Invariants for 𝔾 ( r ) \mathbb{G}_{(r)} -modules Thanks: ∗ partially supported by the Simons Foundation
Eric M. Friedlander ∗
Abstract
. We revisit the constructions given by J. Pevtsova and the author of refined invariants for finite dimensional representations of infinitesimal group schemes 𝔾 ( r ) \mathbb{G}_{(r)} over a field k k of characteristic p > 0 p>0 . Our focus is on the universal p p -nilpotent operator seen as an element in the group algebra of the group scheme 𝔾 ( r ) , X \mathbb{G}_{(r),X} over X X , where X X is either the moduli space V r ( 𝔾 ) V_{r}(\mathbb{G}) of height r r 1 1 -parameter subgroups of 𝔾 \mathbb{G} or the moduli space 𝒞 r ( 𝒩 p ( 𝔤 ) ) \mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g})) of r r -tuples of p p -nilpotent, pair-wise commuting elements of the Lie algebra of 𝔾 \mathbb{G} . We formalize Jordan type function using several variants of the continuous function J T 𝔾 , r , M ( − ) : ℙ V r ( 𝔾 ) → 𝒴 JT_{\mathbb{G},r,M}(-):\mathbb{P}V_{r}(\mathbb{G})\to\mathcal{Y} where 𝒴 \mathcal{Y} is the poset of Young diagrams with p p -columns. One of these variants is designed to be more conducive to computation. The vector bundle construction given by J. Pevtsova and the author is extended to all finite dimensional 𝔾 ( r ) \mathbb{G}_{(r)} -modules, producing coherent sheaves on X X which are locally free on the strata of X X associated to J T 𝔾 , r , M ( − ) JT_{\mathbb{G},r,M}(-) .
2020 Mathematics Subject Classification
20G05, 14L17
0.
Introduction
In the present work, we utilize methods and results of support theory to investigate explicit
invariants for specific finite dimensional modules for certain classes of infinitesimal group schemes
over an arbitrary base field
k
of characteristic
p
.
This is in contrast to the role of support theory in establishing properties for various categories
of representations of finite group schemes as in
[
3
]
. Our techniques do not apply to finite groups,
for we utilize the support theory developed in collaboration with A. Suslin and C. Bendel
in
[
26
]
,
[
27
]
which replaces cohomological varieties by varieties of 1-parameter subgroups.
We further investigate the “Jordan type function” introduced by J. Pevtsova
and the author in
[
15
]
and
[
16
]
.
For a given affine group scheme
\mathbb{G}
of finite type over
k
and a finite dimensional
\mathbb{G}
-module
M
for the
r
-th Frobenius kernel
\mathbb{G}_{(r)}
of
\mathbb{G}
,
we formulate the continuous function
JT_{\mathbb{G},r,M}(-):\mathbb{P}V_{r}(\mathbb{G})\to\mathcal{Y}
from the
scheme-theoretic points of
\mathbb{P}V_{r}(\mathbb{G})
representing height
r
1-parameter subgroups
to the partially ordered set
\mathcal{Y}
of Young diagrams with
p
columns. By definition,
JT_{\mathbb{G},r,M}(-)=JT_{\mathbb{G}_{(r)},r,M}(-)
. The continuity of
JT_{\mathbb{G},r,M}(-)
implies that if a point
x\in\mathbb{P}V_{r}(\mathbb{G})
specializes
to
y\in\mathbb{P}V_{r}(\mathbb{G})
then the Young diagram
JT_{\mathbb{G},r,M}(x)
is greater or equal to
the Young diagram
JT_{\mathbb{G},r,M}(y)
.
This “theory”
M\mapsto JT_{\mathbb{G},r,M}(-)
constitutes a functor
JT_{\mathbb{G},r}:mod(\mathbb{G}_{(r)})\to Hom_{cont}(\mathbb{P}V_{r}(\mathbb{G}),\mathcal{Y})
from finite dimensional
\mathbb{G}_{(r)}
-modules to continuous functions
\mathbb{P}V_{r}(\mathbb{G})\to\mathcal{Y}
which captures much more information about a
\mathbb{G}_{(r)}
-module
M
than does the cohomological support theory
M\mapsto\mathbb{P}H^{*}(\mathbb{G}_{(r)},k)_{M}
.
For example, if
p
does not divide the dimension of
M
, then
\mathbb{P}H^{*}(\mathbb{G}_{(r)},k)_{M}
equals
\mathbb{P}H^{*}(\mathbb{G}_{(r)},k)
and therefore is independent of
M
,
whereas
M\ \mapsto\ JT_{\mathbb{G},r,M}(-)
distinguishes many such
\mathbb{G}_{(r)}
-modules
M
of a given
dimension. As might be expected, determination of
JT_{\mathbb{G},r,M}(-)
is a challenge. With this
in mind, we introduce a simplified Jordan type function
M\ \mapsto\ JT^{exp}_{\mathfrak{g},r,M}(-)
for affine group schemes
\mathbb{G}
which are of
exponential type of height
r
.
After an elementary discussion of Jordan types, we provide a detailed formalism for the
construction of the “universal
p
-nilpotent operator”
\Theta_{\mathbb{G},r}\in k[V_{r}(\mathbb{G})]\otimes k\mathbb{G}_{(r)}\equiv k\mathbb{G}_{(r),V_{r}(\mathbb{G})}
of height
r
constructed by J. Pevtsova and the author in
[
16
]
. In doing so, we expand upon the
connection between 1-parameter subgroups for
\mathbb{G}
and operators on
\mathbb{G}_{(r)}
-modules.
The operator
\Theta_{\mathbb{G},r}
acting on
k[V_{r}(\mathbb{G})]\otimes M
determines the Jordan
type function
JT^{exp}_{\mathfrak{g},r,M}(-)
.
The categorically inclined reader might want to categorify the functor
JT_{\mathbb{G},r}
by replacing
Hom_{cont}(\mathbb{P}V_{r}(\mathbb{G}),\mathcal{Y})
by the category of graded
k[V_{r}(\mathbb{G})]
-modules
equipped with a
p
-nilpotent homogeneous operator.
In order to make the Jordan type function more explicit and computationally accessible,
we restrict our attention to affine group schemes
\mathbb{G}
which are of
exponential type of height
r
. By definition, such groups
\mathbb{G}
are equipped with an
exponential map with the property that height
r
1-parameter subgroups of
\mathbb{G}
are given by applying the exponential map to
r
-tuples of
p
-nilpotent, pair-wise
commuting elements of the Lie algebra
\mathfrak{g}
of
\mathbb{G}
. Thus, the “moduli space”
for height
r
1-parameter subgroups of
\mathbb{G}
(of exponential type
r
) is the variety
\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))
of such
r
-tuples of elements in
\mathfrak{g}
. For
\mathbb{G}=GL_{N}
, the associated exponential
map is the usual (truncated) exponential sending a
p
-nilpotent element of
A\otimes\mathfrak{g}l_{N}
to
its exponential in
GL_{N}(A)
for any commutative
k
-algebra
A
.
For a given finite dimensional
\mathbb{G}_{(r)}
-module
M
with
\mathbb{G}
of exponential type
of height
r
, the Jordan type function becomes a continuous
function
JT_{\mathfrak{g},r,M}(-):\mathbb{P}\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))\to\mathcal{Y}
associated to the “projectivization”
of the universal
p
-nilpotent operator
\Theta_{\mathfrak{g},r}\in k[\mathcal{C}_{r}(\mathcal{N}_{p}\mathfrak{g}))]\otimes k\mathbb{G}_{(r)}
corresponding to
\Theta_{\mathbb{G},r}
. Our exponential Jordan type function
JT_{\mathfrak{g},r,M}^{exp}(-):\mathbb{P}\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))\to\mathcal{Y}
is defined by replacing
\Theta_{\mathfrak{g},r}
by its linear part (denoted
\Theta_{\mathfrak{g},r}^{exp}
) with respect to a natural grading on
the group algebra
k\mathbb{G}_{a(r)}
of
\mathbb{G}_{a(r)}
. In the very special case that
\mathbb{G}
equals
\mathbb{G}_{a(r)}
, this
simplification is related to J. Carlson’s use of cyclic shifted subgroups to identify support
varieties for
(\mathbb{Z}/p)^{\times r}
-modules (see
[
5
]
).
The suitability of
\Theta_{\mathfrak{g},r}^{exp}
follows from work of Pevtsova, Suslin,
and the author
[
17
]
and subsequent work of P. Sobaje
[
24
]
.
Namely, for a
\mathbb{G}_{(r)}
-module
M
of dimension
m
and a point of
x\in\mathbb{P}\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g})
),
JT_{\mathfrak{g},r,M}(x)\ =\ m/p\cdot[p]
if and only if
JT^{exp}_{\mathfrak{g},r,M}(x)\ =\ m/p\cdot[p]
.
Thus, the replacement of
JT_{\mathbb{G},r,M}(-)
by
JT_{\mathfrak{g},r,M}^{exp}(-):\mathbb{P}\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))\to\mathcal{Y}
does not affect (refined) support varieties considered in previous work of Pevtsova
and the author in
[
15
]
and
[
16
]
.
We provide sample computations which demonstrate
the accessibility of
JT_{\mathfrak{g},r,M}^{exp}(-)
for certain
\mathbb{G}_{(r)}
-modules
M
.
