quasipro jective S schemes

0

i

X Z

f f

X Z

y y

i

Z X

where i is a closed immersion Let D b e a nite diagram on X Assume that the

X

0

denes a nite diagram over X and that b oth i and i satisfy TC pullback D

X

We also assume that for all V in D the maps

X

V X V

X

and

V Z V

X

are torindep endent

Then for all spaces Y in sT there is a commutative diagram

X

0

i



k k

H f i Y K H f Y K

sT Z sT X

0 0

Z X

x x

 

f f

X Z

i



k k

H H i Y K Y K

sT sT

Z X

Pro of We have to rene the categories P V D used in the pro of of B further

Z

Let P V D b e the sub category of P V D of those coherent sheaves F satisfying

Z Z

j

0

O F Tor

Z

O

Z

Documenta Mathematica

Annette Huber Jorg Wildeshaus

is again weakly equivalent to K By Q x there The induced space B QP

Z

Z

is a commutative diagram of spaces in sT

X

i f B QP f P

0 0

X X

Z X

x x

i B QP B QP

Z X

This proves the lemma

We also need the following lemma from algebraic geometry

Lemma B Supp ose we are given a cartesian diagram

0

i

Z X

f

y y

i

Z X

of smo oth S schemes where i is a closed embedding then the blowup of X in Z is

the base change by f of the blowup of X in Z provided i and f are torindep endent

n

Pro of In order to see this note that by EGAI I we have to check that f I

n n

0

is isomorphic to J where I is the sheaf of ideals of Z in X and J the O I

X O

X

one of Z in X This follows from torindep endence in the case n Note that in

n n

general we have a surjection f I J Let K b e the kernel Pullback by f is

n

right exact ie we have an exact sequence

f I J f I I

Together with the ab ove surjectivity this implies f I I J J As X resp ec

tively X are regular and Z resp ectively Z are lo cally given by regular sequences the

structural theorem Ha II Theorem A e implies

n n n n

f I I J J

n nk n nk

By the snake lemma K K is surjective and hence f I I J J

n n

for all k But then

n n k n n nk n nk n

f I lim f I J f I lim f I im f I lim J J J

Pushforward is not a ring morphism but it do es resp ect the ltration up to a

shift at least under go o d conditions This is made precise in the following Riemann

Ro ch Theorem which is a slight generalization of de Jeus in Jeu He considers

a sp ecial type of diagram and restricts to a base eld De Jeu imitates the pro of in

T Theorem which is over a eld However his arguments work for our base as

well Indeed the original article Sou Thm treated the more general case

Documenta Mathematica

Classical Motivic Polylogarithm

Theorem B GrothendieckRiemannRoch Let S b e a regular irre

ducible No etherian ane scheme S Let i Z X b e a closed immersion of

constant co dimension d of quasipro jective smo oth S schemes For X Z let

td Gr K b e the usual Todd classes eg T p Let a nite diagram

Q

D b e given that satises the conditions TC with resp ect to i Finally let Y b e a

X

space constructed from schemes in sT

X

a The homomorphism i K i Y K Y has degree d with resp ect to

n Q n Q

the ltration ie

i



j d j

F K Y F K i Y

n Q n Q

b The following diagram commutes

tdZ ch

K i Y Gr K i Y

n Q n Q

i i

 

y y

tdX ch

K Y Gr K Y

n Q n Q

Remark td is a unit with augmentation Hence the horizontal maps in b are

isomorphisms

Pro of We essentially have to prove classical RiemannRo ch for the inclusion Z X

The conditions on our situation are chosen in a way that the diagrams we drag along do

not make any diculties Note also that we can replace Y by the cone of Y U Y

ie we can assume that all pushout maps are isomorphisms Having observed this

we can follow de Jeus arguments in Jeu

The rst step is to prove the analogue of T Theorem or Jeu Prop osition

RiemannRo ch without denominators We only sketch the idea Because

of functoriality B and the homotopy prop erty of K theory we can make the

transformation to the normal cone Hence we can assume without loss of generality

that i is a section of a pro jective bundle over Z The existence of the pro jection

p which is a leftinverse of i allows to make explicit calculations All details of the

argument can b e found in Jeu when replacing K Y K X there by

K X K sT The necessary compatibility of blowup and base change is

X

guaranteed by the previous lemma

We then show that up to multiplication with the appropriate Todd class i has

the required b ehaviour with resp ect to Adams eigenspaces The argument is the same

as in Jeu Prop osition or T Lemma Now the theorem follows by the same

formal manipulations as in the pro of of T Lemma

Corollary B Let i Z X closed immersion of constant co dimension d

and Y b e as in the theorem Let U X r Z Then there is a natural lo calization

sequence

K Z Y K Y K U Y

m X Q m Q m X Q

K Z Y

m X Q

Documenta Mathematica

Annette Huber Jorg Wildeshaus

or in terms of motivic cohomology

id

i i

H Z Y j d H Y j H U Y j

X

M M

M

id

H Z Y j d

X

M

Pro of Part b of Theorem B implies that

M M

j j d

i Gr K Y rel Y U Gr K Y Z

m m

j N j N

id

i

Y rel Y U j is an isomorphism ie H H Z Y j d

X

M

M

We consider the long exact sequence of relative K cohomology or relative motivic

cohomology for the op en embedding U Y Y We can use i to identify the

relative cohomology with cohomology of the closed complement

Only a few K groups are known However the ranks of the K groups of number

elds are understo o d

Theorem B Borel Let K b e a number eld with ring of S integers o

S

where S is a nite set of primes of K Let B Sp ec o As usual r is the number of

S

real places of K and r the number of complex places Then the motivic cohomology

has the following ranks

H B

M

H S r r B

M

H B n r n even

M

H B n r r n o dd

M

i

else H B j

M

Pro of The computation of K B and K B is classical Ba Ch IX Prop and

Ch X Cor The higher K groups for the ring of integers o were calculated by

K

Borel Bo Prop It follows from Quillens computation of the K groups of

nite elds that the ranks are not changed by lo calizing at nite primes

B Cohomology of Abelian Sheaves

We now show how the usual cohomology theories t in the setup of generalized

cohomology This is well do cumented in the literature BrG G Jeu In the

case of a cohomology theory dened by a pseudoasque complex of presheaves F

we compare the dierent p ossible p oints of view These are Zariskicohomology of

the asso ciated complex of sheaves generalized cohomology of the asso ciated space or

simply cohomology of the sections We always get the same cohomology groups B

and B If the complex of presheaves F is part of a twisted duality theory B

we dene Chern classes from K cohomology of spaces to cohomology with co ecients

in F Finally we check compatibility of the lo calization sequence in K cohomology

with the one for cohomology of spaces with co ecients in F B

By a complex we always mean a cohomological complex Of course it can also

b e considered as a homological complex by inverting the signs of the indices

The DoldPupp e functor M Thm attaches to a complex of ab elian groups G

which is concentrated in nonp ositive degrees a simplicial ab elian group K G p ointed

