Hans Georg R uck (with Oliver Bangert)

On uniformizable 𝑇 -modules,
an example
Hans-Georg Rück (with Oliver Bangert)
Institut für Mathematik
Universität Kassel
Drinfeld-modules:
The ring π”½π‘ž [πœƒ], the field 𝐾 = π”½π‘ž (πœƒ),
Λ† .
the valuation π‘£βˆž (πœƒ) = βˆ’1, 𝐾∞ = π”½π‘ž (( 1πœƒ )) and 𝐢∞ = 𝐾
∞
In 𝐢∞ consider π”½π‘ž [πœƒ]-lattice Ξ› of rank π‘Ÿ and corresponding torus 𝐢∞ /Ξ›,
parameterize it by
πœ“ : 𝐢∞ /Ξ› β†’ 𝐢∞
∞
𝑧
∏
7β†’ 𝑒(𝑧) = 𝑧
βˆ‘
𝑧
𝑖
𝑒𝑖 𝑧 π‘ž .
(1 βˆ’ ) =
πœ†
𝑖=0
0βˆ•=πœ†βˆˆΞ›
𝑒(𝑧) converges for each 𝑧, 𝑒(𝑧) is surjective,
ker(𝑒(𝑧)) = Ξ›, hence πœ“ is an isomorphism.
2
The π”½π‘ž [𝑇 ]-module structure of 𝐢∞ :
there is an additive polynomial
𝑃 (𝑋) = πœƒ 𝑋 + 𝑔1 𝑋 π‘ž + . . . + π‘”π‘Ÿ 𝑋 π‘ž
π‘Ÿ
(π‘”π‘Ÿ βˆ•= 0) with 𝑒(πœƒ 𝑧) = 𝑃 (𝑒(𝑧)).
Consider
𝑇 = πœƒ + 𝑔1 𝜏 + . . . + π‘”π‘Ÿ 𝜏 π‘Ÿ ,
where 𝜏 : 𝑋 7β†’ 𝑋 π‘ž , then πœ“ induces π”½π‘ž [𝑇 ]-module structure on 𝐢∞ .
Take 𝑀 = 𝐾 ∞ {𝜏 } as an 𝐾 ∞ [𝑇 ]-module, it has rank π‘Ÿ with basis (1, 𝜏, . . . , 𝜏 π‘Ÿβˆ’1 )
(π‘”π‘Ÿ 𝜏 π‘Ÿ = (𝑇 βˆ’ πœƒ) β‹… 1 βˆ’ 𝑔1 𝜏 βˆ’ . . . βˆ’ π‘”π‘Ÿβˆ’1 𝜏 π‘Ÿβˆ’1 )
3
Generalize this (Anderson):
𝑑 and a polynomial
Consider 𝐢∞
𝑇 = (πœƒπΌπ‘‘ + 𝑁 ) + 𝐺1 𝜏 + . . . + 𝐺𝑠 𝜏 𝑠
with 𝑁, 𝐺1 , . . . , 𝐺𝑠 ∈ M𝑑×𝑑 (𝐾 ∞ ), 𝑁 is nilpotent,
𝑑 is an 𝔽 [𝑇 ]-module, called 𝑇 -module,
then 𝐢∞
π‘ž
and 𝑀 = 𝐾 ∞ {𝜏 }𝑑 is an 𝐾 ∞ [𝑇 ]-module, called corresponding 𝑇 -motive.
𝑀 should be finitely generated, free of rank π‘Ÿ.
There is an exponential
βŽ›
⎞
𝑧1
𝐸(𝑧) = 𝐸(⎝ ... ⎠) =
𝑧𝑑
∞
βˆ‘
βŽ›
π‘žπ‘–
𝑧1
⎞
𝐸𝑖 ⎝ ... ⎠
𝑖
𝑖=0
π‘§π‘‘π‘ž
converging everywhere with 𝐸((πœƒπΌπ‘‘ + 𝑁 ) 𝑧) = 𝑇 𝐸(𝑧).