For
\mathbb{G}
of exponential type of height
r
, we verify that vector bundles on
\mathbb{P}\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))
associated to
\mathbb{G}_{(r)}
-modules
M
of constant Jordan type using
\Theta_{\mathfrak{g},r}
(naturally corresponding to vector bundles previously constructed in
[
16
]
)
are
\mathbb{A}^{1}
-homotopy equivalent to vector bundles using
\Theta_{\mathfrak{g},r}^{exp}
.
We extend these constructions of vector bundles to apply to arbitrary finite dimensional
\mathbb{G}_{(r)}
-modules
M
, yielding coherent sheaves on
\mathbb{P}\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))
whose restrictions on strata associated to the Jordan type function for
M
are locally free.
Unlike the situation for modules of
constant Jordan type, these stratified coherent sheaves associated to a given finite
dimensional
\mathbb{G}_{(r)}
-module will typically depend upon whether we use the Jordan
type function associated to
\Theta_{\mathbb{G},r}
or associated to
\Theta_{\mathfrak{g},r}^{exp}
.
As we observe in the last section, the issue of “stabilization with respect to
r
” of Jordan type
functions of height
\leq r
is closely related to the Jordan type function for finite
dimensional rational
\mathbb{G}
-modules introduced in
[
12
]
if
\mathbb{G}
is of exponential type of infinite height. This corrects a minor error of
[
12
, Prop 4.8]
which incorrectly formulated such a stabilization result.
We thank Julia Pevtsova and Paul Sobaje for their insights into the mathematics we present.
1.
Jordan types and specialization
We begin by recalling some basic properties of Jordan forms of
p
-nilpotent endomorphisms.
Definition 1.1
.
Let
M
be a
k[t]/t^{p}
-module of dimension
m
given by the map
\rho_{M}:k[t]/t^{p}\to M_{m,m}(k)
of
k
-algebras. Let
\underline{a}=(a_{1},\ldots,a_{p})\in\mathbb{N}^{\times p}
(i.e., a
p
-tuple of non-negative
integers) be defined by setting
a_{i}
to be the
number of blocks of size
i
of the Jordan canonical form of the matrix
\rho_{M}(t)
. We say that
the Jordan type of
M
is
\underline{a}
, expressing this as
JT(M)=\underline{a}
or as
JT(M)=a_{p}[p]+a_{p-1}[p-1]+\cdots+a_{1}[1]
.
The following proposition recalls familiar interpretations of Jordan types.
Proposition 1.2
.
Let
M
be a
k[t]/t^{p}
-module of dimension
m
such that
JT(M)=\underline{a}
.
(1)
As a
k[t]/t^{p}
-module,
M
is isomorphic to
\bigoplus_{i=1}^{p}(k[t]/t^{i})^{\oplus a_{i}}
.
(2)
The partition of
m=\sum_{i=1}^{p}a_{i}\cdot i
with
a_{i}
subsets of size
i
is naturally associated to
\underline{a}=(a_{1},\ldots,a_{p})
.
(3)
The usual partial ordering of partitions determines the
partial ordering on
\mathbb{N}^{\times p}
given by
\underline{a}\ \leq\ \underline{b}\quad\iff\quad\forall s,1\leq s\leq p,\quad(\sum_{i=s}^{p}a_{i}\cdot i)\leq(\sum_{i=s}^{p}b_{i}\cdot i).
(4)
Uniquely associated to
\underline{a}
is a Young diagram with
p
-columns such that
the diagram has
a_{i}
rows with
i
boxes and the rows of the diagram are
left-adjusted with descending length. Thus, the partial ordering
\underline{a}\ \leq\ \underline{b}
requires for each
\ell
that the number of boxes in the top
\ell
rows of
\underline{a}
is less or equal to the number of boxes in the top
\ell
rows of
\underline{b}
.
(5)
The Jordan type of
t^{j}
acting on
M\simeq\oplus_{i=1}^{p}(k[t]/t^{i})^{\oplus a_{i}}
(denoted by
\underline{a}^{j}\in\mathbb{N}^{\times p}
)
has associated Young diagram obtained from the Young diagram for
\underline{a}
by removing the first (i.e., left-most)
j-1
columns of
\underline{a}
and adding rows with
each having only 1 box so that the number of boxes of
\underline{a}^{j}
equals
m
.
The lemma below recalls some properties of the partial ordering on Jordan types.
We denote by
rk(t^{j}:M\to M)
the dimension of the subspace
t^{j}(M)\subset M
, the dimension of the image of multiplication by
t^{j}
on the vector space
M
.
Proposition 1.3
.
Consider
m
-dimensional
k[t]/t^{p}
-modules
M,\ N
isomorphic to
\bigoplus_{i=1}^{p}(k[t]/t^{i})^{\oplus a_{i}},\ \bigoplus_{i=1}^{p}(k[t]/t^{i})^{\oplus b_{i}}
.
As above, let
\underline{a}^{s}
(respectively,
\underline{b}^{s}
) denotes the Jordan type of
t^{s}
acting on
M
(resp,
N
) for any
s,\ 1\leq s
0
,
(1.10.1)
C_{d}\ \equiv\ \{\mathfrak{p}\in\operatorname{Spec}\nolimits A:\ rk(\theta_{\mathfrak{p}}:k(\mathfrak{p})\otimes_{A}M_{A}\to k(\mathfrak{p})\otimes_{A}M_{A})\leq d\}
is a closed subset of
\operatorname{Spec}\nolimits R
.
Proof.
By considering the intersection of
C_{d}
with each member of an affine open covering of
\operatorname{Spec}\nolimits A
restricted to which
M_{A}
is free, it suffices to prove the lemma assuming
M_{A}
is free. Moreover,
if
\mathfrak{q}_{1},\ldots,\mathfrak{q}_{r}
are the minimal primes of
\operatorname{Spec}\nolimits A
, then the closures of
\{\mathfrak{q}_{i}\}
are the
(closed) irreducible components of
\operatorname{Spec}\nolimits A
. Thus, by considering the intersection of
C_{d}
with
each of these closures, it suffices to assume that
A
is a domain.
We view
\theta_{A}
as an element of
M_{m,m}(A)
. Then
C_{d}
is the closed subset defined
as the zero locus of the determinants of all
(d+1\times d+1)
-submatrices of
\theta_{A}
.
∎
We remind the reader that an Alexandrov discrete topology on a
T_{0}
space
is a topology in which the arbitrary union of closed subsets is closed.
(See
[
1
]
and
[
21
]
.) The topology on
\mathcal{Y}_{\leq}
which we utilize is such a topology.
Definition 1.11
.
We give
\mathcal{Y}_{\leq}
the topology whose closed subsets are those
subsets
S\subset\mathcal{Y}_{\leq}
satisfying the condition that if
s\in S
and
s^{\prime}\leq s
then
s^{\prime}\in S
.
In particular,
S\subset\mathcal{Y}_{\leq}
is closed if and only if
S\cap\mathcal{Y}_{\leq,m}
is closed in
\mathcal{Y}_{\leq,m}
for all
m>0
.
Theorem 1.12
.
Let
A
be a Noetherian commutative
k
-algebra
and consider a projective
A
-module
M_{A}
of rank
m
equipped with a
p
-nilpotent endomorphism
\theta_{A}
. As usual, we give
\operatorname{Spec}\nolimits A
the Zariski topology. Then
JT_{A,M_{A}}(-):\operatorname{Spec}\nolimits A\quad\to\quad\mathcal{Y}_{\leq,m}\quad\hookrightarrow\quad\mathcal{Y}_{\leq}
is continuous.
Proof.
A subset
C\subset\mathcal{Y}_{\leq,m}
is closed if and only
C\quad=\quad\bigcup_{\underline{a}\in C}\overline{\{\underline{a}\}},
where
\overline{\{\underline{a}\}}
is the closure of the singleton subset
\{\underline{a}\}
.
Thus, it suffices to prove that
JT_{A,M_{A}}^{-1}(\overline{\{\underline{a}\}})
is closed
for all
\underline{a}\in\mathcal{Y}_{\leq,m}
(since
\mathcal{Y}_{\leq,m}
is finite). By definition,
JT_{A,M_{A}}^{-1}(\overline{\{\underline{a}\}})
consists of those
x\in\operatorname{Spec}\nolimits A
such that
JT_{A,M_{A}}(x)\ \leq\ \underline{a}
in
\mathcal{Y}_{\leq,m}.
By Proposition
1.3
(4),
JT_{A,M_{A}}(x)\ \leq\ \underline{a}
if and only if the
rank of
(\theta_{x})^{s}:k(x)\otimes_{A}M_{A}\to k((x)\otimes_{A}M_{A}
is less
than or equal to the rank of
t^{s}
on
\oplus_{i=1}^{p}(k[t]/t^{i})^{\oplus a_{i}}
(which equals
\sum_{i=s+1}^{p}a_{i}\cdot(i-s)
) for all
s
,
1\leq s
1
) which sends
u
to the element
u_{r-1}\in k\mathbb{G}_{a(r)}
.
A scheme-theoretic point
x\in V_{r}(\mathbb{G})
corresponding to the 1-parameter subgroup
\mu_{x}:\mathbb{G}_{a(r),k(x)}\to\mathbb{G}_{(r),k(x)}
determines the map of
k(x)
-algebras
(2.2.1)
(\mu_{x})_{*}\circ\epsilon_{r}:k(x)\otimes k[u]/u^{p}\to k(x)\otimes k\mathbb{G}_{a(r)}\to k(x)\otimes k\mathbb{G}_{(r)},\ u\mapsto(\mu_{x})_{*}(1\otimes u_{r-1}).