Documenta Mathematica

Classical Motivic Polylogarithm

i

by whose homotopy groups K G agree with the cohomology groups h G

i

It induces an equivalence b etween the homotopy category of simplicial ab elian groups

and the homotopy category of complexes of ab elian groups concentrated in non

p ositive degrees By construction of the functor K there is a natural weak equivalence

of spaces

ConeK G K ConeG K G

and hence a natural map K G K G in the homotopy category of p ointed

simplicial sets which is a homotopy equivalence If G is an arbitrary complex of

ab elian groups let G b e the canonical subcomplex in degrees less or equal to N

N

We put

K G K GN

N N

The natural map GN GN induces

N N

K GN K G K G

N N N

which is a weak equivalence This means the K G form an innite lo op sp ectrum

N

whose homotopy groups reect all cohomology groups of the complex

Definition B Let G b e a cohomological complex of sheaves of ab elian groups

on the big Zariski site The sheaed version of the ab ove construction yields an

innite lo op sp ectrum of spaces K G with

i

h G K G

i

where the right hand side is the sheacation of the presheaf

U K G U

i

As a sp ectrum K G denes generalized cohomology groups with indices in Z for

any space X

Proposition B Let G b e a b ounded b elow complex of sheaves on the big Zariski

site Let X b e a scheme Then

i i

H X K G H X G

sT ZAR

Pro of As G is b ounded b elow it has a b ounded b elow resolution by asque sheaves

Now the pro of pro ceeds as in BrG Prop The main ingredient is that K I is a

brant space if I is a asque sheaf

Definition B a Following BrG Sect a complex F of ab elian presheaves

on the big Zariski site is called pseudoasque if it has the MayerVietoris prop

erty ie for op en subschemes U and V of some scheme X we have a long exact

sequence of ab elian groups

i i i

h F U V h F U F V h F U V

i

h F U V

Documenta Mathematica

Annette Huber Jorg Wildeshaus

More precisely the square

F U V F U

y y

F V F U V

is homotopically cartesian

b Let F b e a complex of ab elian presheaves For the ob ject q U in T where U

is a scheme we put

F q U F U

Let X b e a space constructed from schemes Then we put

F X Tot F X

i i

the total complex of the cosimplicial complex F X

i iN

Taking the total complex of a bicomplex as in b of course involves a choice of

signs which we x once and for all Dierent choices of signs dier by a canonical

isomorphism of the total complex

Lemma B Let F b e a b ounded b elow pseudoasque complex of ab elian

presheaves Let F b e its sheacation Then

i i

H X K F h F X

sT

for all spaces X constructed from schemes

Pro of Let I b e a b ounded b elow asque resolution of F This is in particular a

pseudoasque complex of presheaves that is quasiisomorphic to F as a complex of

presheaves b ecause b oth compute Zariskicohomology of F As in the pro of of BrG

Theorem the simplicial sheaf K I is a brant resolution of K F Hence we can

assume without loss of generality that F itself is a complex of asque sheaves

For the case of a scheme X the lemma is the reformulation of BrG Theorem in the

easier case of simplicial presheaves that come from a complex of ab elian presheaves

In the general case

i

H X K F Hom X K F

i

sT

Hom ho colim X K F

i j

holim Hom X K F BouK XI I Prop

i j

i i

h Tot F X h F X

i

This means if we dene a cohomology theory by a pseudoasque complex of

presheaves on the big Zariski site we can freely change from the p oint of view of

generalized cohomology to ordinary Zariskicohomology or cohomology of the sections

of the presheaf

Documenta Mathematica

Classical Motivic Polylogarithm

If X Y is a morphism of schemes we consider as usual its Cechnerve

cosk X Y ie the simplicial Y scheme given by

cosk X Y X X n fold pro duct

n Y Y

with the natural b oundary and degeneracy morphisms

Definition B We say that a morphism X Y of schemes has cohomological

descent for the cohomology theory given by the complex of ab elian Zariskisheaves G

if the natural morphisms

i i

H Y K G H cosk X Y K G

sT sT

are isomorphisms for all i Z

This is of course a very sp ecial case of the general notion of cohomological descent

Lemma B Let j U X b e an op en immersion with closed complement Y

Let F b e a pseudoasque complex of presheaves on ZAR with sheacation F

X

a There are natural isomorphisms

i i

H X rel Y K F H X j j F

sT ZAR

b If Y Y is a morphism with cohomological descent for F then we get a natural

isomorphism

i i

H X rel cosk Y Y K F H X j j F

sT ZAR

Pro of By B the lefthand side of a is canonically isomorphic to the cohomology

of

F i

i

F C Y X Cone F X F Y

where the right hand side is the cone in the category of cohomological complexes We

assume without loss of generality that F is a asque complex The key p oint is the

short exact sequence of complexes of sheaves on X

j j F F i i F

It induces a canonical quasiisomorphism of complexes

j j F Cone F i i F

We now take R X of the righthand side Because F was assumed to b e

Zar

pseudoasque the morphism

Cone F X F Y Cone F X F Y

is a quasiisomorphism This last fact follows from B and B Of course it can

also b e proved even more easily in terms of complexes of ab elian groups rather than