4
Properties of 𝐸: Is 𝐸(𝑧) surjective? What is ker(𝐸(𝑧))?
The following statements are equivalent:
βˆ™ kernel(𝐸(𝑧)) is an π”½π‘ž [πœƒ]-module of rank π‘Ÿ,
βˆ™ 𝐸(𝑧) is surjective.
Then the 𝑇 -module is called uniformizable.
5
How can we handle this? There is a third condition (due to Anderson):
βˆ‘
𝑛
𝐾 ∞ < 𝑇 >= { ∞
𝑛=0 π‘Žπ‘› 𝑇 ∈ 𝐾 ∞ ((𝑇 )) ∣ limπ‘›β†’βˆž π‘£βˆž (π‘Žπ‘› ) = ∞, π‘Žπ‘› ∈ finite ext. of 𝐾∞ },
let again 𝑀 = 𝐾 ∞ {𝜏 }𝑑 ,
consider
𝑀 < 𝑇 >:= 𝑀 βŠ—πΎ ∞ [𝑇 ] 𝐾 ∞ < 𝑇 > .
𝜏 operates on 𝑀 < 𝑇 > by
𝜏 (π‘š βŠ—
βˆ‘
𝑛
π‘Žπ‘› 𝑇 ) = 𝜏 (π‘š) βŠ—
βˆ‘
π‘Žπ‘žπ‘› 𝑇 𝑛 .
Consider
πœ‘ : 𝑀 < 𝑇 >𝜏 βŠ—π”½π‘ž [𝑇 ] 𝐾 ∞ < 𝑇 >β†’ 𝑀 < 𝑇 > ,
then uniformizable is equivalent to β€πœ‘ is an isomorphism” (rigid analytically trivial).
6
Our Examples (motivated by an example by Coleman-Anderson):
(
𝑑 = 2, 𝑇 =
πœƒ 0
0 πœƒ
)
(
+
π‘Ž 𝑏
𝑐 𝑑
)
(
𝜏+
1 0
0 1
)
𝜏 2 with 𝑐 βˆ•= 0.
We study it in the sense of Böckle-Hartl.
A basis of 𝑀 = 𝐾 ∞ {𝜏 }2 as 𝐾 ∞ [𝑇 ]-module is (1, 0), (0, 1), (𝜏, 0), (0, 𝜏 ),
an element in 𝑀 < 𝑇 > is given as
βˆ‘
βˆ‘
βˆ‘
βˆ‘
( 𝑛 𝑀𝑛 𝑇 𝑛 )(1, 0) + ( 𝑛 π‘₯𝑛 𝑇 𝑛 )(0, 1) + ( 𝑛 𝑦𝑛 𝑇 𝑛 )(𝜏, 0) + ( 𝑛 𝑧𝑛 𝑇 𝑛 )(0, 𝜏 )
7
β„•0
Evaluating invariants, some calculations, yields sequences (𝑦𝑛 )π‘›βˆˆβ„•0 ∈ 𝐾 ∞
given by recurrence relations
4
3
2
𝑃 (𝑦𝑛 , π‘¦π‘›βˆ’1 , π‘¦π‘›βˆ’2 ) = 𝐴4 π‘¦π‘›π‘ž + 𝐴3 π‘¦π‘›π‘ž + 𝐴2 π‘¦π‘›π‘ž + 𝐴1 π‘¦π‘›π‘ž + 𝐴0 𝑦𝑛 + 𝐡(π‘¦π‘›βˆ’1 , π‘¦π‘›βˆ’2 ) = 0,
where
𝐴0