Thus, if
M
is a
\mathbb{G}
-module, then
(\mu_{x})_{*}\circ\epsilon_{r})^{*}(k(x)\otimes M)
is equipped with a
k(x)\otimes k[u]/u^{p}
-module structure such that
u
acts as
(\mu_{x})_{*}(1\otimes u_{r-1})
on
k(x)\otimes M
.
To unpack
(\mu_{x})_{*}(u_{r-1})\in k(x)\otimes k\mathbb{G}_{(r)}
, we observe that this is the
k(x)
-linear function sending
f\in k[\mathbb{G}]
to the coefficient of
t^{p^{r-1}}
in the expansion
of
\mu_{x}^{*}(f)\in k(x)\otimes k[\mathbb{G}_{a(r)}]=k(x)[t]/t^{p^{r}}
.
We recall the local Jordan type of a finite dimensional
\mathbb{G}_{(r)}
-module as first formulated
in
[
15
]
and
[
16
]
by Pevtsova and the author.
Definition 2.3
.
(See
[
16
, Cor 3.8]
.)
Let
\mathbb{G}
be an affine group scheme of finite type over
k
,
r\geq 1
, and
M
a finite dimensional
\mathbb{G}_{(r)}
-module. Then the local Jordan type of
M
at a scheme-theoretic point
x\in V_{r}(\mathbb{G})
,
JT_{\mathbb{G},r,M}(x)
, is defined to be the Jordan type of the
k(x)[u]/u^{p}
-module
((\mu_{x})_{*}\circ\epsilon_{r})^{*}(k(x)\otimes M)
.
In the following definition, we recall
the natural grading on
V_{r}(\mathbb{G})
as discussed in
[
26
, §1]
. This is particularly
relevant because there is a natural morphism
\operatorname{Proj}\nolimits k[V_{r}(\mathbb{G})]\to\mathbb{P}H^{*}(\mathbb{G}_{(r)},k)\simeq\Pi(\mathbb{G}_{(r)})
which is a homeomorphism of Zariski spaces
to the
\pi
-point scheme
\Pi(\mathbb{G})
(see
[
17
, Prop 3.1]
).
Definition 2.4
.
The natural right action
V_{r}(\mathbb{G}_{a})\times\mathbb{A}^{1}\to V_{r}(\mathbb{G}_{a})
sends
(\nu:\mathbb{G}_{a(r),A}\to\mathbb{G}_{a(r),A},\alpha\in A)
to
\alpha\cdot\nu:\mathbb{G}_{a(r),A}\to\mathbb{G}_{a(r),A}
defined by
(\alpha\cdot\nu)^{*}(t)=\nu^{*}(\alpha\cdot t)
.
Composing this with 1-parameter subgroups
\mu:\mathbb{G}_{a(r),A}\to\mathbb{G}_{(r),A}
defines
a right action
V_{r}(\mathbb{G})\times\mathbb{A}^{1}\to V_{r}(\mathbb{G})
. This action corresponds to a grading
on the commutative algebra
k[V_{r}(\mathbb{G})]
.
We set
\mathbb{P}V_{r}(\mathbb{G})\quad\equiv\quad\operatorname{Proj}\nolimits k[V_{r}(\mathbb{G})].
If
\mu:\mathbb{G}_{(r),R}\to\mathbb{G}_{R}
is an
R
-point of
V_{r}(\mathbb{G})
for some commutative
k
-algebra
R
and
s
an
R
-point of
A^{1}
(i.e., an element
s\in R
),
then
(s\circ\mu)_{*}(u_{j})=s^{p^{i}}\mu_{*}(u_{j})
for
u_{j}\in k\mathbb{G}_{a(r)}=k[u_{0},\cdots,u_{r-1}]/(u_{i}^{p}).
(See the proof of
[
27
, Prop 6.5]
).)
We rephrase Lemma 1.12 of
[
26
]
in order to elaborate upon this grading of
V_{r}(\mathbb{G})
.
Proposition 2.5
.
The commutative algebra
k[V_{r}(GL_{N})]
is a graded quotient of the symmetric algebra over
k
generated with generators
\{X_{i,j}^{\ell};1\leq i,j,\leq N,0\leq\ell0
.
Identify
V_{r}(\mathbb{G}_{a})
with
\operatorname{Spec}\nolimits k[T_{0},\ldots,T_{r-1}]
with
T^{i}
given grading
p^{i}
and identify the group algebra
k\mathbb{G}_{(r)}
with
k[u_{0},\ldots,u_{r-1}]/(u_{i}^{p})
.
Then
\Theta_{\mathbb{G}.r}\ =\ \sum_{i=0}^{r-1}T_{i}^{p^{r-1-i}}\otimes u_{i}\ +\ \cdots\quad\in\ k[V_{r}(\mathbb{G}_{a})]\otimes k\mathbb{G}_{a(r)},
where the non-explicit, remaining terms of
\Theta_{\mathbb{G}_{a},r}
each consist of a
polynomial homogeneous of weighted degree
p^{r-1}
in the
T_{i}
’s tensor a monomial in the
u_{i}
’s
of degree at least 2. (See
[
27
, Eqn 6.5.1]
.)
The somewhat subtle aspect of Theorem
2.15
is its last assertion of homogeneity of degree
p^{r-1}
whose proof is given in
[
16
]
(reflected in
the examples above).
Theorem 2.15
.
(
[
16
]
)
Let
\mathbb{G}
be an affine group scheme of finite type
over
k
and let
r
be a chosen positive integer.
Then
\Theta_{\mathbb{G},r}
as given in Definition
2.13
satisfies the following properties.
(1)
For any scheme-theoretic point
x\in V_{r}(\mathbb{G})
, the specialization along
\chi_{x}:\operatorname{Spec}\nolimits k(x)\to V_{r}(\mathbb{G})
of
\Theta_{\mathbb{G},r}\in k[V_{r}(\mathbb{G})]\otimes k\mathbb{G}_{(r)}
equals
(\mu_{x})_{*}(1\otimes u_{r-1})\quad\in\ k(x)\otimes k\mathbb{G}_{(r)}
,
where
\mu_{x}
is the 1-parameter subgroup of Definition
2.1
.
(2)
Let
\phi:\mathbb{H}\hookrightarrow\mathbb{G}
be a closed subgroup scheme determining
\phi^{*}:k[V_{r}(\mathbb{G})]\twoheadrightarrow k[V_{r}(\mathbb{H})]
and
\phi_{(r)*}:k\mathbb{H}_{(r)}\hookrightarrow k\mathbb{G}_{(r)}
.
Then the restriction of
\Theta_{\mathbb{G},r}
along
\phi^{*}
,
(\phi^{*}\otimes 1)(\Theta_{\mathbb{G},r})\in k[V_{r}(\mathbb{H})]\otimes k\mathbb{G}_{(r)}
, equals the image of
(1\otimes\phi_{(r)*})(\Theta_{\mathbb{H},r})
.
(3)
For any
\mathbb{G}_{(r)}
-module
M
,
\Theta_{\mathbb{G},r}
determines a
p
-nilpotent,
k[V_{r}(\mathbb{G})]
-linear operator
\Theta_{\mathbb{G},r,M}:k[V_{r}(\mathbb{G})]\otimes M\ \to\ k[V_{r}(\mathbb{G})]\otimes M
.
(4)
The base change along a scheme theoretic point
x\in V_{r}(\mathbb{G})
of
\Theta_{\mathbb{G},r,M}
equals the action of
(\mu_{x})_{*}(1\otimes u_{r-1})
on
k(x)\otimes M
.
(5)
\Theta_{\mathbb{G},r}
is of the form
\sum f_{i}\otimes\phi_{i}\in k[V_{r}(\mathbb{G})]\otimes k\mathbb{G}_{(r)}
with each
f_{i}
homogeneous of degree
p^{r-1}
in
k[V_{r}(\mathbb{G})]
and each
\phi_{i}\in k\mathbb{G}_{(r)}
an element of the augmentation ideal
I_{\mathbb{G}_{(r)}}
of
k\mathbb{G}_{(r)}
.
Proof.
Since
\mu_{x}
is the pull-back along
\chi_{x}
of
\mathcal{U}_{\mathbb{G},r}
, the following diagram commutes:
(2.15.1)
\textstyle{k[V_{r}(\mathbb{G})]\otimes k\mathbb{G}_{a(r)}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}
\scriptstyle{\chi^{*}}
\scriptstyle{(\mathcal{U}_{\mathbb{G},r})_{*}}
\textstyle{k[V_{r}(\mathbb{G})]\otimes k\mathbb{G}_{(r)}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}
\scriptstyle{\chi^{*}}
\textstyle{k(x)\otimes k\mathbb{G}_{a(r)}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}
\scriptstyle{(\mu_{x})_{*}}
\textstyle{k(x)\otimes k\mathbb{G}_{(r)}.}
Consequently, the restriction of
\Theta_{\mathbb{G},r}=(\mathcal{U}_{\mathbb{G},r})_{*}(1\otimes u_{r-1})
equals
(\mu_{x})_{*}(1\otimes u_{r-1})
.
The naturality of
\mathbb{G}\mapsto\Theta_{\mathbb{G},r}
with respect to closed
embeddings
\phi:\mathbb{H}\hookrightarrow\mathbb{G}
as in assertion (2)
follows from the commutativity of (
2.12.3
).