Documenta Mathematica

Annette Huber Jorg Wildeshaus

simplicial ab elian groups In the case of a morphism Y Y with cohomological

descent the left hand side of the statement is by B given by the cohomology of

Cone F X F cosk Y Y

The natural morphism F Y F cosk Y Y is a quasiisomorphism by denition

and Lemma B

L

Theorem B Gillet de Jeu Let F F i b e a pseudoasque com

iZ

plex of ab elian presheaves on the big Zariski site Assume that F denes a twisted

duality theory ie the extra data of G Def exist and all conditions of lo c cit

Def are fullled Then

There are Chern class maps of spaces

c K K F j j

j

They induce morphisms

ij

i

c H Y K F j Y K H

j

sT

sT

for all spaces Y in sT

If Y is a K coherent space then the total Chern class c is a morphism of

algebras on K cohomology of Y

Let i Z X a closed immersion of smo oth S schemes with op en complement

U The map i i F r j F r d j d required in G Def induces

Z X

pushforward on generalized cohomology If Y is a space over X as in B

then the diagram

i



j d j

Gr K Y Gr K Y Z

n Q n X Q

c c

j

j d

y y

i

j n j dn

H Y Z K F j H Y K F j d

X Q Q

sT sT

is commutative

Pro of The construction of the Chern classes is G Thm Gillets formulation is

for schemes but he constructs in fact a morphism of spaces lo c cit p so the

results hold for more general spaces see also GSo The assertion on the ring

structure is GSo Thm We sketch the idea Everything is dened on the level

of co ecients so it do es not dep end on Y Compatibility with multiplication is G

Compatibility with op erators can b e checked on the level of universal Chern

i

classes ie for elements C H B Gl F i Now use the splitting principle

iN n

sT

G

The last part of the prop osition is a generalization of Gillets RiemannRo ch Theorem

G to spaces of our sp ecial type The pro of carries over by the same metho d as in

the pro of of RiemannRo ch for K cohomology B Mutis mutanda the statement

can b e found in Jeu Lemma

Documenta Mathematica

Classical Motivic Polylogarithm

Remark This will allow to dene regulator maps from K cohomology to the coho

mology theories we are interested in

Corollary B Let X Z d Y and F b e as in the theorem In addition assume

that F is pseudoasque Let U b e the complement of Y in X We abbreviate

Y Y U Y Y Z and F K F j Then there is a natural morphism

U X Z X j

of long exact sequences

i id

i i

H Y j H Y j d H Y j H Y j

U Z U

M M

M M

y y y y

i id

i i

Y F Y F H H Y F H Y F H

U j j U j Z j d

sT sT

sT sT

y y y y

i id i i

h F j Y h F j dY h F j Y h F j Y

U Z U

Pro of We start with the long exact sequences for relative cohomology B with

co ecients in the sp ectrum K and in the sp ectrum K F Their compatibility is

nothing but functoriality Relative cohomology is replaced by cohomology of Y Z

X

using B Finally we pass to graded pieces of the ltration Note that the indices

in the denition of motivic cohomology are chosen in a way that they agree with the

indices of other cohomology theories under Chern class maps Equality of the last

two lines is B

Note that the last line has nothing to do with generalized cohomology or spaces

B Continuous Etale Cohomology

There are dierent ways of dening continuous etale cohomology We will see that

they all give the same thing

Fix a number eld K and a prime l Let B b e an op en subscheme of Sp ec o l

K

where o is the ring of integers of K

K

Proposition B Deligne Ekedahl Let f Y X b e a morphism of B

b

schemes of nite type Then there are triangulated categories D X Z and D Y

l c

c

Z admitting the following there is a tstructure whose heart are the constructible

l

ladic systems There are functors

b b

f f D Y Z D X Z

l l

c c

and

b b

f f D X Z D Y Z

l l

c c

having all the usual prop erties of Grothendieck functors

Pro of This is Ek Thm In the case B Sp ec o l the category was already

K

constructed in D

Documenta Mathematica

Annette Huber Jorg Wildeshaus

b

Remark D X Z should b e thought of as the b ounded derived categories of

l

c

b

constructible ladic sheaves on X By Ekedahls construction D X Z is a sub

et l

c

category of a lo calization of a sub category of the derived category of the ab elian

N

category X Z By this notation Ekedahl means the category of pro jective sys

et l

n

tems of etale sheaves on X ringed by the pro jective system Zl The four functors

are dened on the level of this last derived category Ekedahl then shows that they

b

induce welldened functors on D X Z In the case B op en in Sp ec o l we

l K

c

get away with Delignes more straightforward construction

Definition B Version a For k Z let Z k b e the constructible ladic

l

k

sheaf on B given by the pro jective system

n

l

b We dene continuous etale cohomology of s X B by

i

H X k Hom b s Z s Z k i

l l

D X Z

cont

l

c

c If j U X is an op en immersion with complement Y we dene relative contin

uous etale cohomology by

i

H X rel Y k Hom b s Z j s j Z k i

l l

D X Z

cont

l

c

b

X Z We dene continuous etale d More generally let M b e an ob ject of D

l

c

cohomology of X with co ecients in M as

i

H X M Hom b s Z Mi

l

D X Z

cont

l

c

This denition allows to derive all the usual sp ectral sequences from the calculus

of the Grothendieck functors

Remark As checked in H x this denition coincides with Jannsens original one

in Jn sect In our case continuous etale cohomology with co ecients in a

n

X F b ecause all constructible ladic sheaf F is nothing but the naive lim H

n n n

et

n

X F are nite H

n

et

Let us now dene continuous etale cohomology in a way that ts in with the

setting of the previous section

Definition B Version Consider the pro jective system of sheaves

k

on the big etale site over B Let I b e an injective resolution in the

n

l

nN

category of pro jective systems It is given by a pro jective system I of injective res

n

k

on the big etale site with split surjective transition morphisms Jn olutions of

n

l

By taking sections we get a pro jective system of complexes of Zariskipresheaves

k

The functor R lim turns it into a complex F k of Zariskipresheaves R

n

l

l

nN

For any space X put

i i

H X k H X K F k

l

cont sT

In particular if Y X is a morphism of spaces then we put

i i

H X rel Y k H C K F k

l

cont sT

Documenta Mathematica

Classical Motivic Polylogarithm

Lemma B If X is a B scheme then b oth versions of the denition of continuous

etale cohomology agree canonically If Z X is a closed immersion then the same

is true for b oth denitions of relative continuous etale cohomology

Pro of F X is nothing but an explicit version of the derived functor R lim R X

l

from the derived category of pro jective systems of etale sheaves to the derived category

