1702 00047

WEIGHT DECOMPOSITION OF DE RHAM COHOMOLOGY SHEAVES
AND TROPICAL CYCLE CLASSES FOR NON-ARCHIMEDEAN SPACES
YIFENG LIU
Abstract. We construct a functorial decomposition of de Rham cohomology sheaves,

arXiv:1702.00047v1 [math.AG] 31 Jan 2017
called weight decomposition, for smooth analytic spaces over non-Archimedean fields embed-
dable into Cp , which generalizes a construction of Berkovich and solves a question raised
by him. We then investigate complexes of real tropical differential forms and currents
introduced by Chambert-Loir and Ducros, by establishing a relation with the weight de-
composition and defining tropical cycle maps with values in the corresponding Dolbeault
cohomology. As an application, we show that algebraic cycles that are cohomologically
trivial in the algebraic de Rham cohomology are cohomologically trivial in the Dolbeault
cohomology of currents as well.
Contents
1. Introduction 1
2. Review of rigid cohomology 6
3. Weight decomposition in tale topology 8
4. Logarithmic dierential forms 19
5. Tropical cycle class map 24
6. Cohomological triviality 33
References 41
1. Introduction
Let K be a complete non-Archimedean eld of characteristic zero with a nontrivial valu-
ation. Let X be a smooth K-analytic space in the sense of Berkovich. Let OX (resp. cX ) be
the structure sheaf (resp. the sheaf of constant analytic functions [Ber04, 8]) of X in either
analytic or tale topology. We have the following complex of cX -modules in either analytic
or tale topology:
d d d
(1.1) X : 0 OX = 0X
1X
2X
,
known as the de Rham complex, which satises that cX = ker(d : OX 1X ). It is not
exact from the term 1X if dim(X) 1. The cohomology sheaves of the de Rham complex
q,cl
X /dX
q1
are called de Rham cohomology sheaves. For q 0, denote by qX the subsheaf
of q,cl
X /dX
q1
generated by sections of the form
X dfi1 dfiq
ci
fi1 fiq
Date: February 2, 2017.
2010 Mathematics Subject Classification. 14G22.
1
2 YIFENG LIU

where the sum is nite, ci are sections of cX , and fij are sections of OX . In particular, we
have X = cX , and that X is simply the sheaf X dened in [Ber07, 4.3] in the case of
0 1
tale topology.
Theorem 1.1. Suppose that K is embeddable into Cp . Let X be a smooth K-analytic space.
Then for every q 0, we have a decomposition
M
q,cl q1
X /dX = (q,cl q1
X /dX )w
wZ
of cX -modules in either analytic or tale topology. It satisfies that

(i) (q,cl q1
X /dX )w = 0 unless q w 2q;
(ii) qX (q,cl q1
X /dX )2q , and they are equal if q = 1;
1,cl
(iii) (X /dOX )1 coincides with the sheaf X defined in [Ber07, 4.5] in the case of tale
topology.
Such decomposition is stable under base change, cup product, and functorial in X.
The proof of this theorem will be given at the end of Section 4. We call the decomposition
in the above theorem the weight decomposition of de Rham cohomology sheaves.
Corollary 1.2. Suppose that K is embeddable into Cp . Then for every smooth K-analytic
space X, we have 1,cl
X /dOX = X X in tale topology. This answers the question in
[Ber07, Remark 4.5.5] for such K.
Remark 1.3. We expect that Theorem 1.1 and thus Corollary 1.2 hold by only requiring that
the residue eld of K is algebraic over a nite eld (and K is of characteristic zero).
For the rest of Introduction, we work in the analytic topology only. In particular, the
de Rham complex (X , d) is a complex of sheaves on (the underlying topological space of) X.
In [CLD12], Chambert-Loir and Ducros dene, for every K-analytic space X, a bicomplex
(AX, , d , d ) of sheaves of real vector spaces on X concentrated in the rst quadrant. It
is a non-Archimedean analogue of the bicomplex of (p, q)-forms on complex manifolds. In
particular, we may dene analogously the Dolbeault cohomology (of forms) as

ker(d : AXq,q (X) AXq,q +1 (X))
HAq,q (X) := .
im(d : AXq,q 1 (X) AXq,q (X))
By [CLD12] and [Jel16], we know that for every q 0, the complex (AXq, , d ) is a ne

resolution of the sheaf ker(d : AXq,0 AXq,1 ). Thus, HAq,q (X) is canonically isomorphic to

the sheaf cohomology H q (X, ker(d : AXq,0 AXq,1 )). If X is of dimension n and without
boundary, then we may dene the integration
Z

X
for every top form AXn,n (X) with compact support. In particular, if X is moreover
compact, then the integration induces a real linear functional on HAn,n (X).
The next theorem reveals a connection between ker(d : AXq,0 AXq,1 ) and the algebraic
de Rham cohomology sheaves of X.
WEIGHT DECOMPOSITION AND TROPICAL CYCLE CLASSES 3
Theorem 1.4 (Lemma 6.1, Theorem 5.10). Let K be a non-Archimedean field embeddable
into Cp and X a smooth K-analytic space. Let LXq be the subsheaf of Q-vector spaces of
q,cl
X /dX
q1
generated by sections of the form df
f1
1
df
fq
q
where fj are sections of OX .
Then
(1) the canonical map LXq Q cX qX is an isomorphism;
(2) there is a canonical isomorphism LXq Q R ker(d : AXq,0 AXq,1 ).

The above theorem implies that the Dolbeault cohomology HAq,q (X) for X in the theorem

has a canonical rational structure through the isomorphism HAq,q (X) H q (X, LXq ) Q R.
Recall that in the complex world, for a smooth complex algebraic variety X , we have a cycle
class map from CHq (X ) to the classical Dolbeault cohomology Hq,q (X an ) of the associated
complex manifold X an . Over a non-Archimedean eld K, we may associate a scheme X of
nite type over K a K-analytic space X an . The following theorem is an analogue of the
above cycle class map in the non-Archimedean world.
Theorem 1.5 (Denition 5.7, Theorem 5.8, Corollary 5.9). Let K be a non-Archimedean
field and X a smooth scheme over K of dimension n. Then there is a tropical cycle class
map
clA : CHq (X ) HAq,q (X an ),
functorial in X and K, such that for every algebraic cycle Z of X of codimension q,
Z Z
(1.2) clA (Z) =
X an Z an
for every d -closed form AXnq,nq

an (X an ) with compact support.
In particular, if X is proper and Z is of dimension 0, then
Z
clA (Z) = deg Z.
X an
Remark 1.6. Let the situation be as in the above theorem.

(1) Our construction actually shows that the image of clA is in the canonical rational
subspace H q (X an , LXq an ).
(2) The tropical cycle class respects products on both sides. More precisely, for Zi
CHqi (X ) with i = 1, 2, we have clA (Z1 Z2 ) = clA (Z1 ) clA (Z2 ).
(3) We may regard the formula (1.2) as a tropical version of Cauchy formula in multi-
variable complex analysis.
(4) Even when X is proper, one can not use (1.2) to dene clA (Z) as we do not know
whether the pairing
R

HAq,q (X an ) HAnq,nq (X an )
HAn,n (X an ) X
R
is perfect or not at this moment.
For a proper smooth scheme X of dimension n over a general eld K of characteristic
2q
zero, we have a cycle class map cldR : CHq (X ) HdR (X ) with values in the algebraic de
Rham cohomology. It is known that when K = C, the kernel of cldR coincides with the
kernelR of the cycle class map valued in Dolbeault cohomology. In particular, if cldR (Z) = 0,
then Z an = 0 for every -closed (n q, n q)-form on X an . In the following theorem,
4 YIFENG LIU
we prove that the same conclusion holds in the non-Archimedean setting as well, with mild
restriction on the eld K.
Theorem 1.7. (Theorem 6.6) Let K Cp be a finite extension of Qp and X a proper
smooth scheme over K of dimension n. Let Z be an algebraic cycle of X of codimension q
such that cldR (Z) = 0. Then
Z
=0
(ZK Cp )an
for every d -closed form A nq,nq ((X K Cp )an ).

We emphasize again that in the above theorem, we do not know whether clA (Z) = 0 or
not. If we know the Poincar duality for HA, (X an ), then clA (Z) = 0. Nevertheless, we have
the following result for lower degree.
Theorem 1.8. (Theorem 6.2) Let X be a proper smooth scheme over Cp . Then
(1) HA1,1(X an ) is finite dimensional;
(2) for a line bundle L on X such that cldR (L) = 0, we have clA (L) = 0.
To the best of our knowledge, the rst conclusion in the above theorem is the only known

case of the niteness of dim HAq,q (X an ) when both q, q are positive and X is of general
dimension. Note that in the above theorem, we do not require that X can be dened over a
nite extension of Qp .
Remark 1.9. We can interpret Theorem 1.7 in the following way. Let k be a number eld.
Let X be a proper smooth scheme over k of dimension n, and Z an algebraic cycle of X of
codimension q. Suppose that there exists one embedding : k C such that
Z
=0
(Zk, C)an
for every -closed (n q, n q)-form on (X k, C)an . Then for every prime p and every
embedding p : k Cp , we have
Z
=0
(Zk,p Cp )an
for every d -closed (n q, n q)-form on (Z k,p Cp )an .
The article is organized as follows. We review the basic theory of rigid cohomology in
Section 2, which is one of the main tools in our work. We construct the weight decomposition
of de Rham cohomology sheaves in the tale topology in Section 3. In Section 4, we study
the behavior of logarithmic dierential forms in rigid cohomology and deduce Theorem 1.1
for both topologies. We will not use tale topology after this point. We start Section 5 by
reviewing the theory of real forms developed by Chambert-Loir and Ducros; and then we
study its relation with de Rham cohomology sheaves. Next, we dene the tropical cycle
class maps and establish their relation with integration of real forms. In the last Section 6,
we study algebraic cycles that are cohomologically trivial in the sense of algebraic de Rham
cohomology. In particular, we show that they are cohomologically trivial in the sense of
Dolbeault cohomology of currents (of forms if they are of codimension 1).
Conventions and Notation.

Throughout the article, by a non-Archimedean field we mean a complete topological
eld of characteristic zero whose topology is induced by a nontrivial non-Archimedean
valuation | | of rank 1. If the valuation is discrete, then we say that it is a discrete
non-Archimedean field by abuse of terminology.
Let K be a non-Archimedean eld. Put
K = {x K | |x| 1}, K = {x K | |x| < 1}, f = K /K .
K
Denote by K a the algebraic closure of K and K da its completion. A residually algebraic

extension of K is an extension K /K of non-Archimedean elds such that the induced
extension K f /K
f is algebraic. In the text, discrete non-Archimedean elds are usually
denoted by lower-case letters like k, k , etc. And will always be a uniformizer of a
discrete non-Archimedean eld, though we will still remind readers of this.
Let K be a non-Archimedean eld, and A an anoid K-algebra. We then have
the K-analytic space M(A). Denote by A the subring of power-bounded elements,
which is a K -algebra. Put Ae = A K K. f We say that A is integrally smooth if A
is strictly K-anoid and Spf A is a smooth formal K -scheme.

Let K be a non-Archimedean eld. For a real number r > 0, we denote by D(0; r)

the open disc over K with center at zero of radius r. For real numbers R > r > 0,
we denote by B(0; r, R) the open annulus over K with center at zero of inner radius
r and outer radius R. An open poly-disc (of dimension n) over K is the product of
nitely many open discs D(0; ri) (of number n).
For a non-Archimedean eld K, all K-analytic (Berkovich) spaces are assumed to be
Hausdor and strictly K-analytic [Ber93, 1.2.15]. Suppose that K /K is an extension
of non-Archimedean elds. For a K-analytic space X and a K -analytic space Y , we
put
b K K = X M(K) M(K ),
X b K K );
Y K X = Y M(K ) (X
and for a formal K -scheme X and a formal K -scheme Y, we put
b K K = X Spf K Spf K ,
X b K K ).
Y K X = Y Spf K (X
If k is a discrete non-Archimedean eld and X is a special formal k -scheme in the
sense of [Ber96], then we have the notion X , the generic ber of X, which is a k-
analytic space; and Xs , the closed ber of X, which is a scheme locally of nite type
e and a reduction map : X X . For a general non-Archimedean eld K,
over k; s
we say a formal K -scheme X is special if there exist a discrete non-Archimedean
eld k K and a special formal k -scheme X such that X X b k K . For a special

formal K -scheme, we have similar notion : X Xs which is canonically dened.
In this article, all formal K -schemes will be special. Note that if Z is a subscheme
of Xs , then 1 Z is usually denoted as ]Z[X in rigid analytic geometry.
If X is a scheme over an ane scheme Spec A and B is an A-algebra, then we
put XB = X Spec A Spec B. Such abbreviation will be applied only to schemes,
neither formal schemes nor analytic spaces. If X is a scheme over Spec K for a
non-Archimedean eld K, then we write Xs for XKe .
Let K be a non-Archimedean eld and X a K-analytic space. For a point x X,
one may associate nonnegative integers sK (x), tK (x) as in [Ber90, 9.1]. For readers
convenience, we recall the denition. The number sK (x) is equal to the transcendence
6 YIFENG LIU
] over K,
degree of H(x) f and the number t (x) is equal to to the dimension of the
K
q q
Q-vector space |H(x) |/ |K |, where H(x) is the completed residue eld of x.

In the text, the eld K will always be clear so will be suppressed in the notation
sK (x), tK (x).
Let X be a site. Whenever we have a suitable notion of de Rham complex (X , d)
on X, we denote by HdR
(X) := H (X, X ) the corresponding de Rham cohomology
of X, as the hypercohomology of the de Rham complex.
Acknowledgements. The author is partially supported by NSF grant DMS1602149. He
thanks Weizhe Zheng for helpful discussions.
2. Review of rigid cohomology

In this section, we review the theory of rigid cohomology developed in, for example,
[Bert97] and [LS07].
Let R be the category of triples (K, X, Z) where K is a non-Archimedean eld; X is a
scheme of nite type over K; f and Z is a Zariski closed subset of X. A morphism from
(K , X , Z ) to (K, X, Z) consists of a eld extension K /K and a morphism X X Ke K f
whose restriction to Z factors through Z Ke K f . Let V be the category of pairs (K, V ) where
K is a non-Archimedean eld and V is a graded K-vector space. A morphism from (K, V )
to (K , V ) consists of a eld extension K /K and a graded linear map V K K V .
We have a functor of rigid cohomology with support: Ropp V sending (K, X, Z) to

HZ,rig(X/K). Put Hrig (X/K) = HX,rig (X/K) for simplicity. We list the following properties
which will be used extensively in this article:
Suppose that we have a morphism (K , X , Z ) (K, X, Z) with X X Ke K f
and Z Z Ke K f . Then the induced map H (X/K) K H (X /K ) is

Z,rig K Z ,rig
an isomorphism of nite dimensional graded K -vector spaces ([GK02, Corollary 3.8]
and [Ber07, Corollary 5.5.2]).
For Y = X\Z, we have a long exact sequence:
i i i i+1
(2.1) HZ,rig (X/K) Hrig (X/K) Hrig (Y /K) HZ,rig (X/K) .
If both X, Z are smooth, and Z is of codimension r in X, then we have a Gysin
i2r
isomorphism HZ,rig
i
(X/K) Hrig (Z/K).
f is a nite extension
Suppose that K is residually algebraic over Qp (in other words, K
of Fp ). Then the sequence (2.1) is equipped with a Frobenius action of suciently
large degree. In particular, each item V in (2.1) admits a direct sum decomposition
L
V = wZ Vw where Vw consists of vectors of generalized weight w ([Chi98, 1 &
2]).
i
Suppose that X is smooth and Z is of codimension r, then HZ,rig (X/K)w = 0 unless
i w 2(i r) ([Chi98, Theorem 2.3]).
We will extensively use the notion of K-analytic germs ([Ber07, 5.1]), rather than K-
dagger spaces. Roughly speaking, a K-analytic germ is a pair (X, S) where X is a K-analytic
space and S X is a subset. We say that (X, S) is a strictly K-anoid germ if S is a strictly
anoid domain. We say that (X, S) is smooth if X is smooth in an open neighborhood of
S. We have the structure sheaf O(X,S) , and the de Rham complex (X,S) when (X, S) is
smooth. (See [Ber07, 5.2] for details.) In particular, we have the de Rham cohomology

HdR (X, S) when (X, S) is smooth. For a smooth K-analytic germ (X, S) where S = M(A)
for an integrally smooth K-anoid algebra A, we have a canonical functorial isomorphism

HdR
(X, S) Hrig e
(Spec A/K) (see [Bert97, Proposition 1.10], whose proof actually works for
general K).
The following lemma generalizes the construction in [GK02, Lemma 2].
Lemma 2.1. Let (X1 , Y1 ) and (X2 , Y2) be two smooth strictly K-affinoid germs. Then for
a morphism : Y2 Y1 of strictly K-affinoid domains, there is a canonical restriction map
: HdR

(X1 , Y1) HdR (X2 , Y2 ). It satisfies the following conditions:
(i) if extends to a morphism (X2 , Y2) (X1 , Y1) of germs, then coincides with the
usual pullback;
(ii) for a finite extension K of K, if we write Xi (resp. Yi ) for Xi b K K (resp. Yi
b K K )
for i = 1, 2 and for b K K , then coincides with the scalar extension of , in
which we identify HdR (Xi , Yi ) with HdR

(Xi , Yi ) K K for i = 1, 2;
(iii) if Y1 = M(A1 ) and Y2 = M(A2 ) for some integrally smooth K-affinoid algebras
A1 and A2 , then coincides with e : Hrig f /K) H (Spec A
(Spec A 1 rig
f /K) un-
2
f
der the canonical isomorphism HdR (Xi , Yi ) Hrig (Spec Ai /K) for i = 1, 2, where
e : Spec A
f Spec A
2
f is the induced morphism;
1
(iv) if (X3 , Y3 ) is another smooth strictly K-affinoid germ with a morphism : Y3 Y2 ,
then ( ) = .
Proof. Put X = X1 K X2 , Y = Y1 K Y2 , and Y the graph of , which is isomorphic to
Y2 via the projection to the second factor. Denote by ai : X Xi the projection morphism.
We have maps
a a
HdR (X1 , Y1)
1
lim HdR (V )
2
HdR (X2 , Y2),

V
where V runs through open neighborhoods of in X. We show that a2 is an isomorphism.