The action by
\Theta_{\mathbb{G},r,M}
is given by the
restriction to
\{\Theta_{\mathbb{G},r}\}\otimes(k[V_{r}(\mathbb{G})]\otimes M)
of the tensor product
pairing
(2.15.2)
(k[V_{r}(\mathbb{G})]\otimes k\mathbb{G}_{(r)})\otimes(k[V_{r}(\mathbb{G})]\otimes M)\quad\to\quad k[V_{r}(\mathbb{G})]\otimes M.
The fact that this action is
p
-nilpotent follows from the observation that the
i
-th power of
\Theta_{\mathbb{G},r}
equals
(\mathcal{U}_{\mathbb{G},r})_{*}((1\otimes u_{r-1})^{i})
.
Assertion (4) follows directly from assertion (1).
The last assertion is stated and proved in
[
16
, Prop 2.11]
. The proof proceeds by first observing
that it suffices to verify that the restriction of the
k[V_{r}(\mathbb{G})]
-linear map
\mathcal{U}_{\mathbb{G},r}^{*}:k[V_{r}(\mathbb{G})]\otimes k[\mathbb{G}_{(r)}]\to k[V_{r}(\mathbb{G})]\otimes k[\mathbb{G}_{a(r)}]
to
1\otimes k[\mathbb{G}_{(r)}]
reads off the coefficient of
t^{p^{r-1}}
in
k[V_{r}(\mathbb{G})]\otimes k\mathbb{G}_{a(r)}\simeq k[V_{r}(\mathbb{G})]\otimes k[t]/t^{p^{r}}
. The authors then
verify that the actions of
\mathbb{A}^{1}
on
\mathbb{G}_{a(r)}
and
V_{r}(\mathbb{G})
are appropriately compatible.
∎
Definition 2.16
.
Let
\mathbb{G}
be an affine group scheme of finite type over
k
and let
r
be a
positive integer. For any finite dimensional
\mathbb{G}_{(r)}
-module
M
, we define the function
(2.16.1)
JT_{\mathbb{G},r,M}(-):V_{r}(\mathbb{G})\quad\to\quad\mathcal{Y}
from the set of scheme-theoretic points of
V_{r}(\mathbb{G})
to the set of points of the partially orders
set
\mathcal{Y}
by sending
x\in V_{r}(\mathbb{G})
to
JT_{\mathbb{G},r,M}(x)\quad\equiv\quad JT((\mu_{x})_{*}\circ\epsilon_{r})^{*}(k[V_{r}(\mathbb{G})]\otimes M)),
the Jordan type of the
k(x)[u]/u^{p}
-module obtained by restricting
k[V_{r}(\mathbb{G})]\otimes M
along
(\mu_{x})_{*}\circ\epsilon_{r}:k(x)[u]/u^{p}\to k(x)\otimes k\mathbb{G}_{a(r)}\to k[V_{r}(\mathbb{G})]\otimes k\mathbb{G}_{(r)}
;
equivalently,
JT_{\mathbb{G},r,M}(x)
is the Jordan type of
(\mu_{x})_{*}(1\otimes u_{r-1})
acting on
k(x)\otimes M
.
By Theorem
2.15
(1),
JT_{\mathbb{G},r,M}(x)
equals the Jordan type of
k(x)\otimes_{k[V_{r}(\mathbb{G})]}\Theta_{\mathbb{G},r,M}\ \in\ End_{k(x)}(k(x)\otimes M)
.
Let
R
be a graded, commutative
k
-algebra; for example,
R\ =\ k[V_{r}(\mathbb{G})]
.
For any graded
R
-module
M_{*}
, we denote by
(M_{*})^{\sim}
the associated quasi-coherent
sheaf on
\operatorname{Proj}\nolimits(R)
. We shall often consider graded
R
-modules of the form
R\otimes V
where
V
is a
k
-vector space and
R\otimes V
inherits the grading of
R
.
If
M_{*}
is a graded
R
-module, we denote by
M_{*}(n)
the graded module
obtained from
M_{*}
by “shifting down by
n
” the degrees of
M_{*}
; thus,
(M_{*}(n))_{d}=M_{n+d}
. We use
(M_{*})^{\sim}(n)
to denote
(M_{*}(n))^{\sim}
.
The operator
\Theta_{\mathbb{G},r,M}
of Theorem
2.15
(3) multiplies degrees by
p^{r-1}
, so that
viewed as an operator on graded modules one should write
\Theta_{\mathbb{G},r,M}
as
\Theta_{\mathbb{G},r,M}:k[V_{r}(\mathbb{G})]\otimes M\quad\to\quad(k[V_{r}(\mathbb{G})]\otimes M)(p^{r-1}).
Because
k
has positive characteristic,
the informative survey by M. Beltrametti and L. Robbiano in
[
2
]
does not apply to our
consideration of
\mathbb{P}V_{r}(\mathbb{G})
. Because
V_{r}(\mathbb{G})
is usually not
generated by homogeneous elements of degree 1, we supplement the clear exposition of
R. Hartshorne in
[
19
, §II.5]
with
[
10
, Ex 9.5]
when considering the relationship
between graded modules over a commutative, graded
k
-algebra
R
and sheaves of
modules on
\operatorname{Proj}\nolimits(R)
.
Proposition 2.17
.
Consider a polynomial algebra
k[T_{0},\ldots,T_{N}]
with each
T_{i}
given a “weighted” degree
d_{i}
.
Assume that
I_{Y}\subset k[T_{0},\ldots,T_{N}]
is an ideal generated by polynomials
F_{j}(\underline{T})
which are (weighted) homogeneous. Then, the zero locus
Y=Z(I_{Y})\ \subset\operatorname{Proj}\nolimits k[T_{0},\ldots,T_{N}]
is a weighted projective variety associated to the
commutative, graded algebra
k[T_{0},\ldots,T_{N}]/I_{Y}
.
Let
d
be a positive integer divisible by each
d_{i}
.
The graded
k[T_{0},\ldots,T_{N}]/I_{Y}
-module
(k[T_{0},\ldots,T_{N}]/I_{Y})(d)
is a locally free, coherent rank 1 sheaf of
\mathcal{O}_{Y}
-modules on
Y
.
We denote this sheaf by
\mathcal{O}_{Y}(d)
. For each
i
such that
T_{i}
is not in the radical of
I_{Y}
, the restriction
(\mathcal{O}_{Y}(d))_{|V_{i}}
of
\mathcal{O}_{Y}(d)
to open subvariety
V_{i}\ \equiv(\operatorname{Proj}\nolimits k[T_{0},\ldots,T_{N}])-Z(T_{i}))\cap Y\ \subset\operatorname{Proj}\nolimits k[T_{0},\ldots,T_{N}]
is generated as a free, rank 1
\mathcal{O}_{V_{i}}
-module by the image of
T_{i}^{d/d_{i}}\in\mathcal{O}_{Y}(d)(Y)
in
\mathcal{O}_{Y}(d)(V_{i})
.
Proof.
It suffices to assume that
I_{Y}
is a radical ideal . Moreover, by considering each irreducible component
of
Y
, we may assume that
Y
is irreducible. Then
[
10
, Ex 9.5]
applies to verify that
the variety
Z(I_{Y})
is isomorphic to its image in
\operatorname{Proj}\nolimits S
,
where
S\ =k[T_{0},\ldots,T_{N}]_{(d)}\subset k[T_{0},\ldots,T_{N}]
is the graded subalgebra generated by
(weighted) homogeneous polynomials each of whose weights is divisible by
d
and
the image of
Z(Y)
is given by the homogeneous ideal
I_{Y}\cap S\subset S
which is generated
by elements of degree
d
. Thus, replacing
k[T_{0},\ldots,T_{N}]
by
S
with grading modified by dividing degrees by
d
, we may apply
[
19
, Prop II.5.12]
(which is stated for graded submodules of a commutative, graded algebra generated in degree 1).
∎
In particular, Proposition
2.17
applies to the graded commutative algebra
k[V_{r}(\mathbb{G})]
which is generated by elements homogeneous of degree
p^{i},1\leq i0
,
the fact that
M\ \mapsto\ JT_{\mathbb{G},r,M}(x)
commutes with tensor products is given
by
[
17
, Prop 4.7]
.
A surjective map
f:M\to N
of finite dimensional
\mathbb{G}_{(r)}
-modules induces a
surjective map
k(x)\otimes M\ \to\ k(x)\otimes N
of
k(x)[u]/t^{p}
-modules.
Consequently, assertion (5) follows from Corollary
1.4
.
Assertion (6) follows from the definition of a locally split exact sequence (see
[
6
]
)
and the fact for any field
K
and any short exact sequence
0\to N_{K}\to M_{K}\to Q_{K}\to 0
that the Jordan type of
M_{K}
equals the sum of the Jordan types of
N_{K}
and
Q_{K}
if and
only if the short exact sequence splits.
Assertion (7) is equivalent to the assertion that
JT_{\mathbb{G},r,M}(\phi_{*}(y))=JT_{\mathbb{H},r,\phi^{*}(M)}(y)
for every
y\in\mathbb{P}V_{r}(\mathbb{H})
. This
is immediate from the identification (with
x=\phi_{*}(y)
)
JT((\mu_{x})_{*}\circ\epsilon_{r})^{*}(k(x)\otimes M)\quad=\quad JT((\mu_{y})_{*}\circ\epsilon_{r})^{*}(k(y)\otimes\phi^{*}(M)).
∎
The following corollary is rephrasing of some of the properties of Theorem
2.21
.
Corollary 2.22
.