of ab elian groups Hence the complex F X computes the rst version of continuous

l

etale cohomology In particular it has the MayerVietoris prop erty Hence we can

apply the lemmas of the previous section B and get

i i

H X K F h F X

l l

sT

To extend the result to relative etale cohomology we use essentially the same

argument as in Bb

Remark When we say that the isomorphism is canonical we think in particular of

the following situation The cartesian diagram of schemes

0

0

j

i

Y U X

f

y y y

j

i

U X Y

x x x

g

0 0

0 0

j

i

Y X U

f j op en g i closed complements induces a map



f

i i

H X rel Y n H X rel Y n

cont cont

which is compatible with the identication If all schemes are smo oth and X inter

sects Y transversally then we also get the same long exact sequence

id

i i

H X rel Y n d H X rel Y n H X rel Y n

cont cont cont

id

H X rel Y n d

cont

using either denition of relative cohomology

Lemma B If Y Y is a prop er covering ie a prop er and surjective map

then it has cohomological descent for continuous etale cohomology In particular if

Y X is a closed embedding and Y a prop er covering of Y then there is a natural

isomorphism

i i

H X rel Y j H X rel cosk Y Y j

cont cont

where the right hand side is taken in the sense of spaces

Pro of Cohomological descent is a consequence of the same descent for etale coho

mology with torsion co ecients prime to the characteristic of the schemes SGAI I

Exp Vbis By Bb the second part follows

Documenta Mathematica

Annette Huber Jorg Wildeshaus

Proposition B On the Zariski site of smo oth schemes over B the presheaf

F has the prop erties of a twisted duality theory There are regulator maps from

l

K cohomology to continuous etale cohomology

i i

H Y j H Y j

M cont

for all K coherent spaces Y They are compatible with pullback ie if f Y Y

is a map of K coherent spaces we get commutative diagrams



f

i i

H Y j H Y j

M M

c c

j j

y y



f

i i

H Y j H Y j

cont cont

If i Z X is a closed immersion of smo oth schemes constant co dimension d with

op en complement U and Y a space constructed form schemes over X as in B

then the regulator is compatible with pushout ie the diagram

i



nd

n

H Y j H Y Z j d

X

M

M

c

c

j d j

y y

i

nd

n

Y j H Y Z j d H

X

cont cont

is commutative

Pro of We restrict to smo oth schemes for simplicity We have to dene the extra

structure from G and We put

di

H X j H X d j

i

cont

for a ddimensional smo oth connected scheme Pullback on cohomology and pushout

on homology are induced from the functors on sheaves on the etale site We do not

n

this is actually one of Gillets work out the details For a single etale sheaf

l

examples iii

There is really only one case when this regulator is understo o d

Lemma B Let K b e a number eld o b e its ring of integers and l a prime

K

Assume i k then Soules l adic regulator

k

Sp ec o l i K o l Z H

K ik K l

cont

agrees with the one obtained from Prop B

Pro of Put A o l Soules denition in Sou is the comp osition

K

c lim

ik

k

H A Zl i lim K A Zl K A lim

ik ik

et

Documenta Mathematica

Classical Motivic Polylogarithm

where c is as in Sou I I There is a natural map of presheaves F i

ik l

R Zl i Hence in Gillets denition of Chern classes we get a commutative

diagram

c

i

k

H Sp ec A i K A

ik

cont

y

k

H Sp ec A Zl i

et

Hence we only have to consider nite co ecients Furthermore in this simple case of

a regular commutative ring we do not really need to consider the sheaed versions

and generalized cohomology Gillets construction b oils down to a comp osition of the

Hurewiczmap with universal Chern classes

c is dened by the same type of comp osition Sou II For i k the map

ik

with the same universal Chern classes

By the denition of K theory with co ecients we have a commutative diagram lo c

cit I I with X Z B Gl A

l

X X X Zl

n n n

h

q h h

y y y

l

H X Z H X Z H X Zq

n n n

For the prime compare also We

Theorem B Soule Let K b e a number eld o b e its ring of integers and l

K

any prime Let S b e a nite set of prime ideals of o and S S fl g Let o b e

K S

the lo calization of o at S The regulator map

K

i i

0

j Q H Sp ec o j c H Sp ec o

Q l S Q j S

cont M

is always injective and an isomorphism for i and j We have the following

b ehaviour for pairs of indices i j

j j Z isomorphism

j j mot coh vanishes ladic do es not in general

injective of nite co dimension

isomorphism j j

j j conjectured to b e isom ie etale coh to vanish

injective of nite co dimension

isomorphism ie b oth vanish j j

i j else b oth vanish

Pro of We have

i i

Sp ec o j H G Q j H

S Q S l

cont

where G is the Galois group of the maximal extension of K that is unramied

S

outside of S We rst check that these groups vanish for i By Mi I Cor

j

i

all H G are nite This means that the pro jective systems for varying n are

n

S

l

Documenta Mathematica

Annette Huber Jorg Wildeshaus

ArtinRees We do not get a lim contribution to continuous cohomology Moreover

j

i

by lo c cit I c the H G for i are torsion This implies that their

n

S

l

i

pro jective limit is torsion In total we have vanishing cohomology H G Q j

S l

for i

The case i is trivial H G Q E Q where E are the S units while

S l S l S

0

o Sp ec o H Q For H G Q the S Brauergroup the co dimension

0

S l S l

S M

is the same as in the case by EulerPoincare duality cf the discussion in

Jn Lemma and Cor In the remaining cases neither motivic B nor

continuous etale cohomology Jn Lemma is changed by the inversion of S at

least up to torsion We assume S For o dd l the cases j and j for

j are Soules result in Sou Theorem Note that we are in the range where

the previous lemma applies

For l we have to rene the argument On the level of Q co ecients we may by

p

note that the only prime which could Galois descent assume that K contains

p ossibly ramify in this quadratic extension has b een inverted and hence we get an