Then we dene as (a2 )1 a1 .
The proof is similar to that of [GK02, Lemma 2]. To show that a2 is an isomorphism is a
local problem. Thus we may assume that there are elements t1 , . . . , tm OX1 (X1 ) such that
dt1 , . . . , dtm form a basis of 1 (X1 , Y1 ) over O(X1 , Y1 ), and there exist a strictly K-anoid
neighborhood U X of with an element |K |, and an isomorphism

M(Kh1 Z1 , . . . 1 Zm i) K ,
U Y
in which 1 Zi is sent to 1 (ti 1 1 (ti )). Note that Kh1 Z1 , . . . 1 Zm i is an
integrally smooth K-anoid algebra, and Spec Kh1 Z^ 1 , . . . Zm i is canonically isomorphic
1
m
to AKe . Thus by [GK02, Lemma 2], the restriction map HdR (X2 , Y2) HdR (X, U Y ) is an
T
isomorphism. We may choose a sequence of such U with U = . Then lim HdR (X, U

Y ) limV HdR (V ) and thus a2 is an isomorphism.

Properties (i) and (ii) follow easily from the construction. Property (iv) is straightforward
but tedious to check; we will leave it to readers. We now check Property (iii), as it is
important for our later argument. The induced projection morphism
M(Kh1 Z1 , . . . 1 Zm i) K U Y Yi
8 YIFENG LIU
extends canonically to a morphism of formal K -schemes

Spf(Kh1 Z1 , . . . 1 Zm i
b K A ) Spf A ,
i
where A is the coordinate K-anoid algebra of which is isomorphic to A2 . Therefore,

the restriction map HdR (Xi , Yi ) HdR (X, U Y ) coincides with the map
aei : Hrig
f /K) H (Spec Kh1 Z , . ^
(Spec A 1 b
i rig 1 . . Zm iK A /K)
induced from the homomorphism A f Kh1 Z , . ^ f

1 . . Zm iK A of K-algebras. Note that
i
1 b
1
af2 is an isomorphism, and that (af2 ) coincides with the restriction map

Hrig (Spec Kh1 Z1 , . ^
. . 1 Zm i rig
f /K)
b K A /K) H (Spec A2
induced from the homomorphism Kh1 Z1 , . ^

. . 1 Zm i f sending 1 Z to 0 for
b K A A2 i
all i. Property (iii) follows.
The following example will be used in the computation later.
Example 2.2. Let K be a non-Archimedean eld. For an integer t 0 and an element
K, dene the formal K -scheme
Et = Spf K [[T0 , . . . , Tt ]]/(T0 Tt )
and let Et be its generic ber. Let E
t
be the K-anoid algebra
1 1 1 1
Kh|| t+1 T0 , , || t+1 Tt , || t+1 T01 , , || t+1 Tt1 i/(T0 Tt ),
which is integrally smooth. Moreover, M(E
t
) is canonically a strictly K-anoid domain in
E , and the restriction map
t

HdR
(Et ) HdR (Et , M(E
t
)) Hrig g
(Spec E t
/K)
is an isomorphism by [GK02, Lemma 3]. If K is residually algebraic over Qp , then we have

q g
(Spec E q g
/K) = Hrig (Spec E /K)2q .
Hrig t t
3. Weight decomposition in tale topology

In this section, we construct the weight decomposition of de Rham cohomology sheaves in
the tale topology. Therefore, in this section, sheaves like OX , cX , and the de Rham complex
(X , d) are understood in the tale topology.
Definition 3.1 (Marked pair). Let k be a discrete non-Archimedean eld.
(1) We say that a scheme X over k is strictly semi-stable of dimension n if X is locally
of nite presentation, Zariski locally tale over Spec K [T0 , . . . , Tn ]/(T0 Tt ) for
some uniformizer of k, and Xk is smooth over k. For every 0 t n, denote by
Xs[t] the union of intersection of t + 1 distinct irreducible components of Xs . It is a
closed subscheme of Xs with each irreducible component smooth.
(2) A marked k-pair (X , D) of dimension n and depth t consists of an ane strictly semi-
stable scheme X over k of dimension n, and an irreducible component D of Xs[t] that
is geometrically irreducible.
We start from the following lemma, which generalizes [Ber07, Lemma 2.1.2].
Lemma 3.2. Suppose that K is embeddable into kca for some discrete non-Archimedean field
k. Let X be a smooth K-analytic space, and x a point of X with s(x) + t(x) = dimx (X).
Given a morphism of strictly K-analytic spaces X Y , where Y is a special formal K -
scheme, there exist
a finite extension K of K, a finite extension k of k contained in K ,
a marked k -pair (X , D) of dimension dimx (X) and depth t(x),
an open neighborhood U of (X c ) b an
/D k K in XK ,
a point x (Xc ) b
/D k K ,
a morphism of K-analytic spaces : U X, and
a morphism of formal K -schemes X c b
/D k K Y,
such that the following are true:
(i) is tale and (x ) = x;
(ii) the induced morphism (X c ) b
/D k K Y coincides with the composition
c )
(X b
/D k K U
X Y .
Proof. Put t = t(x), s = s(x), and n = t + s. By [Ber07, Proposition 2.3.1], by possibly
taking nite extensions of k (and K), we may replace X by (B k Y ) b k K, where B =
Qt
j=1 B(0; rj , Rj ) for some 0 < rj < Rj and Y is a smooth k-analytic space of dimension s,
and x projects to b B with t(b) = t and y Y with s(y) = s. Denote by P the k -scheme
P1k with the point 0 on the special ber blown up, and by P the formal completion of P
along the open subscheme Ps \{(0), ()}, which is isomorphic to Spf k hX, Y i/(XY )
for some uniformizer of k. By taking further nite extensions of k (and K), we may
Q
assume that there is an Q open immersion t P B containing b such that (b) = 0, where
0 is the closed point in t Ps that is nodal in every component.
For Y , we proceed exactly as in the Step 1 of the proof of [Ber07, Lemma 2.1.2]. We obtain
two strictly k-anoid domains Z W Y . As in the beginning of Step 3 of the proof
of [Ber07, Lemma 2.1.2], we also get an integral scheme Y proper and at over k with an
embedding Y Yan , open subschemes Z W Ys , such that Z = 1 Z , W = 1 W .
Q
Now we put two parts together. Dene Y = t P Y where the ber product is taken
Q e Then W := Q P W
over k , and W = t Ps W where the ber product is taken over k. t
coincides with 1 W in Ykan . Moreover, WK is an open neighborhood of x where WK denotes
S
the inverse image of W in X = (B k Y ) b k K. Write W = l Wi as in Step 2 of the proof
i=1
of [Ber07, Lemma 2.1.2]. By [Ber07, Lemma 2.1.3 (ii)], we may assume that Wi are all
k-anoid by taking nite extensions of k (and K). Making a nite number of additional
blow-ups, we may also assume that there are open subschemes Wi W with Wi = 1 Wi
S
and W = li=1 Wi .
Now we proceed as in Step 4 of the proof of [Ber07, Lemma 2.1.2]. Take an alteration
: X Y after further nite extensions of k (and K), and a point x XKan such that
(x ) = x. By a similar argument, one can show that (x ) Xs ek K f has dimension at
least s. On the other hand, we have s(x ) s and t(x ) t. Thus, s(x ) = s and t(x ) = t.

Denote by C the Zariski closure of (x ) in Xs , equipped with the reduced induced scheme
structure. Suppose that C is contained in t distinct irreducible components of Xs . Then
t t. We take an open subscheme U of X satisfying: D := U C is open dense in C; (D )
is contained in W; U is tale over Spec k [T0 , . . . , Tn ]/(T0 Tt ) for some uniformizer
of k such that D is the zero locus of the ideal generated by (T0 , . . . , Tt , ). Now we
10 YIFENG LIU
blow up the closed ideal generated by (Tt +1 , ), and then the strict transform of the closed
ideal generated by (Tt +2 , ), and continue to obtain an ane strictly semi-stable scheme
X over k such that the strict transform D of D is an irreducible component of Xs[t] . After
further nite extensions of k (and K) and replacing X by an ane open subscheme such
that Xs D is dense in D, we obtain a marked k-pair (X , D) of dimension n and depth t
1
such that : X Y is tale on the generic ber. Note that (an K ) WK is a neighborhood
of x containing 1 D as (D) W. Here, x XKan is an arbitrary preimage of the original
x XKan , which exists by construction.
We take U to be an arbitrary open neighborhood of 1 D contained in (an 1
K ) WK , and
to be K |U . By the same argument in Step 5 of the proof of [Ber07, Lemma 2.1.2], induces a
an
morphism of K -formal schemes X c 1 b c b

/ W k K Y and thus a morphism X/D k K Y.
The conclusions of the lemma are all satised by the construction.
From now on, we assume that K is a residually algebraic extension of Qp .
Definition 3.3 (Fundamental chart). Let X be a K-analytic space and x X a point. A
fundamental chart of (X; x) consists of data (D, (Y, D), (D, ), W, ; y) where
(Y, D) is a marked k-pair of dimension t(x) + s(x) and depth t(x), where k is a nite
extension of Qp ,
D is an open poly-disc over L of dimension dimx (X) t(x) s(x), where L is
simultaneously a nite extension of K and a (residually algebraic) extension of k,
D is an integrally smooth anoid k-algebra, and
c
(3.1) : Spf D [[T0 , . . . , Tt ]]/(T0 Tt )
X /D
is an isomorphism of formal k -schemes, where is a uniformizer of k,

W is an open neighborhood of (Y/D ) b k L = 1 D in Y an ,
e
L L
y is a point in D L W such that it projects to 0 in D and a point in W whose
reduction is the generic point of DLe,
: D L W X is an tale morphism of K-analytic spaces.
Note that the elds k and L will be implicit from the notation (as they are not important).
The isomorphism (3.1) induces an isomorphism Spec D e
f D of k-schemes, and an isomor-
phism
q
M
q 1 j qj
(3.2) : HdR (D L (W, DLe))
Hrig (D/k) k HdR (Et ) k L
j=0
of L-vector spaces [GK02, Lemmas 2 & 3] and [Ber07, Corollary 5.5.2]. Here, Et is the
k-analytic space dened in Example 2.2. Denote by Hwq (D, (Y, D), (D, ), W ) the subspace
of the left-hand side of (3.2) corresponding to the subspace
q
M j qj
Hrig (D/k)w2(qj) k HdR (Et ) k L
j=0
qj
on the right-hand side. In particular, all elements in HdR (Et ) are regarded to be of weight
2(q j). Then we have a direct sum decomposition
q
M
(3.3) HdR (D L (W, 1 DLe)) = Hwq (D, (Y, D), (D, ), W ).
wZ
q q
Finally, we denote by H(w) (D, (Y, D), (D, ), W ) the subspace of HdR (DL W ) as the inverse
image of Hw (D, (Y, D), (D, ), W ) under the restriction map
q
q q
HdR (D L W ) HdR (D L (W, 1 DLe)).
In what follows, if D is of dimension 0, then we suppress it from all notations.
Remark 3.4. Note that Hwq (D, (Y, D), (D, ), W ) = 0 unless q w 2q, and the de-
composition (3.3) is stable under base change along a residually algebraic extension of K
(and L accordingly). We warn that the decomposition (3.3) depends on all of the data
(D, (Y, D), (D, ), W ), not just the L-analytic germ D L (W, 1 DLe). (However, the de-
pendence on D and W is very weak.)
Definition 3.5. Let X be a K-analytic space and x X a point.

(1) Let f Et(X; x) be the category whose objects are fundamental charts of (X; x), and a
morphism
: (D2 , (Y2 , D2 ), (D2 , 2 ), W2 , 2 ; y2 ) (D1 , (Y1 , D1 ), (D1 , 1 ), W1 , 1 ; y1)
consists implicitly extensions of related elds K L1 L2 such that k1 k2 , and

a morphism () : D2 L2 W2 D1 L1 W1 of L1 -analytic spaces sending y2 to y1 ,
and such that
q q
() H(w) (D1 , (Y1 , D1 ), (D1 , 1 ), W1 ) H(w) (D2 , (Y2, D2 ), (D2 , 2 ), W2 )
for all q, w Z. Note that () needs not to respect each factors.

(2) Let Et(X; x) be the category of tale neighborhoods of (X; x). Recall that its objects
are triples (Y, ; y) where : Y X is an tale morphism sending y Y to x, and
morphisms are dened in the obvious way. In the notation (Y, ; y), the morphism
will be suppressed if it is not relevant. For a presheaf F on Xet , the stalk of F at
x is dened to be Fx := lim(Y,;y) F (Y ) where the colimit is taken over the category

Et(X; x).

(3) We have a functor : f Et(X;
x) Et(X; x) sending an object (D, (Y, D), (D, ), W, ; y)

of f Et(X; x) to (D L W, ; y), and a morphism to ().
The following lemma generalizes [Ber07, Proposition 2.1.1].
Lemma 3.6. Suppose that K is embeddable into Cp and X is a smooth K-analytic space.

Fix an arbitrary point x X and let (Y, 0 ; y0) be an object of Et(X; x). Then
(1) there exists an object (D, (Y, D), (D, ), W, ; y) f Et(X; x) such that its image un-
der admits a morphism to (Y, 0 ; y0 );

(2) given two morphisms i : (Di , (Yi, Di ), (Di , i ), Wi , i ; yi) (Y, 0 ; y0 ) in Et(X; x)

for i = 1, 2, there exists an object (D, (Y, D), (D, ), W, ; y) f Et(X; x) together

with morphisms i to (Di , (Yi , Di), (Di , i ), Wi , i ; yi) in f Et(X; x) for i = 1, 2 such
12 YIFENG LIU
that the following diagram

(D1 , (Y1, D1 ), (D1 , 1 ), W1 , 1 ; y1 )
(1 )
3
1

*
(D, (Y, D), (D, ), W, ; y) 4
(Y, 0 ; y0 )

(2 )
2
+
(D2 , (Y2, D2 ), (D2 , 2 ), W2 , 2 ; y2 )
commutes.