We denote by
Hom_{cont}(\mathbb{P}V_{r}(\mathbb{G}),\mathcal{Y})
the set of continuous
maps from
\mathbb{P}V_{r}(\mathbb{G})
with the Zariski topology to
\mathcal{Y}
with the Alexandrov
discrete topology. Give (the scheme-theoretic points of )
\mathbb{P}V_{r}(\mathbb{G})
the partial
ordering of specialization and give
Hom_{cont}(\mathbb{P}V_{r}(\mathbb{G}),\mathcal{Y})
the partial
ordering inherited from that of
\mathcal{Y}
.
Then sending a finite dimension
\mathbb{G}_{(r)}
-module
M
to
JT_{\mathbb{G},r,M}(-)
determines
JT_{\mathbb{G},r}:mod(\mathbb{G}_{(r)})\quad\to\quad Hom_{cont}(\mathbb{P}V_{r}(\mathbb{G}),\mathcal{Y})
commuting with sums and the partial orderings.
The following theorem associates to a finite dimensional
\mathbb{G}
-module
M
the configuration
\{(\mathbb{P}V_{r}(\mathbb{G}))_{M,\leq\underline{a}}\}
which gives considerably more
information about
M
than does the support variety
\mathbb{P}V_{r}(\mathbb{G})_{M}\subset\mathbb{P}V_{r}(\mathbb{G})
of
M
. The fact that
(V_{r}(\mathbb{G}))_{M,\leq\underline{a}}\subset V_{r}(\mathbb{G})
is closed is given with a sketch of proof in
[
15
, Prop 2.8.3]
.
Theorem 2.23
.
(see Proposition 2.8.3 of
[
15
]
)
Let
\mathbb{G}
be an affine group scheme of finite type over
k
and let
r
be a
positive integer. Consider a
\mathbb{G}_{(r)}
-module
M
of dimension
m
.
For each
\underline{a}\in\mathcal{Y}_{m}
, we define the subset
(\mathbb{P}V_{r}(\mathbb{G}))_{M,\leq\underline{a}}\ \equiv\ \{\overline{x}:JT_{\mathbb{G},r,M}(\overline{x})\leq\underline{a}\}\ \subset\ \mathbb{P}V_{r}(\mathbb{G}).
Then
\{(\mathbb{P}V_{r}(\mathbb{G}))_{M,\leq\underline{a}};\ \underline{a}\in\mathcal{Y}_{m}\}
is a
partially ordered configuration of closed subspaces of
\mathbb{P}V_{r}(\mathbb{G})
.
With this notation, the support variety
\mathbb{P}V_{r}(\mathbb{G})_{M}
of
[
27
]
equals the union of those
subspaces
(\mathbb{P}V_{r}(\mathbb{G}))_{M,\underline{a}\leq}
indexed by
\underline{a}\in\mathcal{Y}_{m}
which
satisfy the condition that
a_{i}\not=0
for some
i0
. Then
\mathbb{G}
admits an embedding
in
GL_{2}
as the
r
-th Frobenius kernel of the unipotent subgroup
U_{2}
of strictly upper
triangular matrices. As in Example
2.10
,
\mathbb{P}V_{r}(\mathbb{G})
is a weighted projective space with weights
(1,p,\cdots,p^{r-1})
associated to the graded polynomial algebra
k[T_{0},T_{1},\ldots,T_{r-1}]
where
T_{i}
has
weight
p^{i}
.
Consider a non-zero
r
-tuple
\underline{a}=(a_{0},\ldots,a_{r-1})
of elements of
k(x)
determining a
scheme-theoretic point
x\in V_{r}(X)
projecting to the scheme-theoretic point
\overline{x}\in\mathbb{P}V_{r}(\mathbb{G})
, and consider the associated 1-parameter
subgroup
\mu_{x}:\mathbb{G}_{a(r),k(x)}\to\mathbb{G}_{a(r),k(x)}.
An explicit (but quite complicated) formula for
\mu_{x*}(1\otimes u_{r-1})\in k(x)\otimes k\mathbb{G}_{a(r)}\ =\ k(x)\otimes k[u_{0},\ldots,u_{r-1}]/(u_{i}^{p})
is given in
[
27
, (6.5.12)]
.
The sum of terms linear in the
u_{i}
’s in the
expression for
\mu_{x*}(1\otimes u_{r-1})
is
\sum_{i=0}^{r-1}a_{r-1-i}^{p^{i}}u_{i}\in k(x)\otimes k\mathbb{G}
.
Replacing
\mu_{x*}(1\otimes u_{r-1})
by
\sum_{i=0}^{r-1}a_{r-1-i}^{p^{i}}u_{i}
is a special case of replacing
\Theta_{\mathbb{G},r}
by
\Theta^{exp}_{\mathfrak{g},r}
as discussed in Section
4
.
4.
Exponential Jordan type
In this section, we consider affine group schemes whose 1-parameter subgroups
are given by exponentiating
r
-tuples of
p
-nilpotent, pair-wise commuting elements
of
\mathfrak{g}\equiv Lie(\mathbb{G})
. The more explicit nature of these 1-parameter subgroups
allows us to simplify our Jordan type functions, not necessarily preserving their values
but preserving their close relationship with support varieties.
For an affine group scheme
\mathbb{G}
equipped with an exponential map
\mathcal{E}_{(r)}:\mathbb{G}_{a(r)}\times\mathcal{N}_{p}(\mathfrak{g})\to\mathbb{G}_{(r)}
of height
r
, there is a natural morphism
\Phi_{\mathfrak{g},r}:\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))\ \to\ V_{r}(\mathbb{G})
where
\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))
is the affine variety of
r
-tuples of
p
-nilpotent, pair-wise commuting
elements of
\mathfrak{g}
made explicit in Definition
4.2
.
We say that
\mathbb{G}
is of exponential type of height
r
if
\Phi_{\mathfrak{g},r}
is a homeomorphism
of scheme-theoretic points. Examples include Frobenius kernels of various classical groups,
their parabolic subgroups, and unipotent subgroups of these parabolic subgroups. (See, for
example,
[
25
]
.)
For
\mathbb{G}
of exponential type of height
r
, Theorem
4.8
introduces
\Theta_{\mathfrak{g},r}\in k[\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))]\otimes k\mathbb{G}_{(r)}
which very closely models
\Theta_{\mathbb{G},r}
.
This leads to our simpler
p
-nilpotent operator
\Theta_{\mathfrak{g},r}^{exp}\in k[\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))]\ \otimes k\mathbb{G}_{(r)}
for
\mathbb{G}
of exponential type of height
r
, a
linearization of
\Theta_{g,r}
. A fundamental computation of
[
17
, Thm 1.12]
enables us to conclude that
\Theta_{\mathfrak{g},r}
and
\Theta_{\mathfrak{g},r}^{exp}
determine the same
support varieties for finite dimensional
\mathbb{G}_{(r)}
-modules. By adopting the formalism
we have discussed for
JT_{\mathbb{G},r,M}(-)
, we establish the Jordan type functions
JT_{\mathfrak{g},r,M}(-)
and
JT_{\mathfrak{g},r,M}^{exp}(-)
which satisfy similar properties.
We begin by defining the class of affine group schemes
\mathbb{G}
for which
r
-tuples of
p
-nilpotent, pair-wise commuting elements of the Lie algebra of
\mathbb{G}
determine height
r
1-parameter subgroups of
\mathbb{G}
.
Definition 4.1
.
Let
\mathbb{G}
be an affine group scheme of finite type over
k
with Lie algebra
\mathfrak{g}
.
Fix a positive integer
r
. We say that a morphism
\mathcal{E}_{(r)}:\mathbb{G}_{a(r)}\times\mathcal{N}_{p}(\mathfrak{g})\quad\to\quad\mathbb{G}_{(r)}
equips
\mathbb{G}
with the structure of an exponentiation of height
r
if
\mathcal{E}_{(r)}
satisfies the following conditions for any commutative
k
-algebra
A
,
p
-nilpotent
elements
B,B^{\prime}\ \in A\otimes\mathfrak{g}
,
and elements
\alpha,\beta\in\mathbb{G}_{a(r)}(A)
:
(1)
The specialization of
\mathcal{E}_{(r)}
along
B:k[\mathcal{N}_{p}(\mathfrak{g})]\to A
,
\mathcal{E}_{(r),B}:\mathbb{G}_{a(r),A}\to\mathbb{G}_{(r),A}
, is a 1-parameter subgroup of
\mathbb{G}_{(r),A}
.
(2)
If
B^{\prime}
commutes with
B
, then
\mathcal{E}_{(r),B}(\alpha)\cdot\mathcal{E}_{(r),B^{\prime}}(\beta)=\mathcal{E}_{(r),B^{\prime}}(\beta)\cdot\mathcal{E}_{(r),B}(\alpha)\in\mathbb{G}_{(r)}(A).
(3)
\mathcal{E}_{(r),\alpha\cdot B}(-)\ =\ \mathcal{E}_{(r),B}(\alpha\cdot-):\mathbb{G}_{a(r),A}\to\mathbb{G}_{(r),A}
.
As exhibited in Construction
4.3
, the structure
\mathcal{E}_{(r)}
of exponentiation of height
r
for
\mathbb{G}
enables us to “exponentiate”
r
-tuples of
p
-nilpotent, pair-wise commuting elements
of
\mathfrak{g}
. We first define the variety of such
r
-tuples,
\mathcal{C}_{r}(\mathfrak{g})
.