𝐴1
𝐴2
𝐴3
𝐴4
𝐡(π‘¦π‘›βˆ’1 , π‘¦π‘›βˆ’2 )
=
=
=
=
=
=
2
πœƒπ‘ž+1 π‘π‘ž βˆ’1
2
2
π‘‘π‘ž πœƒπ‘ž π‘π‘ž βˆ’π‘ž + πœƒπ‘ž π‘Žπ‘π‘ž βˆ’1
2
2
2
2
πœƒπ‘ž + π‘‘π‘ž π‘Žπ‘ž π‘π‘ž βˆ’π‘ž βˆ’ π‘π‘ž π‘π‘ž + πœƒπ‘ž π‘π‘ž βˆ’1
2
2
π‘Žπ‘ž + π‘‘π‘ž π‘π‘ž βˆ’π‘ž
1
π‘ž2
π‘ž4
π‘ž3
βˆ’π‘ž 3
βˆ’π‘ž 2
βˆ’π‘ž 2
π‘ž 2 βˆ’1
(βˆ’πœƒ
βˆ’ πœƒ )π‘¦π‘›βˆ’1 + (βˆ’πœƒ 𝐴3 )π‘¦π‘›βˆ’1 + (βˆ’π‘
βˆ’ 1)π‘¦π‘›βˆ’1
βˆ’π‘ž 2 βˆ’π‘ž 3 π‘ž 4
+πœƒ
π‘¦π‘›βˆ’2
8
Result: The 𝑇 -module is uniformizable if and only if
{(𝑦𝑛 )π‘›βˆˆβ„•0 ∣ 𝑃 (𝑦𝑛 , π‘¦π‘›βˆ’1 , π‘¦π‘›βˆ’2 ) = 0 and limπ‘›β†’βˆž π‘£βˆž (𝑦𝑛 ) = ∞}
is an π”½π‘ž [𝑇 ]-module of rank 4.
In other words:
it is uniformizable if and only if we find for each solution of 𝑃 (𝑦0 , 0, 0) = 0
a sequence (𝑦𝑛 )π‘›βˆˆβ„•0 with and π‘£βˆž (𝑦𝑛 ) = ∞.
9
We have the initial equation:
4
3
2
𝑃 (𝑦0 , 0, 0) = 𝐴4 𝑦0π‘ž + 𝐴3 𝑦0π‘ž + 𝐴2 𝑦0π‘ž + 𝐴1 𝑦0π‘ž + 𝐴0 𝑦0 = 0 ,
and the recurrence equation:
4
3
2
𝑃 (𝑦𝑛 , π‘¦π‘›βˆ’1 , π‘¦π‘›βˆ’2 ) = 𝐴4 π‘¦π‘›π‘ž + 𝐴3 π‘¦π‘›π‘ž + 𝐴2 π‘¦π‘›π‘ž + 𝐴1 π‘¦π‘›π‘ž + 𝐴0 𝑦𝑛 + 𝐡(π‘¦π‘›βˆ’1 , π‘¦π‘›βˆ’2 ) = 0 .
We consider their Newton polygon:
let
π‘š=
min (
𝑖=1,2,3,4
βˆ’π‘£βˆž (𝐴𝑖 )
π‘£βˆž (𝐴0 ) βˆ’ π‘£βˆž (𝐴𝑖 )
)
and
𝑀
=
max
(
).
𝑖=0,1,2,3
1 βˆ’ π‘žπ‘–
π‘ž4 βˆ’ π‘žπ‘–
10
11
Conditions that large values π‘£βˆž (π‘¦π‘›βˆ’2 ) and π‘£βˆž (π‘¦π‘›βˆ’1 ) imply also large π‘£βˆž (𝑦𝑛 ):
Lemma: Suppose
π‘£βˆž (π‘π‘ž
2
βˆ’1
+ 1) βˆ’ π‘ž 2 π‘š > π‘£βˆž (𝐴0 ) βˆ’ π‘š .