In particular, the functor : f Et(X;
x) Et(X; x) is initial.
Proof. We may assume that X is of dimension n. Put t = t(x) and s = s(x).
For (1), by [Ber07, Proposition 2.3.1], after taking a nite extension of K, we may assume
that Y = D K X and y0 = (0, x ) (which makes sense) for a point x X with t(x ) = t
and s(x ) = s, where X is a smooth K-analytic space of dimension s + t. Now we only
need to apply Lemma 3.2 to Y = Spf K , the pair (X ; x ), and the structure morphism
X Y = M(K). The existence of (D, ) is due to the argument in Part (iv) of the proof
of [GK02, Theorem 2.3].
For (2), we may assume that K = L1 = L2 . For i = 1, 2, we choose a relative compacti-
cb
cation Yi Yi over ki , where Yi is proper. Then Wi is open in Yi , where Yi = Y i ki K .
Consider the tale morphism
0 : Y := (D1 K W1 ) Y (D2 K W2 ) Y,
and a point y0 Y projecting to y1 (resp. y2 ) in the rst (resp. second) factor. Again
by [Ber07, Proposition 2.3.1], we may nd an object of the form (D K X , ; (0, x)) in
Et(X; x) as in (1) with a morphism to (Y , 0 ; y0 ). Now we apply Lemma 3.2 to X , the

point x , Y = Y1 K Y2 , the morphism
(1 ,2 )
X W1 K W2 Y ,
where i equals the composition
X {0} K X D K X Di K Wi Wi (i = 1, 2)
with the last arrow being the projection. We obtain a marked k-pair (Y, D) of dimension
s + t and depth t, for some discrete non-Archimedean eld k containing k1 , k2 and contained
in (possibly a nite extension of) K; an open neighborhood W of (Yb/D ) b k K in Y an , a
K
point y W , a morphism of K-analytic spaces : W X such that (y ) = x , and a
morphism of formal K -schemes = (1 , 2 ) : Yb/D b k K Y1 K Y2 compatible with
. As i maps the generic point of DKe to the generic point of (Di )Ke , we may replace
(Y, D) by an ane open such that i (DKe ) (Di )Ke for i = 1, 2. In particular, we have
morphisms i : Yb/D c
b k K Y b
i/Di ki K . Note that i does not necessarily descent to any
nite extension of k. By the proof of [GK02, Theorem 2.3], there is an integrally smooth
k-anoid algebra D and an isomorphism as in (3.1).
Now the object (D, (Y, D), (D, ), W, ; y) has been constructed with y = (0, y ) and the
obvious . Let (i ) be the composite morphism D K W D K X Di K Wi for
i = 1, 2. It remains to show that
q
(i) For i = 1, 2, every q, every w, and an element Hrig (Di /ki )w , we have
q
(i ) (i )1 H(w) ((Y, D), (D, ), W ).
(ii) For i = 1, 2 and an arbitrary coordinate T of Eti (where i is a uniformizer of ki), we
have
dT
(i ) (i )1 1
H(2) ((Y, D), (D, ), W ).
T
Note that the composite morphism of formal k -schemes

t b
Spf((E Yb/D
) k D ) Spf D [[T1 , . . . , Tt ]]/(T1 Tt )
induces an isomorphism
q q
(W, 1 DKe ) Hrig g
(Spec E t
HdR e
k
DKe /K)
under which
q
Hwq ((Y, D), (D, ), W ) = Hrig g
(Spec E t
e
k
DKe /K)w
for every q and every w.
For (i), as we have morphisms of formal K -schemes
t b b i c
Spf((E ) k D
b k K ) Y b
/D k K b k K Spf D
Yi/Di i
b
i ki K ,
q
Lemma 2.1 implies that (i ) (i )1 coincides with i in HdR (W, 1 DKe ), where
g
i : Spec E t DKe (Di )Ke
e
k
f
is the induced morphism of (ane smooth) K-schemes.
For (ii), we may assume that Spf Di has a K -point by replacing K by a nite extension

(at the very beginning). Thus we have morphisms of formal K -schemes

t b b i c T
Spf((E b k K ) Y
) k D b
/D k K
b k K Et
Yi/Di i
b
i ki K Spf K [[T ]].
On the generic ber, the image of the induced morphism M(E t b b k K) D(0; 1)
k D
does not contain 0, which implies that it qfactors through a morphism M(E t b b k K)
k D
1 1
M(Khr T, rT i) for a unique r < 1 in |K | as (E ) t b k D is smooth over k . By tak-
ing a nite extension of K, we may assume that r |K |. Then Khr 1 T, rT 1i is integrally

smooth, and we have Spec Khr 1 ^T, rT 1 i (Gm ) . Moreover,

e
K
dT
1
Hrig 1
(Gm /K)2 = Hrig 1
(Gm /K) HdR (D(0, 1), M(Khr 1T, rT 1 i)) = K{ }.
T
Thus Lemma 2.1 implies (ii).
Remark 3.7. The above lemma with its proof implies the following: For part of the data
(D, (Y, D), (D, ), W ) from Denition 3.3 and f O (D L W ), we have
df 1
H(2) (D, (Y, D), (D, ), W ).
f
df
Here, we regard f
, a priori a closed 1-form on D L W , as an element in HdR
1
(D L W ).
Now we are ready to dene the desired direct summand (q,cl q1

X /dX )w in the weight
decomposition of de Rham cohomology sheaves.
14 YIFENG LIU
Definition 3.8 (De Rham cohomology sheaves with weights). Suppose that K is residually
algebraic over Qp and X is a smooth K-analytic space.
For every object U of Xet , dene (q,cl q1 q,cl q1
X /dX )(U)w (X /dX )(U) to be the image
pre
of elements q,clX (U) such that for every point u U, there exists a fundamental chart
q
(D, (Y, D), (D, ), W, ; y) of (U; u) such that , regarded as an element in HdR (D L W ),
q
belongs to H(w) (D, (Y, D), (D, ), W ). The assignment U 7 (q,cl X /d q1
X )(U) w denes a
pre
q,cl q1 pre q,cl q1

sub-presheaf (X /dX )w of X /dX .
We dene (q,cl q1 q,cl q1 pre
X /dX )w to be the sheacation of (X /dX )w , which is canonically
a subsheaf of q,cl
X /dX .
q1
The following lemma can be proved by the same way as for [Ber07, Corollary 5.5.3].
Lemma 3.9. Let K /K be an extension such that K is embeddable into Cp . Let X be a
smooth K-analytic space and : X := X
b K K X the canonical projection. Then the

canonical map of sheaves on Xet
1 (q,cl q1 q,cl q1
X /dX ) L K X /dX
is an isomorphism, where L is the algebraic closure of K in K .

The following theorem establishes the functorial weight decomposition of de Rham coho-
mology sheaves in Theorem 1.1 in the case of tale topology.
Theorem 3.10. If K is embeddable into Cp and X is a smooth K-analytic space, then we
have that
(1) under the situation of Lemma 3.9,
1 (q,cl q1 q,cl q1
X /dX )w L K = (X /dX )w ,
for every w Z;
(2) the image of the composite map

(qX1 ,cl /dqX1 1 )w1 (qX2 ,cl /dqX2 1 )w2 qX1 ,cl /dqX1 1 qX2 ,cl /dqX2 1 q1 +q2 1
qX1 +q2 ,cl /dX
is contained in the subsheaf (qX1 +q2 ,cl /dqX1 +q2 1 )w1 +w2 ;
(3) the sheaf (q,cl q1
X /dX )w is zero unless q w 2q;
(4) the canonical map
M
(q,cl q1 q,cl q1
X /dX )w X /dX
wZ
is an isomorphism;
(5) for every morphism f : Y X of smooth K-analytic spaces, we have
f # (f 1 (q,cl q1 q,cl q1
X /dX )w ) (Y /dY )w
for every w Z. Here, f # denotes the canonical map f 1 X Y and induced

maps of cohomology sheaves.
Proof. Part (1) follows from the denition and Remark 3.4. Part (2) follows from denition
and Lemma 3.6 (2).
For the remaining parts, it suces to work on stalks. Thus we x a point x X with
t = t(x) and s = s(x).
For (3), take an element [] in the stalk of (q,cl q1

X /dX )w at x for some w < q or w > 2q.
We may assume that it has a representative q,cl X (U) for some tale neighborhoods (U; u)
of (X; x). By denition, we have a fundamental chart (D, (Y, D), (D, ), W, ; y) of (U; u)
q
such that = 0 in HdR (D L (W, 1 DLe)) by Remark 3.4. Then there exists an open
q
neighborhood W of 1 DLe in W , such that = 0 in HdR (D L W ). In other words,
q,cl q1
[] = 0 in the stalk of (X /dX )w at x.
For (4), we rst show that the map is injective. Let [] be an element in the stalk
q,cl q1 P P
X,x /dX,x . Suppose that we have [] = []1w = []2w in which both []1w and []2w
are in the stalk of (q,cl q1
X /dX )w at x. We may choose an object (U; u) Et(X, x) such
q,cl q1
that []w has a representative w (X /dX )(U)w for i = 1, 2 and every w Z, and
i i pre
P 1 P P P
w = w2 . In particular, [] has a representative := w1 = w2 on (U; u). Fix
a weight w Z. It suces to show that [w1 ] = [w2 ] in the stalk at x. By Denition
3.8, there exist two fundamental charts (Di , (Yi , Di), (Di , i ), Wi , i ; yi) of (U; u) such that
q
i wi belongs to H(w) (Di , (Yi , Di), (Di , i ), Wi ) for i = 1, 2. By Lemma 3.6, we may nd
another fundamental chart (D, (Y, D), (D, ), W, ; y) f Et(U; u) as in that lemma. Then
q
we have (i ) w H(w) (D, (Y, D), (D, ), W ) for both i = 1, 2. However, (1 ) w1 and
i
q
(2 ) w2 , after restriction to HdR (D L (W, 1 DLe)), must be equal, as they are both the
q
weight w component of in HdR (D L (W, 1 DLe)) under the decomposition (3.3). As
q
the map HdR (D L W ) (q,cl q1 q
X /dX )x factors through HdR (D L (W, DL
1
e )), we have
[]w = []w . Finally, Lemma 3.11 below implies that the map in (4) is surjective as well.
1 2
For (5), we take a point y Y such that f (y) = x. We may take a fundamental chart
(D, (Y, D), (D, ), W, ; y) of (X; x) and replace X by D L W and x by a point (0, x) where
x W with t(x) = t and s(x) = s such that dim W = s + t. By the same proof of Lemma
3.6 (2), we may nd a fundamental chart (D , (Y , D ), (D , ), W , ; y ) of (Y ; y) such that
q q
(f ) H(w) (D, (Y, D), (D, ), W ) H(w) (D , (Y , D ), (D , ), W ). This conrms Part (5)
q q
since H(w) (D, (Y, D), (D, ), W ) (resp. H(w) (D , (Y , D ), (D , ), W )) restricts to the weight
w part in the stalk of q,cl q1 q,cl q1
X /dX (resp. Y /dY ) at x (resp. y), by Lemma 3.11 below.
The following lemma is the most crucial and dicult part in the proof of the weight
decomposition.
Lemma 3.11. Let the assumptions be as in Theorem 3.10. We take a point x X.

For any fixed weight w, an object (D, (Y, D), (D, ), W, ; y) f Et(X, x), and an element
q q,cl q1
H(w) (D, (Y, D), (D, ), W ), the induced class [] X,x /dX,x belongs to the stalk of
(q,cl q1
X /dX )w at x.
Proof. We may assume L = K where L is the nite extension of K implicitly contained in

the data of the fundamental chart. To simplify notation, we denote by V the strictly K-
anoid domain 1 DKe in W . As D will be irrelevant in the discussion, we will regard y as
a point in V . Moreover, by possibly shrinking (Y, D), the decomposition (3.2) and Remark
q q
3.7, we may assume that the image of in HdR (D K (W, V )) is in Hrig (D/K)w .

Step 1. We choose a smooth k -algebra D (of dimension s) such that its -adic com-

pletion is D , where we recall that is a uniformizer of the discrete non-Archimedean eld

k K. In particular, we may identify (Spec D )s with D, and M(D) with a strictly k-anoid
domain in (Spec D )ank . As in Lemma 2.1, we have germs (W, V ) and ((Spec D )k , M(D))
an
and a morphism V M(D) b k K induced from . We choose a neighborhood U of the

16 YIFENG LIU
graph of the previous morphism as in the proof of Lemma 2.1, such that the induced map

HdR (W, V ) HdR (W k (Spec D )an
k , U (V k M(D)))
is an isomorphism. By a similar argument in the proof of [GK02, Lemma 2], we may replace
W by a smaller open neighborhood of V such that there is a morphism W U sending V
into U (V k M(D)) whose induced map

HdR (W k (Spec D )an
k , U (V k M(D))) HdR (W, V )
is the inverse of the previous isomorphism. In other words, we have a morphism : W

(Spec D )an
K sending V into M(D)k K such that, although |V might not coincide with the
b
original morphism V M(D) b k K induced from , we still have that the induced map

Hrig (D/K) HdR ((Spec D )an
k , M(D)) k K HdR (W, V )
coincides with the map induced from the Knneth decomposition (3.2) (where D is trivial).
Step 2. We choose a compactication (Spec D )k S k over k, and dene S to be the
`
k -scheme S k (Spec D )k Spec D . Apply [dJ96, Theorem 8.2] to the k -variety S and Z = .

We obtain a nite extension k /k, an alteration S S k and a k -compactication S S

where S is a projective strictly semi-stable scheme over k such that S\S is a strict normal
crossing divisor of S (concentrated on the special ber). We may further assume that all
irreducible components of Ss are geometrically irreducible. To ease notation, we replace k
by k and possibly K by a nite extension. We may x an irreducible component E of Ss
such that its generic point belongs to Ss and maps to the generic point of S s D. Note
that the complement of E := E Ss in E is exactly Ss[1] E. Denote by E the unique point
in SKan
who reduction is the geometric point of EKe . Then 1 EKe is an open neighborhood of
E .
Dene W via the pullback square in the following diagram:

$
W / (S )an / an
SK

K

W / (S)an
K,
and : W SK an
as the composition in the upper row. We may choose a point y W
q
which lifts y and maps to E in SK an
. The image of the form in Hrig (D/K)w induces a
q
class [ ] Hrig (E /K)w via restriction along the alteration. By taking a nite unramied
extension of k (and possibly a nite extension of K), we may assume that (Fr pf w )N [ ] = 0
for some integer N 1, where #ke = p2f and Fr denotes the relative Frobenius of E /k. e Put
S = Sb/E . We x an open neighborhood U of 1 E in S such that [ ] has a representative
q b k K).
HdR (U
Step 3. We are now going to shrink U such that has controlled behavior on U\ 1 E .
We may cover S by nitely many special open formal k -subschemes Si satisfying the
following conditions: Each Si is tale over
Spf k [[t0 ]]ht1 , . . . , tr , tr+1 , t1 1
r+1 , . . . , ts , ts i/(t0 tr )
for some 0 r = ri s; Ei := E Ss is ane; and if we write fi,j for the image of tj in Si ,

then Ei := E Ss is dened by the equations fi,0 = 0 and fi,1 fi,r 6= 0. Dene the formal
k -scheme Si via the following pullback diagram
Si / Si