etale extension of rings By DwF Theorem and the succeeding remark we have

surjectivity even for l

To conclude we need to show that the Q vector spaces have the right dimension

Let j By Jn pro of of Lemma the dimension of

i

H Sp ec o j

K Q

cont

equals the corank of

i

H Sp ec o Q Z j

K

cont

By Sou and Prop osition this corank for i equals the rank of the

K group if and only if

H Sp ec o Q Z j

K

cont

is torsion This in turn follows from We Theorem

Finally we want to discuss Soules elements in K theory with co ecients Every

thing is in the setting of simplicial sets and sp ectra in the usual sense Generalized

r

cohomology do es not enter Let b e the sphere sp ectrum and l a prime p ower By

denition of the Mo ore sp ectrum there is a cobration sequence

r

r

r

j

i

l

l

l

r

M S

l

Recall that for the ring of integers in a number eld A

K A Z lim K A Zl lim K M

n l n n l

The Mo ore sp ectrum has a unique pro duct for l For l r there are two

r

O Theorem a b pro jective systems of regular pro duct structures on M

l

and Lemma Together with the pro duct structure on K this denes a pro duct on

K A Z for l

l

Documenta Mathematica

Classical Motivic Polylogarithm

For d we dene R Z dl Recall Sou Lemma Sou

d

Soules construction of maps

primitive elements of K R Z K R Z

l d n l n Z l

The original statement is for o dd primes l but using the ab ove adic pro duct the

construction works without any changes for l For a primitive dth ro ot of unity

l r

r

satisfying Let lim K R Zl choose some lim

r r r r dl

r

K R Using the b e the pro jective system of Bott elements with j

r r

formalism of norm compatible units developed in Sou one lets denote the

l

pro jective system

r d n

lim K R Zl Z N

n r r

r

r

Remark It is not clear to the authors whether the adic Soule elements dep end on

the choice of pro duct on the Mo ore sp ectrum By O pp the dierence

r

b etween the two regular pro ducts and on M is given by

r r

r

j j

i

r r r

M M S S M

Lemma B Let b e a ro ot of unity and n The restriction map from

Q Q n into H

l

cont

1

GalQ Q

l

1

Q n H Q

l l

cont

1

GalQ Q

l

n

A

r 1

lim Q H Q

r

Z l l l

l cont

l

r

1

GalQ Q

l

r

n

l

A

1 1

Q Q Q lim

r

Z l l l

l

l

r

is injective

Pro of Note that the argument given in the discussion preceding WiIV Theorem

is incorrect since the transition maps

n n

m

m

H Q H Q

r r

l

l

cont cont

l l

are in general not injective The kernel of the restriction map is given by

1

Q Q n Q H

l l

cont

Since Q Q is prime to l we have to show that

l

1

Q Z n Q H

l l l

cont

1

Q is isomorphic to Z and is torsion But the Galois group G of Q

l l l

hence its rst cohomology equals the functor of coinvariants Our claim follows since

n

Documenta Mathematica

Annette Huber Jorg Wildeshaus

Proposition B Let b e a xed dth ro ot of unity The l adic regulator

r K R Q H Q Q n

l n Z l d l

cont

b

takes to the cyclotomic element in continuous Galois cohomology

l

X

d n

A

r

l b

r

in the description of the last lemma dened by Soule and Deligne cf Sou page

D

Pro of If l is o dd then this is Sou Theoremes and For l the same is true

using the prop erties of the adic regulator see We

B Absolute Hodge Cohomology

Let B Sp ec C or B Sp ec R in this section

In A a denition of absolute Ho dge cohomology and relative cohomology for

general varieties over C was given The variant over R was A

By A resp A absolute Ho dge cohomology of smo oth varieties is given

p

B n functorially by Beilinsons complexes R

H

Lemma B These form a pseudoasque complex of presheaves on the Zariski site

of smo oth B schemes

U where U is a com Pro of By construction B they form a presheaf on pairs U

pactication with complement an NCdivisor For more details cf H Prop

U we get the desired presheaf To say it is pseudo Taking the limit over all choices of

asque means that absolute Ho dge cohomology has the MayerVietoris prop erty In

the context of A and A it is a formal consequence of the existence of triangles

i i id j j for op en immersions j with closed complement i In the context of B

it follows from the MayerVietoris prop erty of De Rhamcohomology and singular

cohomology

We now consider the corresp onding generalized cohomology

Definition B Version If X is a space over B then we dene absolute

Ho dge cohomology by

i i

f

p

H B n XB n H X K R

p

H

H sT

If f Z X is a morphism of spaces then we dene relative cohomology

i i

f

p

B n H X rel Z B n H Conef K R

p

H

H sT

Lemma B There is a functorial isomorphism b etween b oth denitions of absolute

Ho dge cohomology for a smo oth variety X If Y X is a closed immersion of smo oth

schemes then the same is true for relative cohomology

Documenta Mathematica

Classical Motivic Polylogarithm

Pro of Lemma B and Lemma Ba

In order to get the same equalities at least for some singular varieties we have

to check a descent prop erty for Ho dge mo dules For this we need functoriality of

i i with values in complexes of Ho dge mo dules rather than ob jects in the derived

category

Lemma B Let XC b e smo oth and i Y X a closed reduced subscheme of

S

n

pure co dimension Let Y Y For I f ng and M MHM X let

i F

i

i Y Y X

I I i

iI

j U X r Y X

I I i

iI

M j j M MHM X

I I F

I

All Y are equipp ed with the reduced structure Then i i M denes a functor

I I

I

b

fsubsets of f ngg C MHM X

F

Pro of As j is ane b oth j and j map Ho dge mo dules to such Note that lo cally

I I

I

each Y is given by a function f on X The functor i i has an explicit description

i i I

I

for closed subschemes of the type Y given in the pro of of S Prop In fact

I

M M

0 0

M M M i i M

I I I

I

0 0 0 0

I I jI j I I jI j

where the complex sits in degrees less or equal to zero

Proposition B Let XC b e smo oth and i Y X a closed subscheme as in

the lemma Let Y Y q q Y and

n

s

Y cosk Y Y Y

ie

Y k factors Y Y

k Y Y

Then the functor s s dened by the total complex of the cosimplicial complex

is isomorphic to i i s s

n nN

n

Pro of Note that

a

Y Y

k I

k

I fng

where Y Y in the notation of the previous lemma Let M b e in MHM X

I F

fi i g

k

By the previous lemma we get indeed a cosimplicial complex hence s s M is a well

dened complex of Ho dge mo dules Let Y b e the simplicial subscheme given by



a

s

k

Y Y Y

I

k

I i i i

k

Documenta Mathematica

Annette Huber Jorg Wildeshaus

By the Ho dge mo dule version of the combinatorial Lemma B the morphism

s s M s s M is a quasiisomorphism By denition S

i i M M M

fng

and this complex is canonically quasiisomorphic to the total complex of the constant