Definition 4.2
.
Let
\mathbb{G}
be an affine group scheme of finite type over
k
with Lie algebra
\mathfrak{g}
.
Consider the functor on commutative
k
algebras sending
A
to an
r
-tuple
\underline{B}=(B_{0},\ldots,B_{r-1})
of elements in
A\otimes\mathfrak{g}
which satisfies the conditions that
B_{s}^{[p]}=0
for all
s,\ 0\leq s0
.
Lemma 4.7
.
Let
(\mathbb{G},\mathcal{E}_{(r)})
be an affine group scheme of exponential type of height
r
.
The bicontinuous morphism
(4.7.1)
\Phi_{\mathfrak{g},r}:\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))\quad\to\quad V_{r}(\mathbb{G})
is a homeomorphism of Zariski topological spaces which commutes with
\mathbb{A}^{1}
-actions.
Proof.
The assumption of exponential type of height
r
implies that
\Phi_{\mathfrak{g},r}
is a bijection on scheme-theoretic
points. Since
\Phi_{\mathfrak{g},r}
is a morphism, it is continuous. Since
\mathbb{P}\Phi_{\mathfrak{g},r}:\mathbb{P}\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))\to\mathbb{P}V_{r}(\mathbb{G})
is
proper, it is a closed map, thus induces a bijection on closed subsets, thus a homeomorphism. This
implies that the pull-back/restriction
\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))-\{0\}\to V_{r}(\mathbb{G})-\{0\}
is a homeomorphism.
Since the cone point
0
is closed in both
\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))
and
V_{r}(\mathbb{G})
, this implies
that
\Phi_{\mathfrak{g},r}
is also a homeomorphism.
∎
We adapt the construction of
\Theta_{\mathbb{G},r}
in Definition
2.13
and its use
in Theorem
2.15
by replacing
\mathcal{U}_{\mathbb{G},r}
by
\mathcal{U}_{\mathfrak{g},r}
.
Theorem 4.8
.
Let
(\mathbb{G},\mathcal{E}_{(r)})
be an affine group scheme of exponential type of height
r
.
We define the
p
-nilpotent operators
(4.8.1)
\Theta_{\mathfrak{g},r|s}\ \equiv\ (\mathcal{U}_{\mathfrak{g},r|s})_{*}(1\otimes u_{r-1}),\ \Theta_{\mathfrak{g},r}\ \equiv\ (\mathcal{U}_{\mathfrak{g},r})_{*}(1\otimes u_{r-1})\ \in\ k[\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g})))]\otimes k\mathbb{G}_{(r)},
and denote by
\Theta_{\mathfrak{g},r|s,\underline{B}},\ \Theta_{\mathfrak{g},r,\underline{B}}
their specializations at a scheme-theoretic point
\underline{B}\in\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))
.
(1)
\Theta_{\mathfrak{g},r}\ =\ (\Phi^{*}_{\mathfrak{g},r}\otimes 1)(\Theta_{\mathbb{G},r})
.
(2)
\Theta_{\mathfrak{g},r|s,\underline{B}}\ =\ (\mathcal{E}_{r|s,B_{s}})_{*}(u_{r-s-1}),\ \Theta_{\mathfrak{g},r,\underline{B}}\ =\prod_{s=0}^{r-1}\Theta_{\mathfrak{g},r|s,\underline{B}}\quad\in\ k(\underline{B})\otimes k\mathbb{G}_{(r)},
elements of the group algebra of
\mathbb{G}_{(r),k(\underline{B})}
.
(3)
Both
\Theta_{\mathfrak{g},r|s}
and
\Theta_{\mathfrak{g},r}
are homogeneous of degree
p^{r-1}
with
respect to the grading on
k[\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))]
.
(4)
For any
\mathbb{G}_{(r)}
-module
M
, tensor product with
\Theta_{\mathfrak{g},r|s}
and with
\Theta_{\mathfrak{g},r}
determines
p
-nilpotent,
k[\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))]
-linear operators
\Theta_{\mathfrak{g},r|s,M},\ \Theta_{\mathfrak{g},r,M}:k[\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))]\otimes M\ \to\ k[\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))]\otimes M.
(5)
The base changes along a scheme theoretic point
\underline{B}\in\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))
of the operators
\Theta_{\mathfrak{g},r|s,M},\ \Theta_{\mathfrak{g},r,M}
are given by the actions of
\Theta_{\mathfrak{g},r|s,\underline{B}},\ \Theta_{\mathfrak{g},r,\underline{B}}
on
k(\underline{B})\otimes M
.
Proof.
Observe that the elements
\Theta_{\mathfrak{g},r|s},\ \Theta_{\mathfrak{g},r}
are
p
-nilpotent because
1\otimes u_{r-1}\in\mathbb{G}_{a(r),\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))]}
is
p
-nilpotent.
Assertion (1) follows from Proposition
4.5
(2), the definition of
\Theta_{\mathbb{G},r}
in
Definition
2.13
, and the definition of
\Theta_{\mathfrak{g},r}
given above.
The first part of assertion (2) identifying
\Theta_{\mathfrak{g},r|s,\underline{B}}
follows from Proposition
4.4
,
so that the second part is verified using the fact that specialization along
k[\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))]\to k(\underline{B})
,
k[\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))]\otimes\mathbb{G}_{(r)}\to k(\underline{B})\otimes\mathbb{G}_{(r)}
,
commutes with the product map of
\mathbb{G}_{(r)}
.
Assertion (3) follows from Theorem
2.15
(5).
Assertions (4) and (5) and their proofs are direct analogues of Theorem
2.15
(3),(4).
∎
As an immediate corollary of Theorem
4.8
(1), we have the following.
Corollary 4.9
.
Let
(\mathbb{G},\mathcal{E}_{(r)})
be an affine group scheme of exponential type of height
r
.
Consider a geometric point
\underline{B}
of
\mathcal{C}_{r}(\mathfrak{g})
with image
x=\Phi_{\mathfrak{g},r}(\underline{B})
.
For any finite dimensional
\mathbb{G}_{(r)}
-module
M
,
JT(\Theta_{\mathfrak{g},r,M})(\underline{B}))\quad=\quad JT(\Theta_{\mathbb{G},r,M})(x).
In other words,
JT(\Theta_{\mathfrak{g},r,M})(-)\ =\ JT(\Theta_{\mathbb{G},rM}(-)\circ\Phi_{\mathfrak{g},r}(-):\mathbb{P}\mathcal{C}_{r}(\mathfrak{g})\ \to\ \mathcal{Y},
for any finite dimensional
\mathbb{G}_{(r)}
-module
M
.
We denote by
\mathbb{P}\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))
the weighted projective variety
\operatorname{Proj}\nolimits k[\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))]
, parallel
to the notation of
\mathbb{P}V_{r}(\mathbb{G})
for
\operatorname{Proj}\nolimits k[V_{r}(\mathbb{G})]
. We omit the proof of the following
elementary lemma.
Lemma 4.10
.
A geometric point of
\operatorname{Spec}\nolimits K\to\mathbb{P}\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))
corresponds to an equivalence
class
\langle\underline{B}\rangle
of
non-zero
r
-tuples
\underline{B}=(B_{0},\ldots,B_{r-1})
of
p
-nilpotent, pair-wise commuting elements
of
K\otimes\mathfrak{g}
, where
(B_{0},\ldots,B_{r-1})
is equivalent to
(B_{0}^{\prime},\ldots,B_{r-1}^{\prime})
if and only
it there exists some
0\not=\alpha\in K
such that
B_{i}^{\prime}=\alpha^{p^{i}}B_{i}
for each
i\ 0\leq i2
,
\Theta^{exp}_{\mathfrak{g},r}\ \not=\ \Theta_{\mathfrak{g},r}
.
We next give the evident definition of the “exponential Jordan type function” of a
\mathbb{G}_{(r)}
-module, parallel to the
definition of
JT_{\mathbb{G},r,M}(-)
given in Definition
2.16
.
Definition 4.16
.
Let
(\mathbb{G},\mathcal{E}_{(r)})
be an affine group scheme of exponential type of height
r
and let
M
be a finite dimensional
\mathbb{G}_{(r)}
-module. We define
JT^{exp}_{\mathfrak{g},r,M}(-):\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))\quad\to\quad\mathcal{Y}
by sending a geometric point
\underline{B}:\operatorname{Spec}\nolimits k(\underline{B})\to\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))
to
(4.16.1)
JT(\Theta^{exp}_{\mathfrak{g},r,M})(\underline{B})\quad=\quad JT(\sum_{s=0}^{r-1}(\mathcal{E}_{r|s,B_{s}})_{*}(1\otimes u_{r-s-1}),k(\underline{B})\otimes M))\ ,
the Jordan type of the operator
\Theta^{exp}_{\mathfrak{g},r,\underline{B}}
on
k(\underline{B})\otimes M
.
Observe that the operator
(\mathcal{E}_{r|s,B_{s}})_{*}(1\otimes u_{r-s-1})
on
k(\underline{B})\otimes M
can be identified
with the operator
1\otimes u_{r-s-1}\in k(\underline{B})\otimes k\mathbb{G}_{a(r)}
on
(\mathcal{E}_{r|s,B_{s}})^{*}(k(\underline{B})\otimes M)
.
We next formulate the projectivization
\tilde{\Theta}_{\mathfrak{g},r}^{exp}
of
\Theta_{\mathfrak{g},r}^{exp}
and give
the analogues of Theorem
2.18
for
\tilde{\Theta}_{\mathbb{G},r}
and
Proposition
4.11
for
\tilde{\Theta}_{\mathfrak{g},r}
.