Spf k
ht1 , t1 1
1 , . . . , ts , ts i
/ Spf k
[[t0 ]]ht1 , . . . , tr , tr+1 , t1 1
r+1 , . . . , ts , ts i/(t0 tr ).
Then (Si ) = 1 Ei in Si . For 0 < < 1, denote by Si () the open subset of Si

dened by the inequality |fi,0| < ||1. Then Si () form a fundamental system of open
neighborhoods of (Si ) in Si . In fact, the open subset Si () does not depend on the
choice of the tale coordinates as above. We choose an open neighborhood Ui of (Si ) in S
contained in U, together with an absolute Frobenius lifting i : Ui U satisfying properties
p 2f
(a) i fi,j = fi,j for j = 1, . . . , r (as in [Chi98, Lemma 3.1.1]);
(b) |(i g g )(x)| < 1 for all regular functions g on Si and all x Ui at which both g and
p2f
i g are dened (as [Ber07, Lemma 6.1.1]);

q
(c) (i pf w )M = 0 in HdR b k K) for some integer M 1.
(Ui
Since Ui Si is an open neighborhood of (Si ) in Si , there exists some i > 0 such that
S
Si (i ) Ui . Take = min{} > 0, and replace U by the union S () := i Si () in S ,
which is an intrinsically dened open neighborhood of 1 E in S . We suppose that is
very close to 0 in terms of pf , s, ||. Now we replace W by W ( )1 (S () b k K). By
construction, we may remove a Zariski closed subset of W of dimension at most s + t 1

such that the resulting morphism D K W D K W X is tale. In particular,
(D K W ; (0, y )) is an object of Et(X; x).
Step 4. It remains to show the following claim:
For every point u W , there exists a fundamental chart (D , (Y , D ), (D , ), W , ; y )
q
of (W ; u) such that ( ) belongs to H(w) (D , (Y , D ), (D , ), W ).
We start similarly as in the proof of Lemma 3.6. By [Ber07, Proposition 2.3.1], after
taking a nite extension of K, we may assume that W = D K X and u = (0, x ) for a
point x X with t(x ) = t(u) and s(x ) = s(u), where X is a smooth K-analytic space of
dimension s(u) + t(u). Thus we have a morphism : X {0} K X S () b k K. If (u)
1 b
belongs to ( E )k K, the our claim follows in the same way as Claim (i) in the proof of
Lemma 3.6 (2). In general, (u) belongs to Si () b k K for some i, and we assume that its

reduction ( (u)) belongs to (Ss[r ] \Ss[r +1] ) Ei for a unique 0 r ri . (If r = 0, then
we are back to the previous special case.) Without lost of generality, we may assume that
r = ri = r.
Let F E be the irreducible component of Ss[r] \Ss[r+1] where ( (u)) belongs to. By
shrinking Si , we may assume that F is dened by the equations fi,0 = = fi,r = 0, and
there exists an integrally smooth k-anoid algebra F together with an isomorphism
(3.4) c
Spf F [[ti,0 , . . . , ti,r ]] S i/F
of formal k -schemes, also sending ti,j to fi,j . Therefore, we have an isomorphism of graded
k-algebras
(3.5)
HdR (S , 1 F ) Hrig

(Spec Fe /k) k HdR

(Er ),
18 YIFENG LIU
where Er is the k-analytic space in Example 2.2. By [GK02, Lemma 3] and the above
isomorphism, the restriction map

(3.6) HdR (S , 1 F ) HdR

(S (), S () 1 F
is an isomorphism. Now it suces to show that the class of in
b k K 1 F ) H (S (), S () 1 F ) k K
HdR (S ()
b k K, S ()
e
K dR
is of weight w with respect to the decomposition (3.5) and the isomorphism (3.6). Then our
claim follows in the same way as in the proof of Lemma 3.6 (2). Without lost of generality,

we now assume that is an element in HdR (S (), S () 1 F ).
Step 5. To compute the weight, we use the Frobenius lifting i : Ui S () where
Ui S () is an open neighborhood of (Si ) in S , which might be smaller than the one we
start with. Assume that Ui Si contains Si ( ) for some 0 < < . We introduce more
notations as follows: We x a positive integer N such that 0 < 1/N < p2f . Replacing
K by a nite extension, we may assume that there exists a totally ramied extension k+ /k

contained in K with an element + k+ such that + rN
= . We consider the following
k+ -anoid algebras
b k k+ h1 , 1 , . . . , r , 1 i,
F0 = F 1 r
* rN r
+
b k k+
ti,0 + ti,1 + ti,r +
F1 = F rN r , , , ,..., , /(ti,0 ti,r ),
+ ti,0 + ti,1 + ti,r
* rN rp 2fp 2fp 2f +
b k k+
ti,0 + ti,1 + ti,r +
F2 = F rN rp2f
, , p2f , , . . . , p2f , /(ti,0 ti,r ).
+ ti,0 + ti,1 + ti,r
Note that F0 is integrally smooth. We have natural isomorphisms
r
Y
rN r
1 : F1
F0 , ti,j 7 + j , 1 j r, ti,0 7 + j1 ;
j=1
r
Y
p 2f rN rp 2f
2 : F2
F0 , ti,j 7 + j , 1 j r, ti,0 7 + j1 .
j=1

For = 1, 2, we dene a formal k+ -scheme Sihi via the following pullback diagram
Sihi c b
i/F k k ,
/S
+
(3.4)

Spf F /
Spf F [[ti,0 , . . . , ti,r ]] k k+
b k k+ by our choice of N.
so that Si hi is canonically a strictly k+ -anoid domain in Ui
Moreover, induces an isomorphism, denoted again by ,

: Spf F0
Si hi

of formal k+ -schemes. Properties (a) and (b) of the Frobenius lifting i implies that it induces
by restriction a morphism i : Sih1i Sih2i , and the composition 1 2 i 1 : M(F0 )
M(F0 ) is a Frobenius lifting. We x a smooth k+ -anoid germ (V, M(F0 )).
Note that for = 1, 2, we have isomorphisms
b k k+ , 1 F ) b k k+ 1 F )
HdR (S HdR (S ()
b k k+ , S () HdR (Ui
b k k+ , Si hi )

by [GK02, Lemma 3]. In particular, we may equip HdR (Ui
b k k+ , Si hi ) with a weight
decomposition inherited from (3.5). By construction and [Bos81, Corollary 1], we have
a morphism 1 : (V, M(F0 )) (Ui
b k k+ , Si h1i ) such that |M(F ) is very close to
1 0
1 which induces the same morphism on the special ber, and moreover the induced
restriction map
(1 ) : HdR

(Ui rig
f /k )
b k k+ , Si h1i ) H (Spec F0 +
is an isomorphism respecting weights,

b k k+ , Si h2i ) (V, M(F0 )) such that |S h2i is very close to
a morphism 2 : (Ui 2 i
1
2 (not 2 !) which induces the same morphism on the special ber, and moreover
the induced restriction map
(2 ) : Hrig
f /k ) H (U
(Spec F0 + dR
b
i k k+ , Si h2i )
is an isomorphism respecting weights.

In summary, we have weight preserving isomorphisms
b k k+ 1 F )
HdR (S ()
b k k+ , S ()

s
+
b k k+ , Si h1i ) b k k+ , Si h2i )
HdR (Ui HdR (Ui
3

(1 )

+ (2 )

Hrig f /k ).
(Spec F0 +
We will identify the top three objects in the above commutative diagram. Recall that we
b k k+ 1 F ). Let 0 be the element in
regard as an element in HdR (S ()
b k k+ , S ()
q
f /k ) such that ( ) = . By Property (c) of the Frobenius lifting , we
Hrig (Spec F0 + 2 0 i

have that ((1 ) i (2 ) p ) 0 = 0. However, 2 i 1 : (V, M(F0 )) (V, M(F0 ))
fw M
is a Frobenius lifting of the Frobenius endomorphism of Spec F f over kf = k.e Therefore,

0 + 0

and hence are of weight w. The lemma is nally proved!
Remark 3.12. From the proof of Theorem 3.10, we know that the support of (q,cl q1
X /dX )w
is contained in the subset {x X | s(x) 2q w, s(x) + t(x) q}.
4. Logarithmic differential forms

In this section, we study the behavior of logarithmic dierential forms in the rigid co-
homology. Based on this and Theorem 3.10, we nish the proof of Theorem 1.1 for both
topologies.
Let k be a nite extension of Qp . Let S be a proper strictly semi-stable scheme over k
of dimension s. We x an irreducible component E of Ss and let E1 , . . . , EM be all other
T
irreducible components that intersect E. For a subset I {1, . . . , M}, put EI = ( iI Ei ) E
(in particular, E = E) and EI = EI \Ss[|I|+1] . For two subsets I, J of {1, . . . , M}, we write
I J if I J and numbers in J\I are all greater than those in I.
20 YIFENG LIU
For I {1, . . . , M}, we have the open immersion EI EI \Ss[|I|+2] , whose compliment is
`
IJ,|J|=|I|+1 EJ . Thus we have maps
M M

Hrig (EI /k) HE+1
(E \Ss[|I|+2] /k)
,rig I
1
Hrig (EJ /k),
J
IJ,|J|=|I|+1 IJ,|J|=|I|+1
where the second map is the Gysin isomorphism. In the above composite map, denote by
JI the induced map from Hrig
(EI /k) to the component Hrig1
(EJ /k) if I J, and the zero
map if not.
In general, for I J, there is a unique strictly increasing sequence I = I0 I1
I|J\I| = J and we dene
I |J\I|

JI := J|J \I|1 II1 : Hrig (EI /k) Hrig (EJ /k),
and JI = 0 if I J does not hold. Together, for i j, they induce a map
M M +ij
(4.1) ji :
Hrig (EI /k) Hrig (EJ /k),
|I|=i |J|=j
such that ji |H (E /k) is the direct sum of JI for all J with |J| = j. First, we have the
rig I
following lemma.
Lemma 4.1. Let notation be as above. For every 0 q s, the restriction of
q
M
q0 : Hrig (E /k) 0
Hrig (EJ /k)
|J|=q
q
to Hrig (E /k)2q is injective.
Proof. By the long exact sequence of cohomology with support (2.1), the kernel of the map
q+|I|
q0 is a weight preserving extension of k-vector spaces HEI ,rig (E/k) for |I| < q. Therefore, the
q+|I|
lemma follows since HEI ,rig(E/k) is pure of weight q + |I| < 2q by [Tsu99, Theorems 5.2.1 &
6.2.5] (with constant coecients).
Denote by Z i (E) the abelian group generated by EI with |I| = i, modulo the subgroup
L
generated by EI with EI = . Put Z(E) = M
i=0 Z (E) . The image of EI in Z(E) will be
i
denoted by [EI ]. We dene a wedge product

: Z(E) Z(E) Z(E) ,
which is group homomorphism uniquely determined by the following conditions:
Z1 Z2 = (1)ij Z2 Z1 , if Z1 Z i (E) and Z2 Z j (E) ;
[EI ] [EJ ] = 0 if I J 6= ;
[EI ] [EJ ] = [EIJ ] if I J = and I I J.
It is easy to see that is associative and maps Z i (E) Z j (E) into Z i+j (E) . We have an
(injective) class map
M M
cl : Z(E) 0
Hrig (EI /k) k 0 (EI )
I I
sending [EI ] to the canonical generator on (each connected component of) EI .

For an element f O (Skan , 1 E ), that is, an invertible function on some open neigh-
borhood of 1 E in Skan , we can associate canonically an element div(f ) Z 1 (E) . In fact,
there exists an element c k such that |cf | = 1 on 1 E . Thus the reduction cf f is an ele-
ment in OE (E ), and we dene div(f ) to be the associated divisor of cf,f which is an element
in Z 1 (E) . Obviously, it does not depend on the choice of c. Finally, note that by the deni-

tion of rigid cohomology, we have a canonical isomorphism HdR (Skan , 1 E ) Hrig

(E /k).
Proposition 4.2. Let notation be as above. Given f1 , . . . , fq O (Skan , 1 E ), if we regard
df1
f1
df
fq
q q
as an element in HdR q
(Skan , 1 E ) Hrig (E /k), then we have
!
df1 dfq
(4.2) q0 = cl (div(f1 ) div(fq )) .
f1 fq
Proof. The question is local around the generic point of every connected component of EI
with |I| = q. Thus, we may assume that S is ane and admits a smooth morphism
f : S Spec k [T0 , . . . , Tq ]/(T0 Tq )
where is a uniformizer of k, such that
E = E0 and Ei (i = 1, . . . , q) are all the irreducible components of Ss that intersect
E, where Ei is dened by the ideal (f Ti , );
EI is irreducible and nonempty for I {1, . . . , q}.
d(cf )
Since cf
= df
f
; both sides of (4.2) are multi-linear in f1 , . . . , fq O (Skan , 1 E ); and

df
f
= dff
in Hrig
1
(E /k) if |f | = |f | = 1 on 1 E and fe = fe , we may assume that
fi = f Ti . Then as both sides of (4.2) are functorial in f under pullback, we may assume
that S = Spec k [T0 , . . . , Tq ]/(T0 Tq ) and fi = Ti .
Put S = Spec k [T1 , . . . , Tq ] and let g : S S be the morphism sending Ti to Ti (1 i
q). For I {1, . . . , q}, let EI be the closed subscheme of Ss dened by the ideal (, Ti |i I).
Then g induces an isomorphism EI EI . Similarly, we have maps
I +ij
J : Hrig

(EI /k) Hrig (EJ /k)
for I J and ij for i j, where EI = g(EI ). It is easy to see that IJ = JI if we identify
Hrig
(EI /k) with Hrig

(EI /k) through g . Therefore, it suces to show the equality (4.2) for
S , that is, !
0 dT1 dTq 0
q = 1 Hrig (E{1,...,q} /k) k.
T1 Tq
However, S , which is isomorphic to Aqk can be canonically embedded into the proper
smooth scheme Pqk over k . Thus, the rigid cohomology Hrig
(EI /k) and the map ij can
be computed on (Pqk )an . On the generic ber Sk , we similarly dene TI to be the closed
S
subscheme Spec k[T1 , . . . , Tq ]/(Ti | i I) of Sk for I {1, . . . , q}, and TI = TI \ I J TJ .
|J\I|
We may similarly dene maps JI : HdR
(TI ) HdR (TJ ) and ji via algebraic de Rham
cohomology theory. Then we have canonical vertical isomorphisms rendering the diagram
IJ |J\I|

HdR (TI ) / HdR (TJ )

IJ |J\I|

Hrig (EI /k) / Hrig (EJ /k)
22 YIFENG LIU
commutative. From the standard computation in algebraic de Rham cohomology, we have

!
dT1 dTq
q0 = 1 HdR0
(T{1,...,q} ),
T1 Tq
where T{1,...,q} is just the point of origin. Thus, the proposition is proved.
Now we are ready to prove Theorem 1.1. We begin with the case of tale cohomology and
then the case of analytic topology.
Proof of Theorem 1.1 in tale topology. As in the previous section, sheaves like OX , cX , and
the de Rham complex (X , d) are understood in the tale topology.
The direct sum decomposition has been proved in Theorem 3.10 (4). Property (i) follows
from Theorem 3.10 (3).
For Property (ii), the inclusion qX (q,cl q1
X /dX )2q follows from Theorem 3.10 (2) and
Remark 3.7. Now we show that (1,cl X /dOX )2 X . We check the inclusion on stalks. Take
1
1,cl
a point x X with s = s(x) and t = t(x). For every class [] in the stalk of (X /dOX )2
at x, we may nd a fundamental chart (D, (Y, D), (D, ), W, Z, ; y) of (X; x) such that
[] has a representative H(2)1
(D, (Y, D), (D, ), W ). Note that the decomposition (3.2)
specializes to the decomposition
1
HdR (D L (W, 1 DLe)) = Hrig
1 1
(D/L) HdR (Et
b k L).
If the restriction of to HdR1