cosimplicial complex i i M It is easy to see that the natural morphism

Tot i i M s s M

is a quasiisomorphism

Corollary B Let XB b e smo oth Supp ose Y X is an NCdivisor over B

i

all of whose irreducible comp onents are smo oth over B Then the group H Y B j

p

H

as dened in A resp A is isomorphic to the generalized cohomology group

i

H Y B j and to the same noted group in B

p

H

Pro of The condition on Y ensures that Y is indeed a smo oth simplicial scheme It

gives rise to a space over B Cohomological descent for the co ecients as in B

implies cohomological descent for their global sections in the sense of B We can

use Y as the smo oth prop er hypercovering needed in Beilinsons denition Equality

to the generalized cohomology version is again B

This is of course cohomological descent for a closed Cechcovering We have

restricted to this case which is built into the very denition of Ho dge mo dules for

simplicity There is no reason why there should not b e cohomological descent in the

same generality as for constructible sheaves

Lemma B Let XB b e smo oth and Z X a closed immersion of an NC

divisor all of whose irreducible comp onents are smo oth over B Let Z b e the smo oth

simplicial scheme of B then there is a canonical isomorphism

i i

H X rel Z B n H X rel Z B n

p p

H H

where we use the original denition on the left and the second on the right

Pro of This follows by the general metho d of Bb from the descent prop erty that

we have just established

Remark If we had checked cohomological descent in general then we would get

B for arbitrary varieties and B for arbitrary closed immersions

Theorem B On the site of smo oth schemes over B the presheaves

p

B n have the prop erties of a twisted duality theory There are regu R

H

lator maps from K cohomology to absolute Ho dge cohomology

i i

H Y j H Y B j

p

M H

for all K coherent spaces Y They are compatible with pullback ie if f Y Y

is a map of K coherent spaces we get commutative diagrams



f

i i

H Y j H Y j

M M

c c

j j

y y



f

i i

H Y B j H Y B j

p p

H H

Documenta Mathematica

Classical Motivic Polylogarithm

If i Z X is a closed immersion of smo oth schemes constant co dimension d with

op en complement U and Y a space constructed form schemes over X as in B

then the regulator is compatible with pushout ie the diagram

i



nd

n

Y j H H Y Z j d

X

M

M

c c

j

j d

y y

i

nd

n

H H Y Z B j d Y B j

p

p

X

H

H

is commutative

Pro of We use Gillets metho d B All axioms of a twisted duality theory hold eg

H Ch Granted this the pro of pro ceeds as in the l adic case B

Remark Recall N that there is a natural transformation from absolute

Ho dge to Deligne cohomology The comp osition of the ab ove regulator with this

transformation was already constructed in Jeu

Theorem B Borel Let K b e a number eld with r real and r pairs of

complex embeddings into C We consider the ring of integers o as a scheme over Z

K

Then the Beilinson regulator

i i

H Sp ec o j R H Sp ec o R j

p

K Q K R

M H

is an isomorphism for all pairs i j It is injective of co dimension

r r for i j and injective of co dimension one in the case i j

Pro of Note that the cohomological dimension of the category of Ho dge structures is

The case i is trivial and the case is Dirichlets classical result In Bo

the claim and much more is proved for the Borel regulator instead of the Beilinson

regulator By Rp Corollary the two regulators coincide up to a nonvanishing

rational factor

B A Combinatorial Lemma

This section gives a purely combinatorial pro of why two conceivable denitions of the

Cechnerve of a covering are homotopically equivalent This is wellknown at least

for op en coverings and Cechcohomology and probably in general But for lack of

nding an appropriate reference we work out the combinatorics here

Let C n b e the following simplicial set

k

C n f ng

k

b e the simplicial subset of with the obvious face and degeneracy maps Let C n

simplices whose entries are ordered by In fact this is the simplicial version of the

nsimplex

Supp ose we are given a covariant functor from the category of subsets of

f ng to the category of sets We get simplicial sets by setting

An A

k I

I C n

k

An A

I



k

I C n

k

Documenta Mathematica

Annette Huber Jorg Wildeshaus

where A is the value of our functor on the set I fi i g Note that the elements

I k

of C n are ordered tuples but the value of A do es not dep end on the ordering

k I

Lemma B If the functor has constant value A then b oth simplicial sets have

the homotopy

A if i

An

i

else

Pro of Obviously it is enough to consider the case A ie of the simplicial sets

C n C n themselves Both simplicial sets satisfy the extension condition M

rather trivially Hence we can use the combinatorial computation of the homotopy

groups given in M Def We immediately get the result

Proposition B For a general functor A the injection An An of simpli

cial sets is a weak homotopy equivalence

i

Pro of We lter the simplicial sets C n by the simplicial subsets F C n of sim

plices in which at most i dierent integers o ccur This induces a ltration of the simpli

i i i

cial sets An Let G An b e the cobre of the cobration F An F An

It consists of simplices in which precisely i dierent integers o ccur We argue by

induction on i for all functors A at the same time There is a long exact homotopy

sequence attached to the cobration sequence

i i i

F An F An G An

i i

By induction it suces to show that all cobres G AnG An are weakly equiv

alent to the nal ob ject The cobre decomp oses into a union of simplicial sets

corresp onding to a dierent choice of i elements in f ng each If suces to

prove acyclicity for one choice eg for the subset f ig Hence we only have to

i i i i

consider G AiG Ai But this last cobre is isomorphic to G B iG B i

where B is the functor with constant value A For i it is easy to see that

fig

i i

G B iG B i By B the quotients B nB n are acyclic for all n Us

ing the same cobration sequence as for A and the inductive hypothesis this implies