Proposition 4.17
.
Let
(\mathbb{G},\mathcal{E}_{(r)})
be an affine group scheme of exponential type of height
r
.
The operator
\Theta_{\mathfrak{g},r}^{exp}
of Definition
4.12
determines a global section
\tilde{\Theta}_{\mathfrak{g},r}^{exp}\quad\in\quad\Gamma(\mathcal{O}_{\mathbb{P}\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))},\mathcal{O}_{\mathbb{P}\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))}(p^{r-1})\otimes k\mathbb{G}_{(r)})
of the coherent, locally free sheaf
\mathcal{O}_{\mathbb{P}\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))}(p^{r-1})\otimes k\mathbb{G}_{(r)}
of
\mathcal{O}_{\mathbb{P}\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))}
-modules which
satisfies the following properties.
(1)
For any
\mathbb{G}_{(r)}
-module
M
, tensor product with
\tilde{\Theta}_{\mathfrak{g},r}^{exp}
determines the
map of quasi-coherent
\mathcal{O}_{\mathbb{P}\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))}
-modules
\tilde{\Theta}_{\mathfrak{g},r,M}^{exp}:(k[\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))]\otimes M)^{\sim}\quad\to\quad(k[\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))]\otimes M)^{\sim}(p^{r-1}).
(2)
For any non-zero divisior
h\in k[\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))]_{p^{r-1}}
, let
U_{h}
denote the affine open subset
Spec(k[\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))][h^{-1}])_{0}\ \subset\mathbb{P}\mathcal{C}_{r}(\mathfrak{g})
.
The natural isomorphism of
(\mathcal{O}_{\mathbb{P}\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))})_{U_{h}}
-modules
h^{-1}\otimes 1:(\mathcal{O}_{\mathbb{P}\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))}(p^{r-1})\otimes k\mathbb{G}_{(r)})_{|U_{h}}\quad\stackrel{{\scriptstyle\sim}}{{\to}}\quad(\mathcal{O}_{\mathbb{P}\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))}\otimes k\mathbb{G}_{(r)})_{|U_{h}}
sends
(\tilde{\Theta}_{\mathfrak{g},r}^{exp})_{|U_{h}}\ \in\ \Gamma(U_{h},\mathcal{O}_{\mathbb{P}\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))}(p^{r-1})\otimes k\mathbb{G}_{(r)})\ =\ (k[\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))][1/h])_{p^{r-1}}\otimes k\mathbb{G}_{(r)}
to
(h^{-1}\otimes 1)\cdot\Theta_{\mathfrak{g},r}^{exp}\ \in\ k[\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))][1/h]\otimes k\mathbb{G}_{(r)}.
(3)
If
\underline{B}
is a scheme theoretic point of
\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))
projecting to
\langle\underline{B}\rangle
such that
h(\langle\underline{B}\rangle)\not=0
, then the restriction
along
\operatorname{Spec}\nolimits k(\underline{B})\to\mathbb{P}\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))
of
\tilde{\Theta}_{\mathfrak{g},r}^{exp}
is given by
(4.17.1)
k(\underline{B})\otimes_{\mathcal{O}_{\mathbb{P}\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))}}\tilde{\Theta}_{\mathfrak{g},r}^{exp}\quad=\quad c\cdot\sum_{s=0}^{r-1}(\mathcal{E}_{(r),B_{s}})_{*}(1\otimes u_{r-s-1})\ \in\ k(\underline{B})\otimes k\mathbb{G}_{(r)}
for some
0\not=c\in k(\underline{B})
.
(4)
For any
\mathbb{G}_{(r)}
-module
M
, the restriction of
\tilde{\Theta}_{\mathfrak{g},r,M}^{exp}
along
\operatorname{Spec}\nolimits k(\underline{B})\to\mathbb{P}\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))
as in (3) is a non-zero scalar multiple of the image of
\sum_{s=0}^{r-1}(\mathcal{E}_{r|s,B_{s}})_{*}(1\otimes u_{r-s-1})
in
End_{k(\underline{B})}(k(\underline{B})\otimes M)
.
Proof.
By Proposition
2.17
,
\mathcal{O}_{\mathbb{P}\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))}(p^{r-1})\otimes k\mathbb{G}_{(r)}
is a coherent, locally free
sheaf of
\mathcal{O}_{\mathbb{P}\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))}
-modules on
\mathbb{P}\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))
.
Proposition
4.5
(3) implies that
(\Phi_{r|s})^{*}:k[V_{r}(\mathbb{G})]\to k[\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))]
is a
map of graded
k
-algebras so that
\Theta_{\mathfrak{g},r}^{exp}
is homogenous by Theorem
2.15
(5) and Theorem
4.8
(1). Thus,
\Theta_{\mathfrak{g},r}^{exp}
determines
\tilde{\Theta}_{\mathfrak{g},r,M}^{exp}
in
\Gamma(\mathcal{O}_{\mathbb{P}\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))},\mathcal{O}_{\mathbb{P}\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))}(p^{r-1})\otimes k\mathbb{G}_{(r)})
.
The proofs of assertions (1) - (4) are verified by evident analogues for
\tilde{\Theta}^{exp}_{\mathfrak{g},r,}
of the proofs given in the proof of Theorem
2.18
for the corresponding statements for
\tilde{\Theta}_{\mathbb{G},r,}
∎
The following theorem makes explicit the definition of the Jordan type function
JT_{\mathfrak{g},r,M}^{exp}(-)
and states properties analogous to properties for J
T_{\mathbb{G},r,M}^{(}-)
given in Theorem
2.21
.
The proof of Theorem
4.18
follows from Proposition
4.17
exactly as the proof of Theorem
2.21
follows
from Theorem
2.18
.
Theorem 4.18
.
Let
(\mathbb{G},\mathcal{E}_{(r)})
be an affine group scheme of exponential type of height
r
.
For any finite dimensional
\mathbb{G}_{(r)}
-module
M
and any
scheme-theoretic point
\underline{B}
of
\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))
projecting to
\langle\underline{B}\rangle\in\mathbb{P}\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))
,
we define
JT_{\mathfrak{g},r,M}^{exp}(\langle\underline{B}\rangle)\quad\equiv\quad JT(\sum_{s=0}^{p-1}(\mathcal{E}_{r|s,B_{s}})_{*}(1\otimes u_{r-s-1}),k(\underline{B})\otimes M),
the Jordan type of the
p
-nilpotent element
k(\underline{B})\otimes_{\mathcal{O}_{\mathbb{P}\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))}}\tilde{\Theta}_{\mathfrak{g},r}^{exp}\in k(\underline{B})\otimes\mathbb{G}_{(r)}
acting on
k(\underline{B})\otimes M
.
(1)
For any finite dimensional
\mathbb{G}_{(r)}
-module
M
,
the association
\langle\underline{B}\rangle\ \mapsto JT_{\mathfrak{g},r,M}^{exp}(\langle\underline{B}\rangle)
determines a continuous map
JT_{\mathfrak{g},r,M}^{exp}(-):\mathbb{P}\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))\quad\to\quad\mathcal{Y}.
(2)
For
\mathbb{G}_{(r)}
-modules
M,N
of dimensions
m,n
and any scheme-theoretic point
\underline{B}\in\mathbb{P}\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))
,
JT_{\mathfrak{g},r,M\oplus N}^{exp}(\langle\underline{B}\rangle)\ =\ JT_{\mathfrak{g},r,M}^{exp}(\langle\underline{B}\rangle)+JT_{\mathfrak{g},r,N}^{exp}(\langle\underline{B}\rangle)\ \in\ \mathcal{Y}_{m+n}.
(3)
Assume either that
r=1
or that
\langle\underline{B}\rangle\in\mathbb{P}\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))
is a generic point.
For
\mathbb{G}_{(r)}
-modules
M,N
of dimensions
m,n
,
JT_{\mathfrak{g},r,M\otimes N}^{exp}(\langle\underline{B}\rangle)\ =\ JT_{\mathfrak{g},r,M}^{exp}(\langle\underline{B}\rangle)\otimes JT_{\mathfrak{g},r,N}^{exp}(\langle\underline{B}\rangle)\ \in\ \mathcal{Y}_{m\cdot n}.
(4)
If
f:M\to N
is a surjective map of
\mathbb{G}_{(r)}
-modules, then
JT_{\mathfrak{g},r,N}^{exp}(\langle\underline{B}\rangle)\ \leq JT_{\mathfrak{g},r,M}^{exp}(\langle\underline{B}\rangle)
for any
\langle\underline{B}\rangle\in\mathbb{P}\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))
; in other words,
JT_{\mathfrak{g},r,N}^{exp}(-)\leq JT_{\mathfrak{g},r,M}^{exp}(-)
.
(5)
A short exact sequence
0\to N\ \to\ M\ \to\ Q\to 0
of finite dimensional
\mathbb{G}_{(r)}
-modules
is locally split if and only if
JT_{\mathfrak{g},r,M}^{exp}(-)=JT_{\mathfrak{g},r,N}^{exp}(-)+JT_{\mathfrak{g},r,Q}^{exp}(-).