(D L (W, 1 DLe)) belongs to HdR 1 b k L), then we are done.
(Et
Otherwise, restricts to Hrig (D/L)2 . It suces to show that classes in Hrig
1 1
(D/L)2 can be
represented by logarithmic dierential of invertible functions tale locally, up to a constant
multiple.
We repeat certain process in Step 2 of the proof of Lemma 3.11. Choose a smooth
k -algebra D (of dimension s) such that its -adic completion is D , a compactication
`
(Spec D )k S k over k, and dene S to be the k -scheme S k (Spec D )k Spec D . Then we
obtain a nite extension k /k, an alteration S S k and a k -compactication S S
where S is a projective strictly semi-stable scheme over k such that S\S is a strict normal
crossing divisor of S. We may further assume that all irreducible components of Ss are
geometrically irreducible. To ease notation, we replace k by k and possibly L by a nite
extension. We may x an irreducible component E of Ss such that its generic point belongs
to Ss and maps to the generic point of S s D. Thus there is a unique point E (S c )

such that (E ) is the generic point of E.
Now we apply the setup in the beginning of this section to S and E. Note that E Ss
coincides with E . It suces to show that every class in Hrig 1
(E /k)2 can be represented by
the logarithmic dierential of an invertible function on some tale neighborhood of E . Put
`
E [i] = E Ss[i] for i 1. We have E [1] \E [2] = Mi=1 E{i} . There are exact sequences
(4.3) 1
Hrig 1
(E/k) Hrig (E /k) HE2[1] ,rig (E/k) Hrig
2
(E/k),
and
HE2[2] ,rig (E/k) HE2[1] ,rig (E/k) HE2[1] \E [2] ,rig (E\E [2] /k) HE3[2] ,rig (E/k).
As the codimension of E [2] in E is at least 2, we have HE2[2] ,rig (E/k) = HE3[2] ,rig (E/k) = 0. Thus,
M
M
HE2[1] ,rig (E/k) HE2[1] \E [2] ,rig (E\E [2]/k) 0
Hrig (E{i} /k).
i=1
Since the composition

M
M M
M M
M
0
Hrig (E{i} /k) HE2{i} ,rig (E/k) HE2[1] ,rig (E/k) 0
Hrig (E{i} /k)
i=1 i=1 i=1
L
is an isomorphism, we may replace the term HE2[1] ,rig (E/k) in (4.3) by M
i=1 Hrig (E{i} /k), which
0
is isomorphic to the k-vector space Z 1 (E) k, and the boundary map

Z 1 (E) k Hrig
2
(E/k)
becomes the cycle class map in rigid cohomology. As Hrig
1
(E/k) is of pure weight 1, we have
the isomorphism

(4.4) 1
Hrig (E /k)2
ker(Z 1 (E) k Hrig
2
(E/k)).
P
Now take a divisor D = M i=1 ci [E{i} ] with ci Z such that its cycle class in Hrig (E/k)
2
2
Hcris (E/k) is trivial. Then there exists some integer > 0 such that D is algebraically
equivalent to zero, and in particular OE (D) is an element in Pic0E/ek (k). e Since Pic0E/ek is
a projective scheme over the nite eld k, e one may replace by some multiple such that
OE (D) is a trivial line bundle. Therefore, there exists a function f OE (E ) with div(f) =
D. We may assume that that f lifts to a function f O (Skan , 1 E ). (Otherwise, we may
take an ane open subscheme Spec D of S such that (Spec D )s is densely contained in E
and f|(Spec D )s lifts to a function f O(Spec Dk ), and repeat the above process to Spec D .)
By Proposition 4.2, df f
has image D under the map Hrig 1
(E /k) HE2[1] ,rig(E/k) Z 1 (E) k.
Thus, (ii) is proved.
For Property (iii), when X has dimension 1, it follows from (the proof of) [Ber07, Theorem
4.3.1]. In general, it suces to show that (1,cl 0
X /dX )1 X by [Ber07, Theorem 4.5.1 (i)]
and Theorem 3.10 (4). However, this follows from the denition of X , Theorem 3.10 (5),
and the case of curves.
Proof of Theorem 1.1 in analytic topology. Now sheaves like OX , cX , qX , and the de Rham
complex (X , d) are understood in the analytic topology. The corresponding objects in the
tale topology will be denoted by OXet , cXet , qXet , and (Xet , d).
Note that we have a canonical morphism : Xet X of sites, and OX = OXet ,
cX = cXet , qX = qXet for every q. We claim that the canonical map q,cl X /dX
q1

q,cl q1 q,cl q,cl
(Xet /dXet ) is an isomorphism. It will follow from: (a) X = Xet as subsheaves of
qX ; (b) dXq1
= dXq1

et
as subsheaves of qX ; (c) R1 dX q1

et
= 0. Assertion (a) is obvious.
Both (b) and (c) will follow from the general fact that R F = 0 for i > 0 and any sheaf of
i
Q-vector spaces F on Xet . In fact for every x X, we have (Ri F )x = H i (H(x), i1 x F ),

where H(x) is the completed residue eld of x and ix : M(H(x)) X is the canonical mor-
phism, and we know that the pronite Galois cohomology H i (H(x), i1 x F ) is torsion hence
trivial for i > 0.
Now for w Z, we dene (q,cl q1 q,cl q1
X /dX )w = (Xet /dXet )w . Then we have a decompo-
sition
M q,cl
q,cl q1
X /dX = (X /dX q1
)w ,
wZ
stable under base change and functorial in X and satisfying Property (i).
24 YIFENG LIU
For Property (ii), we have the inclusion qX qXet as subsheaves of q,cl q1

X /dX , which
is canonically isomorphic to (q,cl q1
Xet /dXet ). Thus, we have the inclusion of sheaves X
q
(q,cl q1
X /dX )2q . When q = 1, we have to show that Xet X . We check this on the
1 1
stalk at an arbitrary point x X. Take an element [] in ( 1Xet )x . We may assume that it

has a representative 1X (U) for some open neighborhood U of x satises = df f
+ dg
for some nite tale surjective morphism : U U and f O (U ), g O(U ). Then
= deg()1 df f
+ dg where f (resp. g) is the multiplicative (resp. additive) trace of f (resp.
g ) along .
5. Tropical cycle class map

In this section, we study the sheaf ker(d : AXq,0 AXq,1) and its relation with de Rham
cohomology sheaves. We construct tropical cycle class maps and show their compatibility
with integration. In this section, sheaves like OX , cX , and the de Rham complex (X , d) are
understood in the analytic topology.
Definition 5.1 (Sheaf of rational Milnor K-theory). Let (X, OX ) be a ringed site. We dene
the q-th sheaf of rational Milnor K-theory KXq for (X, OX ) to be the sheaf associated to the
presheaf assigning an open U in X to KqM (OX (U))Q ([Sou85, 6.1]). Here, KqM (OX (U)) is

the abelian group generated by the symbols {f1 , . . . , fq } where f1 , . . . , fq OX (U), modulo
the relations
{f1 , . . . , fi fi , . . . , fq } = {f1 , . . . , fi , . . . , fq } + {f1 , . . . , fi , . . . , fq },
{f1 , . . . , f, . . . , 1 f, . . . , fq } = 0.
Example 5.2. Let X be a smooth scheme of nite type over an arbitrary eld k of dimension
n. Then by [Sou85, 6.1, Remarque], we have an isomorphism

(5.1) H q (X , KXq )
clK : CHq (X )Q := CHq (X ) Q
for every integer q. It can be viewed as a universal cycle class map.
If Z is an irreducible closed subscheme of X of codimension q that is a locally com-
plete intersection, then clK (Z) has an explicit description as follows: Choose a nite ane
open covering Ui of X and fi1 , . . . , fiq OX (Ui ) such that Z Ui is dened by the ideal
(fi1 , . . . , fiq ). Let Uij be the nonvanishing locus of fij . Then {Uij | j = 1, . . . , q} is an
T
open covering of Ui \Z. Thus the element {fi1 , . . . , fiq } KqM (OX ( qj=1 Uij )) gives rise
to an element in H q1 (Ui \Z, KXq ) and hence in HZU q
i
(Ui , KXq ). One can show that the
q q
image in HZUi (Ui , KX ) does not depend on the choice of {fi1 , . . . , fiq }. Therefore, we ob-
tain a class c(Z) in H 0 (X , H qZ (X , KXq )). By [Sou85, Thorme 5], we know that the map
H i (X , KXq ) H i (X \Z, KXq ) is a bijection (resp. an injection) if i q 2 (resp. i = q 1),
and thus H iZ (X , KXq ) = 0 for i q1. Thus, the local to global spectral sequence induces an
isomorphism HZq (X , KXq ) H 0 (X , H qZ (X , KXq )). Then clK (Z) is the image of c(Z) under
the map H 0 (X , H qZ (X , KXq )) HZq (X , KXq ) H q (X , KXq ).
We recall some facts from the theory of real forms on non-Archimedean analytic spaces
developed by Chambert-Loir and Ducros in [CLD12]. (See also [Gub13] for a slightly dierent
formulation.) Let X be a K-analytic space. There is a bicomplex (AX, , d , d ) of sheaves

of real vector spaces on (the underlying topological space of) X, where AXq,q is the sheaf
of (q, q )-forms ([CLD12, 3.1]). Moreover, they dene another bicomplex (DX, , d , d ) of

sheaves of real vector spaces on X, where DXq,q is the sheaf of (q, q )-currents, together with
a canonical map
X : (AX, , d , d ) (DX, , d , d )

of bicomplexes given by integration ([CLD12, 4.2 & 4.3]). It is known that AXq,q = DXq,q =
0 unless 0 q, q dim(X).
Definition 5.3 (Dolbeault cohomology). Let X be a K-analytic space. We dene the
Dolbeault cohomology (of forms) to be

ker(d : AXq,q (X) AXq,q +1 (X))
HAq,q (X) := ,
im(d : AXq,q 1 (X) AXq,q (X))
and the Dolbeault cohomology (of currents) to be

ker(d : DXq,q (X) DXq,q +1 (X))
HDq,q (X) := ,
im(d : DXq,q 1 (X) DXq,q (X))

together with an induced map X : HAq,q (X) HDq,q (X).
By [Jel16, Corollary 4.6] and [CLD12, Corollaire 3.3.7], the complex (AXq, , d ) is a ne
resolution of ker(d : AXq,0 AXq,1 ). In particular, we have a canonical isomorphism
H (X, ker(d : AXq,0 AXq,1 )) HAq, (X).
Suppose that X is of dimension n. By denition, we have a bilinear pairing

DXq,q (U) AXnq,nq (U)c R,

for every open U X, where AXnq,nq (U)c AXnq,nq (U) is the subset of forms whose
support is compact and disjoint from the boundary of X. In particular, if X is compact and
without boundary, then we have an induced pairing

h , iX : HAq,q (X) HAnq,nq (X) R.
Definition 5.4. Let X be a K-analytic space. We have the sheaf of rational Milnor K-theory
KX for the ringed topological space (X, OX ) (Denition 5.1).
(1) We dene a map of sheaves
Xq : KXq ker(d : AXq,0 AXq,1 )
as follows. For a symbol {f1 , . . . , fq } KXq (U) with f1 , . . . , fq OX

(U), we have
the induced moment morphism (f1 , . . . , fq ) : U (Gm,K ) . Composing with the
an q
evaluation map log | | : (Gan

m,K ) R , we obtain a continuous map
q q
trop{f1 ,...,fq } : U Rq .
If we endow the target with coordinates x1 , . . . , xq where xi = log |fi |, then we
dene
Xq ({f1 , . . . , fq }) = dx1 dxq ker(d : AXq,0 (U) AXq,1 (U)).
It is easy to see that Xq factors through the relations of Milnor K-theories, and thus
induces a map of corresponding sheaves.
26 YIFENG LIU
(2) If X is moreover smooth, then we dene another map of sheaves

qX : KXq q,cl q1
X /dX
as follows. For a symbol {f1 , . . . , fq } KXq (U) with f1 , . . . , fq OX

(U), we put
df1 dfq
qX ({f1 , . . . , fq }) = ,
f1 fq
where the right-hand side is regarded as an element in q,cl X (U) and hence in
q,cl q1 q
(X /dX )(U). It is easy to see that X factors through the relations of Milnor
K-theories, and thus induces a map of corresponding sheaves.
(3) We introduce the following quotient sheaves:
TXq = KXq / ker Xq , LXq = KXq / ker qX
whenever the maps are dened.
Proposition 5.5. Let K be a non-Archimedean field and X a smooth K-analytic space.
Then Xq induces an isomorphism
TXq Q R ker(d : AXq,0 AXq,1 ).
Proof. If suces to show the isomorphism on stalks. We x a point x X with s = s(x)
and t = t(x). We rst describe a typical section of ker(d : AXq,0 AXq,1 ) around x. We say
a collection of data (U; f1 , . . . , fN ) where U is an open neighborhood of x and f1 , . . . , fN

OX (U) is basic at x if, under the induced tropicalization map
f1 ,...,f log | |
tropU : U
N
(Gan N
m,K ) TN Z R RN
where TN is the cocharacter lattice of GNm , there exists a rational polyhedral complex C of
dimension s + t with a unique minimal polyhedron U , which is of dimension t, such that
tropU (U) is an open subset of C and tropU (x) is contained in U . For every polyhedron of
C, we denote by L( ) the underlying linear Q-subspace of TN,Q := TN Z Q. Then we have
an inclusion
X
q L( ) q TN,Q
U C
of Q-vector spaces, and thus a map

X
HomQ (q TN,Q , R) HomQ ( q L( ), R).
U C
By [JSS15, Proposition 3.16], the canonical map

HomQ (q TN,Q , R) ker(d : AXq,0 (U) AXq,1 (U))
P
factors through HomQ ( U C q L( ), R), and moreover every element in the stalk
q,0 q,1 P
ker(d : AX,x AX,x ) has a representative in HomQ ( U C q L( ), R) for some basic
data (U; f1 , . . . , fN ). This implies that the induced map TXq Q R ker(d : AXq,0 AXq,1 )
is injective, as well as surjective since elements in HomQ (q TN,Q , Q) are in the image of
Xq .
Remark 5.6. Proposition 5.5 implies that for all q, q 0, we have a canonical isomorphism

H q (X, TXq Q R) HAq,q (X).

In particular, the real vector space HAq,q (X) has a canonical rational structure coming from

the isomorphism H q (X, TXq Q R) H q (X, TXq ) Q R.
Definition 5.7 (Tropical cycle class map). Let K be a non-Archimedean eld and X a
smooth scheme over K.
(1) The tropical cycle class map (in forms) clA is dened to be the composition
cl
clA : CHq (X )Q K H q (X , KXq ) H q (X an , KXqan )
H q (X an , q an )
X
H q (X an , TXqan ) H q (X an , TXqan Q R)
HAq,q (X an ),
which can be regarded as a cycle class map valued in Dolbeault cohomology of forms.
(2) The tropical cycle class map (in currents) clD is dened to be the further composition
cl
HAq,q (X an ) X HDq,q (X an ),
clD : H q (X , KXq ) A
an
which can be regarded as a cycle class map valued in Dolbeault cohomology of cur-
rents.
It is clear that both clA and clD are homomorphisms of graded Q-algebras.
The following theorem establishes the compatibility of tropical cycle class maps and in-
tegration, which can be viewed as a tropical version of Cauchy formula in multi-variable
complex analysis.
Theorem 5.8. Let K be a non-Archimedean field and X a smooth scheme over K of di-
mension n. Then for every algebraic cycle Z of X of codimension q, we have the equality
Z
hclD (Z), iX an =
Z an
for every d -closed form AXnq,nq

an (X an )c with compact support.
Proof. We may assume that Z is prime, that is, a reduced irreducible closed subscheme of
X of codimension q. Let Zsing Z be the singular locus, which is a closed subscheme of X
of codimension > q. Put U = X \Zsing , Zsm = Z\Zsing , X = X an , U = U an , and Z = Zsm
an
.
In particular, Z is a Zariski closed subset of U. To ease notation, we put

