5 Differential forms on algebraic varieties [0365]
Original official author HTML, exact retained edition. Historical TeX conversion verdicts remain unchanged. Cited-edition alignment and mathematical self-containment are not assessed.
Complete original source context · Original author HTML
5 Differential forms on algebraic varieties
On a complex analytic manifold , we use open analytic charts to define -forms on by pull-back. The idea in the non-archimedean setting is similar replacing the above charts by tropical charts from the previous section in order to pull-back Lagerberg’s superforms to .
In this section, is an algebraically closed field endowed with a complete non-trivial non-archimedean absolute value . Let be the associated valuation and let be the value group. The theory could be done for arbitrary fields (see [CD12]), but it is no serious restriction to assume that is algebraically closed as the theory is stable under base extension and in the classical setting, the analysis is also done over . We will introduce -forms on the analytification of a -dimensional algebraic variety over .
5.1
We recall from 4.15 that a tropical chart consists of an open subset of for a very affine open subset of such that for an open subset of . Here, is the canonical moment map. It is a closed embedding to the torus . The tropical variety is a tropical cycle of and is the tropicalization map. The embedding is only determined up to translation by an element in and hence the tropical constructions are canonical up to integral -affine isomorphisms.
Suppose that we have another tropical chart . Then is a tropical chart (see Proposition 4.16) and we get a canonical affine homomorphism of the underlying tori with on (see 4.12). The associated affine map maps the tropical variety onto (use Lemma 4.9). Then we define the restriction of the superform to a superform on by using the pull-back to with respect to . This plays a crucial role in the following definition:
Definition 5.2
A differential form of bidegree on an open subset of is given by a covering of by tropical charts of and superforms such that for every . If is another differential form of bidegree on given by with respect to the tropical charts covering , then we consider and as the same differential forms if and only if for every and . We denote the space of -differential forms on by . As usual, we define the space of differential forms on by .
5.3
It is obvious from the definitions that the differential forms form a sheaf on . Using the corresponding constructions for superforms on tropical cycles, it is immediate to define the wedge product and differential operators , , on differential forms on . By 4.6, we have if .
For a morphism and open subsets (resp. ) of (resp. ) with , we get a pull-back defined in the following way: Suppose that is given by the covering and the superforms as above. Then there is a covering of by tropical charts which is subordinate to . This means that for every , there is with and for the corresponding very affine open subsets. Then is the differential form on given by the covering and the superforms . We leave the details to the reader. This construction is functorial as usual.
Remark 5.4
We obtain the same sheaf of differential forms on as in [CD12], §3. In the latter reference, all analytic moment maps were used to define differential forms on and so it is clear that our differential forms here are also differential forms in the sense of [CD12]. To see the converse, we argue as follows: By Proposition 4.16, tropical charts form a basis in . It follows from Proposition 7.2 that an analytic moment map may be locally in approximated by an algebraic moment map such that in an open neighbourhood of in . Here, is a suitable very affine open subset of with . It follows from [CD12], Lemma 3.1.10, that we may use algebraic moment maps to define differential forms in the sense of [CD12]. Using that factorizes through (see 4.12), we get the claim.
Definition 5.5
Let be a differential form on an open subset of . The support of is the complement in of the set of points of which have an open neighbourhood such that . Let be the space of differential forms of bidegree with compact support in .
Proposition 5.6
Let be a tropical chart of and let be given by . Then in if and only if in .
Proof: See [CD12], Lemme 3.2.2.
Remark 5.7
5.8
In analogy with differential geometry on manifolds, we set for any open subset of and a smooth function on is just a differential form of bidegree . Since tropicalization maps are continuous, it is clear that a smooth function is a continuous function on . By the Stone-Weierstrass theorem, the space of smooth functions with compact support in is a dense subalgebra of (see [CD12], Proposition 3.3.5).
Definition 5.9
Let be an open covering of an open subset of . A smooth partition of unity on with compact supports subordinated to the covering is a family of non-negative smooth functions with compact support on with the following properties:
- (i)
The family is locally finite on .
- (ii)
We have on .
- (iii)
For every , there is such that .
Proposition 5.10
Let be an open covering of an open subset of . Then there is a smooth partition of unity on with compact supports subordinated to the covering .
Proof: It is enough to show that for every , there is a non-negative smooth function with compact support in and with . Since is a locally compact Hausdorff space which is also -compact, the open subset is paracompact and hence standard arguments from differential geometry yield the existence of the desired partition of unity (see [Wa83], Theorem 1.11).