i i

that all G B iG B i are acyclic

Note that A could also b e a functor to the category of ab elian groups or to the

dual of the category of ab elian groups

References

Ba H Bass Algebraic K theory Benjamin Inc

B AA Beilinson Notes on absolute Ho dge cohomology in Applications

of Algebraic K theory to Algebraic Geometry and Number Theory Pro

ceedings of a Summer Research Conference held June in

Boulder Colorado Contemp Math vol Part I AMS Providence

pp

B AA Beilinson Higher regulators and values of Lfunctions Jour So

viet Math pp

Documenta Mathematica

Classical Motivic Polylogarithm

B AA Beilinson On the derived category of p erverse sheaves in

YuI Manin ed K Theory Arithmetic and Geometry LNM

SpringerVerlag

B AA Beilinson Polylogarithm and Cyclotomic Elements typewritten

preprint MIT

BBD AA Beilinson J Bernstein P Deligne Faisceaux p ervers in B

Teissier JL Verdier Analyse et Topologie sur les Espaces singuliers

I Asterisque So c Math France

BD AA Beilinson P Deligne Motivic Polylogarithm and Zagier Conjec

ture preprint

BDp AA Beilinson P Deligne Polylogarithms and Regulators the very

preliminary version handwritten preprint

BD AA Beilinson P Deligne Interpretation motivique de la conjecture

de Zagier reliant p olylogarithmes et regulateurs in U Jannsen SL

Kleiman JP Serre eds Motives Pro ceedings of the Research Con

ference on Motives held July August in Seattle Washington

Pro c of Symp in Pure Math Part I I AMS pp

BL AA Beilinson A Levin The Elliptic Polylogarithm in U Jannsen

SL Kleiman JP Serre Motives Pro ceedings of the Research Con

ference on Motives held July August in Seattle Washington

Pro c of Symp in Pure Math Part I I AMS pp

BLp AA Beilinson A Levin Elliptic Polylogarithm typewritten preprint

MIT

Bl S Blo ch Application of the dilogarithm function in algebraic K theory

and algebraic geometry in Pro c Int Symp Alg Geometry Kyoto

pp

BlK S Blo ch K Kato Lfunctions and Tamagawa Numbers of Motives

in P Cartier et al eds The Grothendieck Festschrift Volume I

Birkhauser pp

Bo A Borel Stable real cohomology of arithmetic groups Ann Sc Ec

Norm Sup pp

Bo A Borel Cohomologie de SL et valeurs de fonctions zeta Ann Scuola

n

Normale Sup eriore di Pisa pp

Bo A Borel et al Algebraic D mo dules Perspectives in Mathematics

Academic Press

BouK AK Bouseld DM Kan Homotopy Limits Completions and Lo cal

izations LNM Springer

Br KS BrownAbstract homotopy theory and generalized sheaf cohomolo

gy Trans AMS pp

BrG KS Brown SM Gersten Algebraic K theory as generalized sheaf co

homology in Alg K Theory I LNM pp Springer

Brw W Browder Algebraic K theory with co ecients Zp in MG Bar

ratt ME Mahowald eds Geometric Applications of Homotopy The

ory I Pro ceedings Evanston LNM SpringerVerlag pp

Documenta Mathematica

Annette Huber Jorg Wildeshaus

D P Deligne Equations Dierentielles a Points Singuliers Reguliers LNM

SpringerVerlag

D P Deligne Theorie de Ho dge I I Publ Math IHES pp

D P Deligne Theorie de Ho dge I I I Publ Math IHES pp

D P Deligne La Conjecture de Weil I I Publ Math IHES pp

D P Deligne Le Group e Fondamental de la Droite Pro jective Moins Trois

Points in Y Ihara K Rib et JP Serre Galois Groups over Q Pro

ceedings of a Workshop held March at the MSRI Berkeley

California Springer pp

Den C Deninger Higher regulators and Hecke Lseries of imaginary quadratic

elds I I Ann of Math pp

DwF WG Dwyer EM Friedlander Algebraic etale K theory Trans Am

Math So c no pp

E H Esnault On the Lo day Symbol in the DeligneBeilinson Cohomolo

gy K Theory pp

EGAI I A Grothendieck J Dieudonne Elements de geometrie algebrique I I

Publ Math IHES

Ek T Ekedahl On the Adic Formalism in P Cartier et al e ds The

Grothendieck Festschrift Volume I I Birkhauser pp

GH J Gamst K Ho echsmann Pro ducts in sheaf cohomology Tohoku

Math J pp

G H Gillet RiemannRo ch theorems for higher algebraic K theory Adv

in Math pp

GSo H Gillet C Soule Filtrations on higher algebraic K theory unpub

lished revised version in preparation

GSo H Gillet C Soule Descent Motives and K Theory Preprint

Go eJr PG Go erss JF Jardine Lo calization theories for simplicial

presheaves preprint

Go AB Goncharov Polylogarithms and Motivic Galois Group in U

Jannsen SL Kleiman JP Serre Motives Pro ceedings of the Re

search Conference on Motives held July August in Seattle

Washington Pro c of Symp in Pure Math Part I I AMS pp

Gr M Gros Regulateurs syntomiques et valeurs de fonctions L padiques

I I Invent Math pp

HZ RM Hain S Zucker Unip otent variations of mixed Ho dge structure

Inv math pp

Ha R Hartshorne Algebraic Geometry Graduate Texts in Mathematics

Springer Verlag

H A Hub er Mixed Motives and Their Realization in Derived Categories

LNM Springer

H A Hub er Mixed Perverse Sheaves for Schemes over Number Fields

Comp Math pp

Documenta Mathematica

Classical Motivic Polylogarithm

HW A Hub er J Wildeshaus The Classical Polylogarithm Abstract of a

series of lectures given at the workshop on p olylogs in Essen May

available under httpwwwmathuiuceduKtheory

Jeu R de Jeu Zagiers Conjecture and Wedge Complexes in Algebraic K

theory Comp Math pp

Jn U Jannsen Continuous etale cohomology Math Ann pp