AXq,q ,cl = ker(d : AXq,q (X) AXq,q +1 (X)), DXq,q ,cl = ker(d : DXq,q (X) DXq,q +1 (X)).
We x a form AXq,q,cl (X)c . By [CLD12, Lemme 3.2.5], belongs to AXq,q,cl (U)c .
Step 1. Using Example 5.2, we describe explicitly the class clD (Zsm ). We choose a nite
ane open covering Ui of U and fi1 , . . . , fiq OU (Ui ) such that Zsm Ui is dened by the
ideal (fi1 , . . . , fiq ). Let Uij be the nonvanishing locus of fij . Put Ui = Uian and Uij = Uijan .
Then {Uij | j = 1, . . . , q} is an open covering of Ui \Z. Thus the element Uq ({fi1 , . . . , fiq })
gives rise to an element in H q1 (Ui \Z, AUq,0,cl ) H q1 (Ui \Z, AUq, ), and we denote its image
under the composite map
H q1 (Ui \Z, AUq, ) H q1 (Ui \Z, DUq, ) HZU
q
i
(Ui , DUq, )
by c(Z)i . It is easy to see that c(Z)i does not depend on the choice of fi1 , . . . , fiq . Therefore,
{c(Z)i } gives rise to an element c(Z) H 0 (U, H qZ (DUq, )). Again by [CLD12, Lemme 3.2.5],
28 YIFENG LIU
H iZ (DUq, ) = 0 for i < q, and we have an isomorphism HZq (U, DUq, ) H 0(U, H qZ (DUq, )). The
image of c(Z) in H q (U, DUq,) HDq,q (U) coincides with clD (Zsm ).
Step 2. We study HUq iZ (Ui , DUq, ) in more details. Put
q,
DU,Z (Ui ) = ker(DUq, (Ui ) DUq, (Ui \Z)),
with the induced dierential d , and put
q,q q,q +1
q,q ker(d : DU,Z (Ui ) DU,Z (Ui ))
HZ,D (Ui ) = q,q 1 q,q .
im(d : DU,Z (Ui ) DU,Z (Ui ))
As DUq, is a complex of asque sheaves, we have the following commutative diagram

HDq,q 1 (Ui \Z) / q,q
HZ,D (Ui ) / HDq,q (Ui ) / HDq,q (Ui \Z)

H q 1 (Ui \Z, DUq,) / HUq i Z (Ui , DUq, ) / H q (Ui , DUq, ) / H q (Ui \Z, DUq, ).
In particular, when q = q we have

q,q
HUq i Z (Ui , DUq, ) HZ,D (Ui ) ker(DUq,q,cl (Ui ) DUq,q (Ui \Z)) DUq,q,cl (Ui ).
Let i AUq,q1,cl (Ui \Z) be a Dolbeault representative of Uq ({fi1 , . . . , fiq }) as a cohomology

class in H q1(Ui \Z, AUq,0,cl ), with induced class [i ] HDq,q1(Ui \Z). By partition of unity,
P
we may write = i i with i AUnq,nq (Ui )c . Note that
X X
hclD (Z), iX an = hclD (Zsm ), iU = h ([]), iU = h ([]), iiU = h ([i ]), iiU
i i
and Z XZ
= i .
Z an i Ui Z
To prove the theorem, it suces to show that

Z
h ([i ]), iiU = i
Ui Z
for every i.
Step 3. In what follows, we suppress the subscript i. We summarize our data as follows:
an ane smooth scheme U over K of dimension n, with U = U an ,
a smooth irreducible closed subscheme Z of codimension q dened by the ideal
(f1 , . . . , fq ) where f1 , . . . , fq OU (U), with Z = Z an ,
AUq,q1,cl (U\Z) a Dolbeault representative of Uq ({f1 , . . . , fq }) as a cohomology
class in H q1 (U\Z, AUq, ), and
AUnq,nq (U)c .
Our goal is to show that Z
h ([]), iU = .
Z
Here we recall that [] HDq,q1(U\Z) is the class induced by , and : HDq,q1(U\Z)

q,q q,q
HZ,D (U) is the coboundary map, in which the target HZ,D (U) is a subspace of DUq,q,cl (U).
As Z is a closed Zariski subset of U of codimension q, the image of AUnq,nq (U)c under d

is in AUnq,nq+1,cl(U\Z)c . By denition, the following diagram
h , iU
HDq,q1(U\Z) AUnq,nq+1,cl (U\Z)c / R
O
d

h , iU
DUq,q,cl (U) AUnq,nq (U)c / R
is commutative. Therefore, we have
Z
h ([]), iU = d .
U \Z
Thus it suces to show that Z Z

d = .
U \Z Z
Obviously, the equality does not depend on the choice of the Dolbeault representative.
Step 4. Let Ui U be the nonvanishing locus of fi . Then we have an open covering
T
U = {Ui } of U\Z, where Ui = Uian . For I {1, . . . , q}, put UI = iI Ui .
Let us recall the construction of a Dolbeault representative . We inductively construct
elements i H qi1 (U\Z, AUq,i,cl ) represented by an (alternative) closed ech cocycle
i = {i,I AUq,i,cl (UI ) | |I| = q i}
for i = 0, . . . , q 1. The class 0 is simply
{0,{1,...,q} = Uq ({f1 , . . . , fq }) AUq,0,cl (U{1,...,q} )}.
Suppose that we have i1 for some 1 i q 1. By Poincar lemma, we have an exact
sequence
0 AUq,i1,cl AUq,i1 AUq,i,cl 0.
As AUq,i1 is a ne sheaf, the ech cohomology H qi(U , AUq,i1) is trivial. Thus there exists
i = {i,J AUq,i1 (UJ )||J| = q i} with U i = i1 , where U denotes the ech dierential
for the covering U . Now we set i = d i := {d i,J AUq,i,cl (UJ ) | |J| = q i}. The last
closed ech cocycle q1 = {q1,{i} AUq,q1,cl (Ui ) | i = 1, . . . , q} is simply a Dolbeault
representative of Uq ({f1 , . . . , fq }) H q1 (U\Z, AUq, ).
For > 0 and I {1, . . . , q}, put
VI = {x U | fi (x) D(0, ), i I; fj (x) D(0, ), j 6 I},
and U = U\V . Here, D(0, ) is the closed disc of radius with center at zero, and U\V is
the closure of U\V in U. As d AUnq,nq+1(U\Z)c , there is a real number 0 > 0 such
that d = 0 on V0 . Thus for every 0 < < 0 , we have
Z Z Z Z

(5.2) d = d = d ( ) = d (q1 ).
U \Z U \V U \V U
Since U is a closed subset of U, the forms and hence have compact support on U .
Step 5. Now we have to use integration on boundaries VI and the corresponding Stokes
formula. We use the formulation of boundary integration through contraction as in [Gub13,
log | |
2]. We consider rst a tropical chart tropW : W (Gan m,K )
N
RN , where W is an
open subset of U . Since V is a K -analytic space of dimension n |I| for some extension
I I
30 YIFENG LIU
KI /K of non-Archimedean elds, the image I := tropW (W VI ) consists of closed faces of

codimension |I| of tropW (W ). For every i I, we choose a tangent vector i for the closed
face {i} of of codimension 1, as dened in [Gub13, 2.8].
Suppose that I = {m1 , . . . , mj } where 1 m1 mj q. If is an (n, n i)-
superform on W with compact support, then we dene
Z Z
:= h; m1 , . . . , mj i{1,...,j} .
I I
It is easy too see that the above integral does not depend on the choice of i ; however,
R
it
does depend on the order. We may patch the above integral to dene the integral VI for
an (n, n |I|)-form on VI with compact support. The negative signs for i ensure that
we have the following Stokes formula
Z X Z
d = (1)(j,I{j}) I{j}

VI j6I V
for an (n, n |I| 1)-form on VI with compact support, for |I| 1. Here, (j, J) is the
position from the rear of the index j when J is ordered in the usual manner. However, for
the initial Stokes formula, we have
Z Z X Z
d = =
U U I
|I|=1 V
for an (n, n 1)-form on U with compact support.

In particular, we have
Z Z X Z
(5.3) d (q1 ) = q1 = q1,I .
U U I
|I|=1 V
In general, for 1 i q 1, we have

XZ X Z
qi,I = d qi,I
VI VI
|I|=i |I|=i
X Z
= d (qi,I )
VI
|I|=i
X X Z
(j,I{j})
= (1) I{j}
qi,I
V
|I|=i j6I
X Z X
= (1)(j,J) qi,J\{j}
VJ jJ
|J|=i+1
X Z
= (U qi )J
J
|J|=i+1 V
X Z
= q(i+1),J .
VJ
|J|=i+1
Combining with (5.2), (5.3), we have

Z Z Z
(5.4) d =
{1,...,q}
0,{1,...,q} = {1,...,q}
Uq ({f1 , . . . , fq })
U \Z V V
for every 0 < < 0 .

Step 6. By (5.4), the theorem is reduced to the formula

Z Z
(5.5) {1,...,q}
Uq ({f1 , . . . , fq }) =
V Z
for suciently small > 0. We may choose a nite admissible covering of U by anoid
Nk log | |
domains Wk , a tropical chart tropWk : Wk (Gan m,K ) RNk , an (n q, n q)-
superform k on tropWk (Wk ) whose support is contained in the interior of tropWk (Wk ), such
P
that = k tropWk k . It suces to check (5.5) on each Wk . Now we x an arbitrary k and
suppress it from notation. Suppose that the moment morphism W (Gan m,K )
N
is dened

by functions g1 , . . . , gN OU (W ). To check (5.5), we may assume that the morphism
(f1 , . . . , fq ) : W (AqK )an is purely of relative dimension n q and W0 6= , where W0 is
the ber over the origin. Put W = W (V \Z).
Applying [CLD12, Proposition 4.6.6] successively, we know that there is some > 0, such
that tropW (W )n is isomorphic to tropW (W0 )nq [ log , +)q . Here, for a polyhedral
complex C of dimension n, we denote by Cn the union of all polyhedra of dimension n.
Therefore, (5.5) follows for every 0 < < , as on tropW (W ) we may take i to be x i ,
where (x1 , . . . , xq ) is the natural coordinate on [ log , +)q .
Corollary 5.9. Let K be a non-Archimedean field and X a proper smooth scheme over K
of dimension n. Then for every algebraic cycle Z of X of dimension 0, we have
Z
clA (Z) = deg Z.
X an
The last result in this section establishes the relation of maps Xq and qX .
Theorem 5.10. Let K be a non-Archimedean field embeddable into Cp , and X a smooth
K-analytic space. Then ker Xq = ker qX . In other words, we have a canonical isomorphism
TXq LXq .
Proof. It suces to check the equality on stalks. Thus we x a point x X with s = s(x)
and t = t(x).
P
Step 1. Let U be an open neighborhood of x. Take an element F = N i=1 ci {fi1 , . . . , fiq }
q
KX (U) where ci Q and fij OX (U). By [Ber07, Propositions 2.1.1, 2.3.1], Knneth
formula, and (the proof of) Theorem 1.1 (ii), there exist
a proper strictly semi-stable scheme S over k of dimension s, where k is a nite
extension of Qp ;
an irreducible component E of Ss that is geometrically irreducible;
an open neighborhood W of 1 ELe in SLan where L is a nite extension of K containing
k;
a closed subset Z of dimension at most s 1 of Sk ;
Q
a point y V := D tk=1 B(0; rk , Rk ) W which projects Q
to E in W ;
a morphism : V U that is tale away from D tk=1 B(0; rk , Rk ) (W ZLan ),
such that (y) = x;
for each i, j, integers dij1, . . . , dijt and gij O (W, 1 ELe), such that
Q d

dfij
d gij tk=1 Tk ijk
Q d
fij gij tk=1 Tk ijk
32 YIFENG LIU
is an exact 1-form on V . Here, Tk is the coordinate function on B(0; rk , Rk ) for

1 k t, which will be regarded as a function in O (V ) via the obvious pullback;
and : V W is the projection morphism.
Q d
In particular, if we put hij = gij nk=1 Tk ijk , then | fij h1
ij | is equal to a constant cij R>0
on V .
Step 2. We dene three tropical charts as follows.
The rst one uses fij (1 i N, 1 j q), which induce a moment morphism
log | |
U (Gan
m,K )
Nq
, and thus a tropicalization map tropU : U (Gan m,K )
Nq
RN q .
The second one uses functions gij (1 i N, 1 j q), which induce a mo-
ment morphism W (Gan m,L )
Nq
, and thus a tropicalization map tropW : W
log | |
(Gan
m,L )
Nq
RN q .
The third one uses functions Tk (1 k t) and gij (1 i N, 1 j q),
which induce a moment morphism V (Gan m,L )
t+N q
, and thus a tropicalization map
log | |
tropV : V (Gan
m,L )
t+N q
Rt+N q .
We have a commutative diagram
tropW
WO / RNO q

tropV
V / Rt+N q

tropU

U / RN q
Pt
in which sends a point (xk , xij ) Rt RN q to (yij ) where yij = log cij + yij + k=1 xk ,
and is the projection onto the last Nq factors. Note that
N
X q
^
Xq (F ) = ci dyij ,
i=1 j=1
and thus !
N
X q
^ t
X
Xq (F )
= ci dxij + dijk dxk
i=1 j=1 k=1
P
as a q-form on Rt+N q . We may write Xq (F ) = I{1,...,t},|I|q dxI I for some (q |I|)-
form I on RN q .
Step 3. We show that (ker qX )x (ker Xq )x . Thus we assume that qX (F ) is an exact
q-form on U and we need to show that Xq (F ) = 0 on a possibly smaller open neighborhood
of x. It suces to that Xq (F ) = 0 when restricted to tropV (V ). This is true as, by
Proposition 4.2, we have that I = 0 when restricted to tropW (W ) for every I.
Step 4. We show that (ker Xq )x (ker qX )x . Thus we may assume that Xq (F ) = 0 when
restricted to tropU (U) and we need to show that qX (F ) is an exact q-form on a possibly
smaller open neighborhood of x. Then Xq (F ) = 0 when restricted to tropV (V ), and thus
I = 0 when restricted to tropW (W ) for every I. By Proposition 4.2, the image of qX (F )
q
in HdR (V ) is 0 after possibly replacing W by a smaller open neighborhood of 1 DLe, as the
q
map q (4.1) is injective on Hrig
0
(E /k)2q . In particular, there is an open neighborhood V of y
in V such that the induced morphism : V U is nite tale where U U is the image of
|V , and qX (F )|V = d for some (q 1)-form on V . Thus qX (F )|U = deg(|V )1 d
where is the trace of along : V U . The theorem follows.
6. Cohomological triviality
In this section, we study the relation between algebraic de Rham cycle classes and tropical
cycle classes.
In this section, sheaves like OX , cX , qX , and the de Rham complex (X , d) are understood
in the analytic topology. We x an embedding R Cp throughout this section. Moreover,
we have to use adic topology. By [Sch12, Theorem 2.24], we may associate to a K-analytic
space X an adic space X ad , and we have a canonical continuous map X : X ad X of
topological spaces.
Lemma 6.1. Let K be a non-Archimedean field embeddable into Cp , and X a smooth K-
analytic space. Then the canonical map LXq Q cX qX is an isomorphism for every
q 0.
Proof. By denition, it suces to show that the map LXq Q cX qX is injective on stalks.
P q
Thus we x a point x X with s = s(x) and t = t(x). Take an element M l
l=1 bl X (F )
q q
LX (U) Q cX (U) such that F = 0 in X (U), where U is a connected open neighborhood
of x, and bl cX (U), F l KXq (U). It suces to show that possibly after shrinking U, the
elements qX (F l ) are linearly dependent in qX (U) over Q.
P l l l
Write F l = N l l l
i=1 ci {fi1 , . . . , fiq } where ci Q and fij OX (U). We copy Step 1 of the
PM
proof of Theorem 5.10 to the element F := l=1 bl F l . Then for every I {1, . . . , t} with
|I| q, we have that