To prove the crucial claim at the beginning of the proof, we may assume that is coming from a tropical chart (see Proposition 4.16). Then is a open subset of with and hence there is an open subset in with . There is a smooth non-negative function on with compact support in such that . Since the tropicalization map is proper, the smooth function has compact support in and hence fulfills the claim.
So far, we have seen properties of differential forms which are completely similar to the archimedean case. The next result of Chambert-Loir and Ducros ([CD12], Lemme 3.2.5) shows that the support of a differential form of degree at least one is disjoint from .
Lemma 5.11
Let be an open subset of . We consider and with . Then .
Proof: Using Proposition 4.16 and shrinking the open neighbourhood of , we may assume that is a tropical chart on which is given by the superform . By Proposition 4.14, there is a very affine open subset of and a compact neighbourhood of in such that is of dimension . By Proposition 4.16, there is a tropical chart with and . By 4.12, there is an affine homomorphism such that . Using the same factorization for the tropicalizations, we see that the restriction of to is given by . The inclusion yields that factorizes through (use 4.12). Since , we get . As , we conclude that . This proves .
Corollary 5.12
Let be an open subset of and let be a Zariski open subset of . If with , then .
Proposition 5.13
Let be a differential form with . Then there is a very affine open subset of such that and such that is given on by a superform .
Proof: By assumption, the support of is a compact subset of . We conclude that there are finitely many tropical charts covering such that is given on by the superform . Recall that is an open subset of . By 4.13, is a non-empty very affine open subset of . We define the open subset of by . Since , Corollary 5.12 yields . Using 4.12, we see that for an affine homomorphism of tori. Then we have
and we denote this open subset of by . It follows that the preimage of with respect to is equal to . We conclude that is a tropical chart of . Note that is given on by . By Proposition 5.6, agrees with on for every and hence they define a superform . By construction, gives the differential form on . It follows from Remark 5.7 that has compact support in . Since has compact support in , we conclude that is a superform on which defines on .
5.14
Let for an open subset of , where . Obviously, we may view as a -form on with compact support. We call a very affine open subset as in Proposition 5.13 a very affine chart of integration for . Then is given by a superform . We define the integral of over by
Here, we view as a tropical cycle (see 4.6) and we integrate as in 3.4.
Lemma 5.15
For , the following properties hold:
- (a)
If is a very affine chart of integration for , then every non-empty very affine open subset of is a very affine chart of integration for .
- (b)
The definition of is independent of the choice of the very affine chart of integration for .
Proof: By Corollary 5.12, and (a) follows. To prove (b), it is enough to show
| (3) |
for a non-empty very affine open subset of by using (a). The differential form is given on (resp. ) by (resp. ). By 4.12, there is an affine homomorphism of the underlying canonical tori such that . It follows that is given on also by . By Proposition 5.6, we have . The Sturmfels–Tevelev multiplicity formula shows that (see Proposition 4.11). Then Proposition 3.10 shows that (3) holds.
Proposition 5.16
Let and let . Then we have
Proof: By Lemma 5.15, we may choose a simultaneous very affine chart of integration for both and . Then the claim follows by the corresponding property of the integration of superforms.
We have also Stokes’ theorem for differential forms on the open subset of :
Theorem 5.17
For and , we have and hence .
Proof: By Proposition 5.13, there is a very affine open subset of such that and such that is given on by a superform . Then is a very affine chart of integration for and and the claim follows from Proposition 3.5.
Remark 5.18
Integration of differential forms on complex manifolds is defined by using a partition of unity with compact supports subordinated to a covering by holomorphic charts. Surprisingly, this was not necessary in our non-archimedean algebraic setting as we have defined integration by using a single suitable tropical chart. In fact, the use of a smooth partition of unity with compact supports subordinate to an open covering of by tropical charts would not work here directly. To illustrate this, suppose that is given on by . If the functions are of the form for some and , then we could set . However, the functions could not be expected to have this form and so this approach fails.
Chambert-Loir and Ducros define integration more generally for differential forms on paracompact good analytic spaces (see [CD12], §3.8). The idea is to use a covering by the interiors of affinoid subdomains. Then there is a smooth partition of unity with supports subordinated to this covering which reduces the problem to defining integration over an affinoid subdomain. But in the affinoid case, one can find a single tropical chart of integration similarly as in Proposition 5.13. It follows from Remark 7.6 and Proposition 7.11 that both definitions give the same integral on the analytification of an algebraic variety.