If π‘£βˆž (π‘£π‘›βˆ’2 ) β‰₯ βˆ’π‘š + πœ– and π‘£βˆž (π‘£π‘›βˆ’1 ) β‰₯ βˆ’π‘š + π‘ž 2 πœ–,
then there is 𝑦𝑛 with 𝑃 (𝑦𝑛 , π‘¦π‘›βˆ’1 , π‘¦π‘›βˆ’2 ) = 0 and π‘£βˆž (𝑣𝑛 ) β‰₯ βˆ’π‘š + π‘ž 4 πœ–.
12
Now try to find for each 𝑦0 values 𝑦1 , 𝑦2 with π‘£βˆž (𝑦1 ), π‘£βˆž (𝑦2 ) > βˆ’π‘š:
Lemma: Suppose
(π‘£βˆž (𝐴0 ) βˆ’ π‘š) βˆ’ (βˆ’π‘ž 4 𝑀 ) < π‘ž 2 and π‘£βˆž (π‘π‘ž
2
βˆ’1
+ 1) βˆ’ π‘ž 2 𝑀 > π‘£βˆž (𝐴0 ) βˆ’ π‘š ,
then for each 𝑦0 with 𝑃 (𝑦0 , 0, 0) = 0
we find 𝑦1 with 𝑃 (𝑦1 .𝑦0 , 0) = 0 and π‘£βˆž (𝑦1 ) > βˆ’π‘š
and 𝑦2 with 𝑃 (𝑦2 .𝑦1 , 𝑦0 ) = 0 and π‘£βˆž (𝑦2 ) > βˆ’π‘š.
13
(π‘£βˆž (𝐴0 ) βˆ’ π‘š) βˆ’ (βˆ’π‘ž 4 𝑀 ) < π‘ž 2
14
Result: The 𝑇 -module with
(
) (
)
(
)
πœƒ 0
π‘Ž 𝑏
1 0
𝑇 =
+
𝜏+
𝜏2
0 πœƒ
𝑐 𝑑
0 1
(𝑐 βˆ•= 0)
is uniformizable if
(π‘£βˆž (𝐴0 ) βˆ’ π‘š) βˆ’ (βˆ’π‘ž 4 𝑀 ) < π‘ž 2 and π‘£βˆž (π‘π‘ž
2
βˆ’1
+ 1) βˆ’ π‘ž 2 𝑀 > π‘£βˆž (𝐴0 ) βˆ’ π‘š .
15
Can these conditions be fulfilled?
Example:
If
π‘£βˆž (π‘Ž), π‘£βˆž (𝑑) β‰₯ βˆ’
π‘ž
,
π‘ž+1
π‘£βˆž (𝑏) > βˆ’π‘ž,
0 ≀ π‘£βˆž (𝑐) <
1
,
π‘ž+1
then it is uniformizable.
16
Special case:
If the Newton polygon of 𝑃 (𝑦0 , 0, 0) is a line, i.e. π‘š = 𝑀 , then the 𝑇 -module is
uniformizable if and only if the least slope of 𝑃 (𝑦1 , 𝑦0 , 0) is smaller.
17
When is it not uniformizable? – if we find 𝑦0 with 𝑃 (𝑦0 , 0, 0) = 0 which cannot be
extended to a converging sequence.
Result: If
π‘£βˆž (π‘π‘ž
2
βˆ’1
+ 1) ≀ βˆ’π‘ž 2 (π‘ž 2 βˆ’ 1)𝑀 ,
then the 𝑇 -module is not uniformizable.
18
Example:
If
π‘ž2 + π‘ž
π‘£βˆž (π‘Ž), π‘£βˆž (𝑑) β‰₯ βˆ’ 2
,
π‘ž +1
π‘ž
βˆ’π‘ž 2 βˆ’ π‘£βˆž (π‘Ž) βˆ’ π‘£βˆž (𝑑)
βˆ’
β‰₯ π‘£βˆž (𝑐) >
,
π‘žβˆ’1
π‘žβˆ’1
βˆ’1
𝑏 = πœƒπ‘βˆ’π‘ž ,
then it is not uniformizable.
19