M
X Nl
X X Y ^ M
(6.1) bl cli dli(k)k cl div gijl 0
Hrig (EJ /L)
l=1 i=1 kI j6im J,|J|=q|I|
vanishes, for some nite extension of non-Archimedean elds L/cX (U). Here, is taken over
V
all injective maps I {1, . . . , q}; the multi-wedge product j6im div gijl is taken in the
increasing order for the index j; and {1} is determined by . Note that Hrig 0
(EJ /L) is

canonically isomorphic to Q0 (EJ ) Q L, and for every l,

Nl
X X Y ^ M
cli dli(k)k cl div gijl Q0 (EJ ) .
i=1 kI j6im J,|J|=q|I|
Thus, there exist bl Q, not all being zero, such that (6.1) vanishes for every I if we replace
bl by bl .
This implies that there is an open neighborhood V of y in V such that the induced
morphism : V U is nite tale where U U is the image of |V , and qX (F ) = d
P q q
for some q1 (V ) where F = M 1
l=1 bl X (F ). Then X (F ) = deg() d where is
l

the trace of along : V U . The lemma follows.
The following theorem shows the niteness of HA1,1 and studies the tropical cycle class of
line bundles.
Theorem 6.2. Let X be a proper smooth scheme over Cp . Then
(1) HA1,1(X an ) is finite dimensional;
34 YIFENG LIU
2
(2) for a line bundle L on X whose (algebraic) de Rham Chern class cldR (L) HdR (X )
is trivial, we have clA (L) = 0.
Proof. We put X = X an . By Theorem 1.1 and Lemma 6.1, we know that H 1 (X, LX1 Q Cp )
H 1 (X, LX1 ) Q Cp is a direct summand of H 1 (X, 1,clX /dOX ).
For (1), it suces to show that dimCp H 1 (X, 1,cl X /dOX ) < . In fact, we have a spec-
tral sequence Er abutting to HdR (X) = H (X, X ) with the second page terms E2p,q =
p,q
H p (X, q,cl q1
X /dX ). Thus, it suces to show that both H (X, Cp ) and H (X, X ) are nite
3 2
dimensional. Since the homotopy type of X is a nite CW complex, dimCp H (X, Cp ) <
i
for every i Z. By GAGA, H i (X, X ) is canonically isomorphic to the algebraic de Rham

cohomology HdR i
(X ) for every i, and thus nite dimensional.
For (2), note that the map clA : CHq (X )Q HAq,q (X) factors through H q (X, LXq ). We
denote by cl(L) the corresponding class in H 1 (X, LX1 ). It suces to show that cl(L) is zero
in H 1 (X, 1,cl
X /dOX ). Now we regard cl(L) as an element in the latter cohomology group.
Note that cl(L) maps to zero under the coboundary map : H 1 (X, 1,cl 3
X /dOX ) H (X, Cp ),
1 3
as the composite map CH (X )Q H (X, Cp ) ts into the following commutative diagram
CH1 (X )Q / H 1 (X , 1,cl
X /dOX )
/ H 1 (X, 1,cl
X /dOX )

H 3 (X , Cp ) / H 3(X, Cp )
in which H 3 (X , Cp ) vanishes. Thus, it suces to show that the image of cl(L) vanishes under
the map ker( : H 1 (X, 1,cl 3 2 2
X /dOX ) H (X, Cp )) H (X, X )/H (X, Cp ). However, by
comparing the denitions of two cycle class maps, we know that it is also the image of
cldR (L) under the composite map HdR 2
(X ) H 2 (X, X ) H 2 (X, X )/H 2(X, Cp ), thus
vanishes.
The following lemma is the analytic version of the corresponding statement in the algebraic
setting.
Lemma 6.3. Let K be a non-Archimedean field. Let X be a geometrically connected proper
n1
smooth scheme over K of dimension n. Then we have H n (X an , nX an /dX an ) K.
Proof. Put X = X an . By the spectral sequence E2p,q = H p (X, q,cl q1

X /dX ) H
p+q
(X, X )

and the GAGA comparison isomorphism HdR (X) HdR (X ), it suces to show that
n1
H (X, F ) = 0 for i > n and every abelian sheaf F on X. Then we have H n (X, nX /dX
i
)
2n
HdR (X ) K.
By [Ber93, Proposition 1.3.6 & Lemma 1.6.2] and [Sch12, Theorem 2.21], we have a canon-
1
ical isomorphism H i (X, F ) H i (X ad , X F ).
In fact, we will show that H (X , F ) = 0 for i > n and every abelian sheaf F on X ad .
i ad
Recall that a formal model of X is a proper at formal K -scheme X with an isomorphism

X X. A formal model X induces a continuous map X : X ad X. By [Sch12, Theorem
2.22] and [SP, 094L, 0A2Z], we have an isomorphism X ad limX X of spectral spaces, where

the (coltered) limit is taken over all formal models X of X. By [SP, 0A37], we have an
isomorphism limX H i(X, X F ) H i(X ad , F ). Now as (the underlying space of) X is a

Noetherian topological space of dimension (at most) n, it follows that H i (X, XF ) = 0 for
i > n by Grothendieck vanishing theorem [SP, 02UZ]. The lemma then follows.
Definition 6.4. Let X be a compact smooth Cp -analytic space of dimension n. We have a

total integration map
Z
: H n (X, TXn Q R) HAn,n (X) R.
X
By the isomorphism LXn in Theorem 5.10, Lemma 6.1, and by extending the above
TXn
map linearly over Cp , we obtain a Cp -linear map
TrA n n n n
X : H (X, X ) H (X, LX Q Cp ) Cp ,
called the trace map for X.

The following proposition can be regarded as certain algebraicity property of the tran-
scendental map of integration.
Proposition 6.5. Let k Cp be a discrete non-Archimedean subfield, and X a geometrically
connected proper smooth scheme over k of dimension n. If we put Xa = X k Cp , then the
map TrAXaan factors through the canonical map
H n (Xaan , nXaan ) H n (Xaan , nXaan /dX

n1
a
an ) Cp ,
where we have used Lemma 6.3 for the last isomorphism.

n1
In particular, H n (Xaan , (nXaan /dXa
an )w ) Cp if w = 2n, and is trivial otherwise.
Proof. We put X = Xaan . Dene nX to be the quotient sheaf in the following exact sequence
0 nX (nX /dX
n1
)2n nX 0.
It is functorial in X. The best hope is that nX is trivial; but we do not know so far. However,
one can show that nX is supported on {x X | s(x) 2}. This suggests that one should
expect H i (X, nX ) = 0 for i n1, which suces for the proposition. In fact, such vanishing
result can be proved if we have semi-stable resolution instead of alteration. In the absence
of semi-stable resolution, we need an ad hoc argument.
We may assume that X is projective, as we will eventually take an alteration of X . Take
a cohomology class H n1 (X, nX ). Since X is (Hausdor and) compact, by [SP, 09V2,
01FM], there is a nite open covering U = {Ui | i = 1, . . . , N} of X such that is represented
by an (alternative) ech cocycle = {I nX (UI ) | I {1, . . . , N}, |I| = n} on U, where
T
UI = iI Ui as always. By rening U , we may assume that I is in the image of the
map (nX /dX n1
)(UI )pre
2n X (UI ) for every I (See Denition 3.8 for the notation). By
n
[Pay09, Theorem 4.2], taking blow-ups, and possibly taking a nite extension of k inside
Cp , we have a (proper at) integral model Y of X such that if Z1 , . . . , ZM are all reduced
irreducible components of Ys , then the covering { 1 Zi b k Cp | i = 1, . . . , M} renes U . By
[dJ96, Theorem 8.2], possibly after taking further nite extension of k inside Cp , we have a
(proper) strictly semi-stable scheme Y over k with an alteration Y Y. For simplicity,
we may also assume that every irreducible component of Ys[t] (0 t n) is geometrically
irreducible. In particular, if we denote by Z1 , . . . , ZM
all irreducible components of Ys , then
1 b 1 1
the covering { Zi k Cp | i = 1, . . . , M } renes f U := {f Ui | i = 1, . . . , N}. We x an
index function : {1, . . . , M } {1, . . . , N} for the renement; in other words, 1 Zi b k Cp
1
is contained in f U(i) .
We x a uniformizer of k, and put X = (Y k Cp )an . We claim that f = 0, where
f is the canonical image of f 1 in H n1(X , nX ).

36 YIFENG LIU
For every 1 i M and 0 < < 1, we denote by Ui () the open subset of 1 Zi

b k Cp

X as in Step 3 in the proof of Lemma 3.11. Then we have that
[ \
1 Zi
b k Cp = Ui (), 1 (Zi \Xs[1] )
b k Cp = Ui ().
0<<1 0<<1
By denition, U () := {Ui () | 1 i M } form an open covering if and only if > 1/2.

Fix a real number 1/2 < < 1. We study a typical n-fold intersection of U(). Without lost
T T
of generality, we consider U{1,...,n} () := ni=1 Ui (). If ni=1 Zi = , then U{1,...,n}() = . So
Tn F
we may assume that i=1 Zi = Ll=1 Cl , where each Cl is a geometrically irreducible proper
smooth curve over k. e
Take a typical member C of {C1 , . . . , CL } and put UC () = U{1,...,n} () 1 C
b k Cp . Recall
the k-analytic space E dened in Example 2.2. Let E () E be the subspace such
n1 t t

that |Ti | < ||1 for every 0 i t. By [GK02, Lemma 3], the canonical map HdR (Et )
[n] 1 b
HdR (E ()) is an isomorphism. Put C = C\Xs and UC () = UC () C k Cp . By the
t
proof of [GK02, Theorem 2.3], we have isomorphisms

HdR (UC (), UC ()) Tot(HdR
n1
(E
()) k Hrig (C /k)) k Cp

Tot(HdR n1
(E ) k Hrig (C /k)) k Cp
of graded Cp -vector spaces. By Theorem 1.1 (ii), there is an open neighborhood U1 of UC ()
in UC () such that the image of f 1 {(1),...,(n)} |U1 in an X (U1 ) zero. Put U2 := UC ()
1 (C\C )
b k Cp . Then H (U2 ) is isomorphic to a nite copy of H (En ) k Cp , and in
dR dR
particular, the image of f 1 {(1),...,(n)} |U2 in an
X (U2 ) is zero. Finally, note that UC () =
U1 U2 , which implies that f = 0.
Going back to X, we have the following commutative diagram
TrA
X
H n1 (X, nX ) / H n (X, nX ) / Cp
f f deg(f )
TrA
X
H n1(X , nX ) / H n (X , nX ) / Cp ,
where the right vertical arrow is the multiplication by deg(f ). Therefore, TrA X factors through
n1
the map H n (X, nX ) H n (X, nX /dX ).
The last statement follows from the combination of
TrAX is surjective as one can write down an (n, n)-form on X with nonzero total
integral;
n1 L n1
H n (X, nX /dX ) = wZ H n (X, (nX /dX )w ) by Theorem 1.1;
n1
H (X, X /dX ) Cp by Lemma 6.3; and
n n
n1 n1
the image of H n (X, nX ) H n (X, nX /dX ) is contained in H n (X, (nX /dX )2n ).

The next theorem shows that algebraic cycles that are cohomologically trivial in the al-
gebraic de Rham cohomology are cohomologically trivial in of Dolbeault cohomology of
currents as well.
Theorem 6.6. Let k Cp be a finite extension of Qp and X a proper smooth scheme over
k of dimension n. Let Z be an algebraic cycle of X of codimension q such that cldR (Z) = 0.
If we put Xa = X k Cp and Za = Z k Cp , then clD (Za ) = 0, that is,

Z
=0
Zaan
for every d -closed form A nq,nq (Xaan ).
We need some preparation before the proof of the theorem. We start from the following
lemma.
Lemma 6.7. Let the assumption and notation be as in Theorem 6.6. Then for every i and
q, the canonical map
q q
lim H i (Xkan i an
, TX an ) H (Xa , TX an )

k a
k
is an isomorphism, where the colimit is taken over all finite extensions k of k in Cp .
Proof. We have an isomorphism of spectral spaces Xaad limk Xkad . Thus by [SP, 0A37], it
1 1 q
1 q
suces to show that the canonical map limk k X an TX an Xaan TXaan is an isomorphism,
k k
where k : Xaad Xkad
is the canonical map. However, this follows from the fact that for every
f O (Xaan , V ) where V is a rational anoid domain, there is a function g O (Xkan

,V )
1 1
for some k such that k (V ) = V and f k g has norm 1 on some open neighborhood of
V . Here, we have used [Ber07, Lemma 2.1.3 (ii)].
We review some facts about cup products from [SP, 01FP]. Let X be a topological space,
k a eld, n 0 an integer. Let be a sheaf of k-vector spaces on X. Suppose that we have
two bounded complexes F , G of sheaves of k-vector spaces on X, with a map of complexes
of sheaves of k-vector spaces
: Tot(F k G ) [n].
Then we have a bilinear pairing
: H i (X, F ) H 2ni(X, G ) H 2n (X, [n]) = H n (X, )
for every i Z. Now suppose that we have four bounded complexes Fj , Gj (j = 1, 2) of

sheaves of k-vector spaces on X, maps 1 : F1 F2 , 2 : G2 G1 , and j : Tot(Fj k
Gj ) [n] (j = 1, 2), such that 1 (idF1 2 ) = 2 (1 idG2 ). Then we have the
following commutative diagram
1
(6.2) H i (X, F1 ) H 2ni (X, G1 ) / H n (X, )
O
H i (X,1 ) H 2ni (X,2 )
2
H i (X, F2 ) H 2ni (X, G2 ) / H n (X, )
for every i Z.
Proof of Theorem 6.6. Without lost of generality, we may assume that X is geometrically
irreducible over k (of dimension n). Put X = X an .
38 YIFENG LIU
Step 1. By Proposition 5.5 and Theorem 5.10, we have the following commutative dia-
gram

HAq,q (Xaan ) HAnq,nq (Xaan ) / HAn,n (Xaan )

H q (Xaan , TXqaan Q R) H nq (Xaan , TXnq
a
an Q R)
/ H n (Xaan , TXnaan Q R)

H q
(Xaan , q,cl q1
Xaan /dXaan ) H nq
(Xaan , nq,cl nq1
Xaan /dXaan ) / H n n1
(Xaan , nXaan /dXa
an ),
in which the rst cup product is induced by the wedge product of real forms.
To prove the theorem, it suces to consider an arbitrary element H nq (Xaan , TXnq
a
an ).
In view of Lemma 6.7, after replacing k by a nite extension in Cp , we may assume that
H nq (X, TXnq ).
Note that the tropical cycle class map clA (Denition 5.7) factors as
CHq (X )Q H q (X, TXq ) H q (Xaan , TXqaan ) H q (Xaan , TXqaan Q R) HAq,q (Xaan ),
in which we denote the rst map by clT . By Theorem 5.8 and Proposition 6.5, it suces
n1
to show that the image of clT (Z) in H n (X, nX /dX ), which is isomorphic to k by
Lemma 6.3, is zero. Denote by the image of clT (Z) in H (X, q,cl
q q1
X /dX ), and regard
as in H nq (X, nq,cl
X /dnq1
X ). In fact, we can prove that = 0 if we have semi-stable
resolution instead of alteration. In the absence of semi-stable resolution, we need an ad hoc
argument.
Step 2. To proceed, we need the adic topology of X. Recall that we have a continuous
map X : X ad X. Let (X ad , d) the de Rham complex on X ad . Then we have a canonical
1
map X (X , d) (X ad , d) of complexes of sheaves of k-vector spaces on X ad . Denote by
ad (resp. ad ) the image of (resp. ) under the canonical map
H i(X, i,cl i1 i ad i,cl i1

X /dX ) H (X , X ad /dX ad )
for i = q (resp. i = n q). Note that when i = n, the above map is an isomorphism, by the
same argument for Lemma 6.3.
We claim that there exists an alteration f : X X possibly after a nite extension of k
in Cp , such that f ad is in the image of the canonical map
H 2n2q (X ad , nq X ad ) H nq (X ad , nq,cl nq1

X ad /dX ad ),
where X = X an .
Assuming the above claim, we deduce the theorem as follows. Applying (6.2) to X ad and
n1
the sheaf := nX ad /dX ad , we obtain the following commutative diagram
H q (X ad , q,cl q1
X ad /dX ad ) H
nq
(X ad , nq,cl nq1
X ad /dX ad )
/ H n (X ad , )
O
1 2

H 2q (X ad , q X ad ) H 2n2q (X ad , nq X ad ) / H n (X ad , )
O
1 2

H 2q (X ad , X ad ) H 2n2q (X ad , X ad ) / H n (X ad , )
in which the maps among various complexes of sheaves are dened in the obvious way. By
the above claim, there exists H 2n2q (X ad , nq X ad ) such that 2 ( ) = f ad . Thus,
f ad f ad = f ad 2 ( ) = 1 (f ad ) = 1 (cldR (f Z)) = cldR (f Z) 2 ( ),
where we regard cldR (f Z) as an element in H 2q (X ad , X ad ) under the comparison map
(which is in fact an isomorphism)
2q
HdR (X ) = H 2q (X , X ) H 2q (X ad , X ad ).
As cldR (Z) = 0, we have cldR (f Z) = 0 and hence f ad f ad = 0. Thus, f f = 0,
and in particular, Z Z
1
= deg(f ) f = 0.
Zaan (f Z)an
a
The theorem is proved.

q
Step 3. Now we fucus on the claim in Step 2. For an integral model Y of X, dene KX,Y
to be the sheaf on Ys associated to the presheaf
U 7 lim KqM (OX (U)) Q, U Ys

1
U U
where the colimit is taken over all open neighborhoods U of 1 U in X. We remark that
q
there is a canonical morphism KX,Y YX 1
KXq which is in general not an isomorphism,
where Y : X ad Ys is the induced continuous map. Put ,q 1 q
X,Y = Y X X . Then we have
a complex of sheaves of k-vector spaces (,
X,Y , d) on Ys . And we have a canonical map
qX,Y : KX,Y
q
,q,cl ,q1
X,Y /dX,Y
q
similar to Denition 5.4. Denote by LX,Y the image sheaf of qX,Y in the above map. Since
sheacation commutes with pullback and taking colimit, we have a canonical isomorphism
q
lim Y1 KX,Y 1
X KXq

Y
of sheaves on X , where the ltered colimit is taken over all integral models Y of X. On
ad
the other hand, we have an obvious isomorphism

lim Y1 , 1
X,Y X X .