Jn U Jannsen On the ladic Cohomology of Varieties over Number Fields

and its Galois Cohomology in Y Ihara K Rib et JP Serre Galois

Groups over Q Pro ceedings of a Workshop held March at

the MSRI Berkeley California Springer pp

Jn U Jannsen Mixed Motives and Algebraic K Theory LNM

SpringerVerlag

Jr JF Jardine Simplicial ob jects in a Grothendieck top os in Applica

tions of Algebraic K theory to Algebraic Geometry and Number Theory

Pro ceedings of a Summer Research Conference held June in

Boulder Colorado Contemp Math vol Part I AMS Providence

pp

Jr JF Jardine Simplicial presheaves Journal of Pure and Appl Algebra

pp

Jr JF Jardine Stable homotopy theory of simplicial presheaves Can J

Math pp

Jr JF Jardine Generalized Etale Cohomology Theories Progress in

Mathematics BirkhauserVerlag

Kn D Kan On c s s complexes Am Journ Math pp

KNF M Kolster T Nguyen Quang Do V Fleckinger Twisted S units padic

class number formulas and the Lichtenbaum conjectures Duke Math J

pp

Kr C Kratzer strucure en K theorie algebrique Comm Math Helv

no pp

Ks M Kashiwara A study of variation of mixed Ho dge structure Publ

RIMS Kyoto Univ pp

Le M Levine Lambdaop erations K theory and motivic cohomology in

VP Snaith ed Algebraic K theory pap ers from the nd Great Lakes

conference Canada March in memory of RW Thomason Provi

dence RI AMS Fields Inst Comm pp

Lo JL Lo day K theorie algebrique et representations de group es Ann

scient Ec Norm Sup e serie t pp

M P May Simplicial ob jects in algebraic top ology Van Nostrand

Mi JS Milne Arithmetic Duality theorems Academic Press

N J Nekovar Beilinsons Conjectures in U Jannsen SL Kleiman JP

Serre Motives Pro ceedings of the Research Conference on Motives held

July August in Seattle Washington Pro c of Symp in Pure

Math Part I AMS pp

Documenta Mathematica

Annette Huber Jorg Wildeshaus

Neu J Neukirch The Beilinson Conjecture for Algebraic Number Fields in

M Rap op ort N Schappacher P Schneider eds Beilinsons Conjec

tures on Sp ecial Values of LFunctions Perspectives in Mathematics

Academic Press pp

O S Oka Multiplications on the Mo ore Sp ectrum Memoirs of the Fac of

Sciences Kyushu Univ Ser A Vol No pp

Q D Quillen Homotopical Algebra LNM Springer

Q D Quillen Higher Algebraic K Theory in Alg K Theory I LNM

Springer pp

Rm D Ramakrishnan Regulators algebraic cycles and values of L

functions in M R Stein R Keith Dennis eds Algebraic K Theory

and Algebraic Number Theory Pro c of a Seminar held at the EastWest

Center in Honolulu Hawaii on Jan Cont Math

pp

Rp M Rap op ort Comparison of the regulators of Beilinson and of Borel

in M Rap op ort N Schappacher P Schneider eds Beilinsons Con

jectures on Sp ecial Values of LFunctions Perspectives in Mathematics

Academic Press pp

S Morihiko Saito Mo dules de Ho dge Polarisables Publ RIMS Kyoto

Univ pp

S Morihiko Saito Mixed Ho dge Mo dules Publ RIMS Kyoto Univ

pp

S Morihiko Saito Ho dge Conjecture and Mixed Motives I in JA Carl

son CH Clemens DR Morrison Complex Geometry and Lie Theory

Pro c of Symp in Pure Math AMS pp

Sch AJ Scholl Height pairings and Sp ecial Values of Lfunctions in U

Jannsen SL Kleiman JP Serre Motives Pro ceedings of the Re

search Conference on Motives held July August in Seattle

Washington Pro c of Symp in Pure Math Part I AMS pp

SGA A Grothendieck et al Revetements Etales et Group e Fondamental

LNM SpringerVerlag

SGAI I M Artin A Grothendieck JL Verdier et al Theorie des Topos et

Cohomologie Etale des Schemas Tome LNM SpringerVerlag

SGAI I I M Artin A Grothendieck JL Verdier et al Theorie des Topos et

Cohomologie Etale des Schemas Tome LNM SpringerVerlag

Sou C Soule K theorie des anneaux dentiers de corps de nombres et coho

mologie etale Inv Math pp

Sou C Soule On higher padic regulators in E M Friedlander M R

Stein eds Algebraic K theory Pro ceedings Evanston LNM

SpringerVerlag

Sou C Soule The rank of etale cohomology of varieties over padic and

number elds Comp Math pp

Documenta Mathematica

Classical Motivic Polylogarithm

Sou C Soule Op erations en K theorie algebrique Can Jour Math

no pp

Sou C Soule Elements cyclotomiques en K theorie Asterisque

pp

T G Tamme The Theorem of RiemannRo ch in M Rap op ort N Schap

pacher P Schneider eds Beilinsons Conjectures on Sp ecial Values of

LFunctions Perspectives in Mathematics Academic Press pp

We C Weibel Etale Chern classes at the prime in PG Go erss and JF

Jardine eds Algebraic K Theory and Algebraic Topology Pro c Lake

Louise NATO Adv Sci Inst Ser C Math Phys Sci Kluwer

Acad Publ pp

Wi J Wildeshaus Realizations of Polylogarithms LNM Springer

Verlag

WiI J Wildeshaus Mixed structures on fundamental groups in Wi pp

WiI I I J Wildeshaus Polylogarithmic Extensions on Mixed Shimura varieties

Part I Construction and basic prop erties in Wi pp

WiIV J Wildeshaus Polylogarithmic Extensions on Mixed Shimura varieties

Part I I The classical p olylogarithm in Wi pp

Annette Hub er Jorg Wildeshaus

Math Institut Math Institut

Einsteinstr Einsteinstr

M unster M unster

Germany Germany

hubermathunimuensterde wildeshmathunimuensterde

Documenta Mathematica

Documenta Mathematica