Y
Passing to the quotient, we have a canonical isomorphism

q
lim Y1 LX,Y 1
X LXq .

Y
40 YIFENG LIU
Note that originally, belongs to H nq (X, TXnq ) H nq (X, LXnq ). By a similar argu-
ment for Lemma 6.3, there is an integral model Y of X such that is in the image of the
canonical map
nq
H nq (Ys , LX,Y ) H nq (X ad , 1 LXnq ) H nq (X, LXnq ).
By [dJ96, Theorem 8.2], possibly after taking further nite extension of k inside Cp , we
have a projective strictly semi-stable scheme Y over k with an alteration f : Y Y. Put
X = Y an . If we put qX ,Y = Y qX ad , then (X ,Y , d) is a complex of sheaves of k-vector
spaces on Ys and we have a canonical map (,
X ,Y , d) (X ,Y , d). The claim in Step 2
will follow if we can show that the composite map
,nq,cl ,nq1
(6.3) H nq (Ys , LXnq
,Y ) H
nq
(Ys , X ,Y /dX ,Y )
H nq (Ys , nq,cl nq1
X ,Y /dX ,Y ) H
2n2q+1
(Ys , nq1X ,Y )
is zero. The advantage of (X ,Y , d) is that the entire complex admits a canonical Frobenius
action. More precisely, we x a uniformizer of k; let Ys be the log scheme Ys equipped
with log structure as in [HK94, (2.13.2)]; and Spf W (k) e be the formal log scheme Spf W (k)
e
equipped with log structure 1 7 0. Here, we use Zariski topology in the construction of log
schemes and log crystal sites instead of tale one in [HK94]. There is a canonical morphism
e )
u : (Ys / Spf W (k)
log-cris Ys of sites. Then by (the proof of) [HK94, Theorem 5.1], we
have a canonical isomorphism
Ru Olog-cris W (ek) k (X ,Y , d)
Ys / Spf W (e

k)
in the derived category of abelian sheaves on Ys , where OYlog-cris

/ Spf W (e
k)
denotes the structure
s
sheaf in the log crystal site. Since OYs / Spf W (ek) admits a Frobenius action over Spec k, e we
obtain a Frobenius action on the entire complex (X ,Y , d) in the derived category.

For w Z, denote by (q,cl q1 q,cl q1
X ,Y /dX ,Y )w the maximal subsheaf of X ,Y /dX ,Y generated
by sections of generalized weight w. We claim that
(a) the image of the canonical map KXq ,Y LXq ,Y q,cl q1
X ,Y /dX ,Y is contained in the
subsheaf (q,cl q1
X ,Y /dX ,Y )2q for every q;
(b) the image of the canonical map ,q,cl ,q1 q,cl q1
X ,Y /dX ,Y X ,Y /dX ,Y is contained in the
L
subsheaf 2q q,cl q1
w=0 (X ,Y /dX ,Y )w for every q.
Then the triviality of the map (6.3) follows easily from an argument of spectral sequences. In
fact, we have a spectral sequence Erp,q abutting to H (Ys , X ,Y ) equipped with a Frobenius
action, such that E2p,q = H p (Ys , q,cl q1
X ,Y /dX ,Y ). By (a) and (b), the restriction of all
dierentials dnq,nq
r in the spectral sequence with r 2 to the image of the map
H nq (Ys , LXnq
,Y ) H
nq
(Ys , nq,cl nq1 nq,nq
X ,Y /dX ,Y ) = E2
is zero by weight consideration. Thus, (6.3) is the zero map.

Step 4. The last step is devoted to the proof of the two claims (a) and (b) in Step 3. We
remark that they are not formal consequences of Theorem 1.1.
By denition, q,cl q1 q 1
X ,Y /dX ,Y is the sheaf on Ys associated to the presheaf U 7 HdR ( U),
,q,cl ,q1 q
and X ,Y /dX ,Y is the sheaf on Ys associated to the presheaf U 7 HdR (X , 1 U) by
[Ber07, Lemma 5.2.1]. We check (a) and (b) on stalks and thus x a point x Ys .
To prove (a), it suces to consider the case where q = 1. Let U Y be an open

ane neighborhood of x. Take f O (X , 1 U); we have to show that the image of df f
in
1
HdR ( Us ) is of generalized weight 2 for a possibly smaller open neighborhood U of x. First,
1
we may replace f by the restriction of an (algebraic) function f O(Uk ) without changing

the image of df f
in HdR
1
( 1 Us ). Since Y is projective, we may choose a closed embedding
Y PN k into a projective space. Choose an open ane neighborhood V of x in Pk such
N

that V Y U and f |VY = g|VY for some g O (Vk ). Since Pk is smooth, by Remark
N
3.7, dgg
belongs to Hrig1
(Vs /k)2 and thus its image in HdR 1
( 1 Vs ) is of generalized weight 2.
By functoriality of log crystal sites for the morphism Y PN k , we conclude that the image
df
of f in HdR ( (V Y )s ) is of generalized weight 2. Here, (V Y )s is the inverse image
1 1 1
in X .
Claim (b) is a consequence of [GK05, Theorem 0.1] and [Chi98, Theorem 2.3]. In fact,
we have a functorial map of spectral sequences Erp,q Erp,q abutting to HdR
(X , 1 U)
1
HdR ( U) with the rst page being

E1p,q = Hrig
q e )
(Us(p) / Spf W (k) k E1p,q = Hlog-cris
q e )
(Us(p) / Spf W (k) k,
W (e
k )[1/p] W (e
k)
where Us(p) is the disjoint union of irreducible components of Us[p] , equipped with the induced
log structure from Ys . By [GK05, Theorem 3.1, Lemma 4.6] and [Chi98, Theorem 2.3], we
know that the weights of (the nite dimensional k-vector space) E1p,q are in the range [q, 2q],
q
and thus the weights of HdR (X , 1U) are in the range [0, 2q].
References
[SP] The Stacks Project Authors, Stacks Project. available at http://math.columbia.edu/algebraic_geometry/stacks
34, 35, 37
[Ber90] V. G. Berkovich, Spectral theory and analytic geometry over non-Archimedean fields, Mathemat-
ical Surveys and Monographs, vol. 33, American Mathematical Society, Providence, RI, 1990.
MR1070709 5
[Ber93] , tale cohomology for non-Archimedean analytic spaces, Inst. Hautes tudes Sci. Publ.
Math. 78 (1993), 5161 (1994). MR1259429 5, 34
[Ber96] , Vanishing cycles for formal schemes. II, Invent. Math. 125 (1996), no. 2, 367390, DOI
10.1007/s002220050078. MR1395723 5
[Ber04] , Smooth p-adic analytic spaces are locally contractible. II, Geometric aspects of Dwork
theory. Vol. I, II, Walter de Gruyter GmbH & Co. KG, Berlin, 2004, pp. 293370. MR2023293 1
[Ber07] , Integration of one-forms on p-adic analytic spaces, Annals of Mathematics Studies, vol. 162,
Princeton University Press, Princeton, NJ, 2007. MR2263704 2, 6, 7, 8, 9, 10, 11, 12, 14, 17, 23,
31, 37, 40
[Bert97] P. Berthelot, Finitude et puret cohomologique en cohomologie rigide, Invent. Math. 128 (1997),
no. 2, 329377, DOI 10.1007/s002220050143 (French). With an appendix in English by Aise Johan
de Jong. MR1440308 6, 7
[Bos81] S. Bosch, A rigid analytic version of M. Artins theorem on analytic equations, Math. Ann. 255
(1981), no. 3, 395404, DOI 10.1007/BF01450712. MR615859 19
[CLD12] A. Chambert-Loir and A. Ducros, Formes diffrentielles ralles et courants sur les espaces de
Berkovich (2012). arXiv:math/1204.6277. 2, 24, 25, 27, 31
[Chi98] B. Chiarellotto, Weights in rigid cohomology applications to unipotent F -isocrystals, Ann. Sci.
cole Norm. Sup. (4) 31 (1998), no. 5, 683715, DOI 10.1016/S0012-9593(98)80004-9 (English,
with English and French summaries). MR1643966 6, 17, 41
[dJ96] A. J. de Jong, Smoothness, semi-stability and alterations, Inst. Hautes tudes Sci. Publ. Math. 83
(1996), 5193. MR1423020 16, 35, 40
42 YIFENG LIU
[GK02] E. Grosse-Klnne, Finiteness of de Rham cohomology in rigid analysis, Duke Math. J. 113 (2002),
no. 1, 5791, DOI 10.1215/S0012-7094-02-11312-X. MR1905392 6, 7, 8, 10, 12, 16, 18, 19, 36
[GK05] , Frobenius and monodromy operators in rigid analysis, and Drinfelds symmetric space, J.
Algebraic Geom. 14 (2005), no. 3, 391437, DOI 10.1090/S1056-3911-05-00402-9. MR2129006 41
[Gub13] W. Gubler, Forms and currents on the analytification of an algebraic variety (after Chambert-Loir
and Ducros) (2013). arXiv:math/1303.7364. 24, 29, 30
[HK94] O. Hyodo and K. Kato, Semi-stable reduction and crystalline cohomology with logarithmic poles,
Astrisque 223 (1994), 221268. Priodes p-adiques (Bures-sur-Yvette, 1988). MR1293974 40
[Jel16] P. Jell, A Poincar lemma for real-valued differential forms on Berkovich spaces, Math. Z. 282
(2016), no. 3-4, 11491167, DOI 10.1007/s00209-015-1583-8. MR3473662 2, 25
[JSS15] P. Jell, K. Shaw, and J. Smacka, Superforms, tropical cohomology and Poincar duality (2015).
arXiv:math/1512.07409. 26
[LS07] B. Le Stum, Rigid cohomology, Cambridge Tracts in Mathematics, vol. 172, Cambridge University
Press, Cambridge, 2007. MR2358812 6
[Pay09] S. Payne, Analytification is the limit of all tropicalizations, Math. Res. Lett. 16 (2009), no. 3,
543556, DOI 10.4310/MRL.2009.v16.n3.a13. MR2511632 35
[Sch12] P. Scholze, Perfectoid spaces, Publ. Math. Inst. Hautes tudes Sci. 116 (2012), 245313, DOI
10.1007/s10240-012-0042-x. MR3090258 33, 34
[Sou85] C. Soul, Oprations en K-thorie algbrique, Canad. J. Math. 37 (1985), no. 3, 488550, DOI
10.4153/CJM-1985-029-x (French). MR787114 24
[Tsu99] N. Tsuzuki, On the Gysin isomorphism of rigid cohomology, Hiroshima Math. J. 29 (1999), no. 3,
479527. MR1728610 20
Department of Mathematics, Northwestern University, Evanston IL 60208, United States

E-mail address: liuyf@math.northwestern.edu

1702 00047

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

1702 00047

Uploaded by

Copyright:

Available Formats

WEIGHT DECOMPOSITION OF DE RHAM COHOMOLOGY SHEAVES

AND TROPICAL CYCLE CLASSES FOR NON-ARCHIMEDEAN SPACES

Abstract. We construct a functorial decomposition of de Rham cohomology sheaves,

of cX -modules in either analytic or tale topology. It satisfies that

for every d -closed form AXnq,nq

Remark 1.6. Let the situation be as in the above theorem.

for every d -closed form A nq,nq ((X K Cp )an ).

for every d -closed (n q, n q)-form on (Z k,p Cp )an .

Conventions and Notation.

Let K be a non-Archimedean eld. For a real number r > 0, we denote by D(0; r)

2. Review of rigid cohomology

and Z Z Ke K f . Then the induced map H (X/K) K H (X /K ) is

where V runs through open neighborhoods of in X. We show that a2 is an isomorphism.

extends canonically to a morphism of formal K -schemes

where A is the coordinate K-anoid algebra of which is isomorphic to A2 . Therefore,

induced from the homomorphism A f Kh1 Z , . ^ f

induced from the homomorphism Kh1 Z1 , . ^

is an isomorphism by [GK02, Lemma 3]. If K is residually algebraic over Qp , then we have

3. Weight decomposition in tale topology

morphism of K -formal schemes X c 1 b c b

is an isomorphism of formal k -schemes, where is a uniformizer of k,

In what follows, if D is of dimension 0, then we suppress it from all notations.

Definition 3.5. Let X be a K-analytic space and x X a point.

: (D2 , (Y2 , D2 ), (D2 , 2 ), W2 , 2 ; y2 ) (D1 , (Y1 , D1 ), (D1 , 1 ), W1 , 1 ; y1)

consists implicitly extensions of related elds K L1 L2 such that k1 k2 , and

for all q, w Z. Note that () needs not to respect each factors.

The following lemma generalizes [Ber07, Proposition 2.1.1].

that the following diagram

Et(X; x) as in (1) with a morphism to (Y , 0 ; y0 ). Now we apply Lemma 3.2 to X , the

(at the very beginning). Thus we have morphisms of formal K -schemes

smooth, and we have Spec Khr 1 ^T, rT 1 i (Gm ) . Moreover,

Now we are ready to dene the desired direct summand (q,cl q1

q,cl q1 pre q,cl q1

is an isomorphism, where L is the algebraic closure of K in K .

for every w Z. Here, f # denotes the canonical map f 1 X Y and induced

For (3), take an element [] in the stalk of (q,cl q1

Proof. We may assume L = K where L is the nite extension of K implicitly contained in

pletion is D , where we recall that is a uniformizer of the discrete non-Archimedean eld

and a morphism V M(D) b k K induced from . We choose a neighborhood U of the

is the inverse of the previous isomorphism. In other words, we have a morphism : W

We obtain a nite extension k /k, an alteration S S k and a k -compactication S S

for some 0 r = ri s; Ei := E Ss is ane; and if we write fi,j for the image of tj in Si ,

Then (Si ) = 1 Ei in Si . For 0 < < 1, denote by Si () the open subset of Si

i g are dened (as [Ber07, Lemma 6.1.1]);

is an isomorphism respecting weights,

is an isomorphism respecting weights.

is a Frobenius lifting of the Frobenius endomorphism of Spec F f over kf = k.e Therefore,

4. Logarithmic differential forms

denoted by [EI ]. We dene a wedge product

sending [EI ] to the canonical generator on (each connected component of) EI .

commutative. From the standard computation in algebraic de Rham cohomology, we have

If the restriction of to HdR1

Since the composition

is isomorphic to the k-vector space Z 1 (E) k, and the boundary map

Q-vector spaces F on Xet . In fact for every x X, we have (Ri F )x = H i (H(x), i1 x F ),

For Property (ii), we have the inclusion qX qXet as subsheaves of q,cl q1

stalk at an arbitrary point x X. Take an element [] in ( 1Xet )x . We may assume that it

5. Tropical cycle class map

evaluation map log | | : (Gan

(2) If X is moreover smooth, then we dene another map of sheaves

as follows. For a symbol {f1 , . . . , fq } KXq (U) with f1 , . . . , fq OX

of Q-vector spaces, and thus a map

By [JSS15, Proposition 3.16], the canonical map