(6)
If
\phi:\mathbb{H}\to\mathbb{G}
is a morphism of affine group schemes of exponential type of height
r
, then
for any finite dimensional
\mathbb{G}_{(r)}
-module
M
JT_{\mathfrak{h},r,\phi*(M)}^{exp}(-):\mathbb{P}\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{h}))\ \to\ \mathcal{Y}
equals
JT_{\mathfrak{g},r,M}^{exp}(-)\circ\phi_{*}:\mathbb{P}\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{h}\ \to\ \mathbb{P}\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))\ \to\ \mathcal{Y}.
5.
Examples of exponential Jordan types
We provide some examples of computations of the simplified
Jordan type function
JT^{exp}_{\mathfrak{g},r,M}(-)
of Theorem
4.18
.
One challenge to computing Jordan types is the behavior of Jordan types with
respect to the sum of two commuting,
p
-nilpotent operators. Another challenge
is the fact that for
r>1
the Jordan type functions
JT_{\mathbb{G},r,M}(-),\ JT_{\mathfrak{g},r,M}^{exp}(-)
rarely commute with tensor products of
\mathbb{G}_{(r)}
-modules at all points of
\mathbb{P}\mathcal{C}_{r}(\mathfrak{g})
.
Proposition 5.1
.
By Proposition
4.13
, Examples
3.1
and
3.2
remain valid if
(\mathbb{G},\mathcal{E}_{(r)})
is an affine group scheme of exponential type of height
r
and if
JT_{\mathbb{G},r,M}(-)
is replaced by
JT^{exp}_{\mathfrak{g},r,M}(-)
.
The justification of Example
3.3
also justifies its validity
when
JT_{\mathbb{G},r,M}(-)
is replaced by
JT^{exp}_{\mathfrak{g},r,M}(-)
.
Example
3.6
remains unchanged, since
JT_{\mathbb{G},1,M}(-)\ =\ JT^{exp}_{\mathfrak{g},1,M}(-)
.
On the other hand, as we see in the following examples, some computations simplify if we replace
JT_{\mathbb{G}_{(r)},M}(-)
by
JT^{exp}_{\mathfrak{g},M}(-)
.
Example 5.2
.
Adopt the notation of Examples
2.10
and
2.14
.
Let
V
be an
n
-dimensional
k
-vector space, and let
\alpha_{0},\ldots\alpha_{r-1}
be
be pair-wise commuting
p
-nilpotent endomorphisms of
V
. Denote by
M
the
(\mathbb{G}_{a}^{\times r})_{1}
-module such that
x_{i}\in k(\mathbb{G}_{a}^{r})_{1}
acts as
\alpha_{i}
and denote
by
N
the
\mathbb{G}_{a(r)}
-module such that
u_{i}\in k\mathbb{G}_{a(r)}
also acts as
\alpha_{i}
.
Recall from Example
2.10
the morphism
\Phi:\mathbb{P}^{r-1}\to w\mathbb{P}(1,p,\ldots,p^{r-1})
.
Since
\Theta_{\mathbb{G}_{a,r}}^{exp}\ =\ \sum_{i=0}^{r-1}T_{i}^{p^{r-1-i}}\otimes u_{i}\ \in\ k[V_{r}(\mathbb{G}_{a(r)})]\otimes k\mathbb{G}_{a(r)}
,
\Phi^{*}(\Theta_{\mathbb{G}_{a},r}^{exp})\quad=\quad\sum_{i=0}^{r-1}t_{i}^{p^{r-1}}\otimes x_{i}\ =\ F^{r-1*}(\Theta_{\mathbb{G}_{a}^{\times r},1}).
Thus,
JT_{\mathfrak{g}_{a},r,N}^{exp}(-)\circ\Phi\ =\ JT_{\mathbb{G}_{a}^{\times r},1,M}\circ F^{r-1}:\ \mathbb{P}^{r-1}\ \to\ \mathcal{Y}.
In the following examples, we use (
4.16.1
) of Definition
4.16
which becomes
JT^{exp}_{\mathfrak{g},r,M}(\underline{B})\quad=\quad JT(\sum_{s=0}^{r-1}(exp_{B_{s}})_{*}(1\otimes u_{r-s-1}),k(\underline{B}))
provided that
(\mathbb{G},\mathcal{E}_{r})
is associated with an embedding
i:\mathbb{G}\hookrightarrow GL_{N}
of
exponential type of height
r
.
Example 5.3
.
Set
\mathbb{G}=SL_{2(2)}
, so that
\mathfrak{g}=sl_{2}
. We consider the irreducible
\mathbb{G}
-module
M\ =\ S(\lambda_{0})\otimes S(\lambda_{1})^{(1)},\quad 0\leq\lambda_{0},\lambda_{1}i
and proceed to determine
JT_{\mathfrak{g}l_{N},r,V^{(i)}}(\underline{B})
for
\underline{B}=(B_{0},\ldots,B_{r-1})
, an
r
-tuple of
p
-nilpotent, pairwise commuting
elements of
k(\underline{B})\otimes\mathfrak{g}l_{N}
.
Then the action of
(\exp_{B_{s}})_{*}(u_{r-s-1})
on
V^{(i)}
is non-zero
if and only if
i=r-s-1
, in which case this action equals the action of
B_{s}\in k(\underline{B})\otimes\mathfrak{g}l_{N}
acting on
k(\underline{B})\otimes V
. Consequently,
JT_{\mathfrak{g}l_{N},r,V^{(i)}}^{exp}(\underline{B})=JT(B_{s},V)
where
s=r-i-1
.
More generally, we can take a polynomial representation
V_{s}
of
GL_{N}
for
each
s,\ 0\leq sm
. In this case, we would be comparing
\mathbb{A}^{1}
-equivalent vector bundles on the singular variety
w\mathbb{P}(1,p,\ldots,p^{r-1})
.
An approach might be to look at local invariants of these bundles at singular points of
w\mathbb{P}(1,p,\ldots,p^{r-1})
.
7.
Jordan types for linear algebraic groups
In
[
12
, Defn 4.4]
, we formulated the local Jordan type of a finite dimensional rational
\mathbb{G}
-module
M
for
\mathbb{G}
a linear algebraic group equipped with a structure of exponential type.
There is some subtlety to this construction in order to “package” the Jordan types of
M_{|\mathbb{G}_{(r)}}
of restrictions of
M
to all Frobenius kernels of
\mathbb{G}
. This involves reordering
r
-tuples of
p
-nilpotent,
pairwise commuting elements
(B_{0},\ldots,B_{r-1})
of
\mathfrak{g}
so that they appear as
(B_{r-1},\ldots,B_{0})
.
The definition of the Jordan type of a finite dimensional rational
\mathbb{G}
-module
M
is then formulated
in terms of the action of a colimit of
p
-nilpotent operators, rather than as a colimit of Jordan
types of restrictions
M_{|\mathbb{G}_{(r)}}
of
M
to the Frobenius kernels
\mathbb{G}_{(r)}
of
\mathbb{G}
.
In this final section, we observe that this approach leads to Jordan type functions
for finite dimensional rational
\mathbb{G}
-modules
M
provided that
\mathbb{G}
is of “exponential type of infinite height”. In Definition
7.3
, we extend the construction of
JT^{exp}_{\mathfrak{g},r,M}(-):\mathcal{C}_{r}(N_{p}(\mathfrak{g}))\ \to\ \mathcal{Y}
of Definition
4.16
for
\mathbb{G}_{(r)}
-modules
to
JT_{\mathbb{G},\infty,M}^{exp}(-):\mathcal{C}_{\infty}(\mathfrak{g})\to\mathcal{Y}
for finite dimensional
rational
\mathbb{G}
-modules. As shown in Proposition
7.4
(correcting an error of
[
12
, Prop 4.8]
),
JT_{\mathbb{G},\infty,M}^{exp}(-)
is a twisted stabilization of
JT_{\mathbb{G},r,M_{|\mathbb{G}_{(r)}}}^{exp}(-)
.
We say that an affine group scheme
\mathbb{G}
of finite type over
k
is a linear algebraic group
if
\mathbb{G}
is reduced and irreducible. In what follows,
k\mathbb{G}\ \equiv\varinjlim_{r}k\mathbb{G}_{(r)}
, the
“distribution algebra” of
\mathbb{G}
. We define
\mathcal{C}_{\infty}(\mathcal{N}_{p}(\mathfrak{g}))
to be the ind-scheme
\mathcal{C}_{\infty}(\mathcal{N}_{p}(\mathfrak{g}))\ \equiv\ \varinjlim_{r}\mathcal{C}_{r}(N_{p}(\mathfrak{g}));\quad\mathcal{C}_{r}(N_{p}(\mathfrak{g}))\to\mathcal{C}_{r+1}(\mathcal{N}_{p}(\mathfrak{g})),
where
(B_{0},\ldots,B_{r-1})\mapsto(B_{0},\ldots,B_{r-1},0).
We define the “coordinate algebra”
k[\mathcal{C}_{\infty}(\mathcal{N}_{p}(\mathfrak{g}))]\quad\equiv\quad\varinjlim_{r}k[\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))].
We define a scheme-theoretic (respectively, geometric) point of
\mathcal{C}_{\infty}(\mathcal{N}_{p}(\mathfrak{g}))
to be an
element in the direct limit of scheme-theoretic (respectively, geometric) points of
\{\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))\}
.
We can view such a point
\iota_{x}:\operatorname{Spec}\nolimits K\to\mathcal{C}_{\infty}(\mathcal{N}_{p}(\mathfrak{g}))
as given by a “continuous” map of
algebras
\iota_{x}^{*}:k[\mathcal{C}_{\infty}(\mathcal{N}_{p}(\mathfrak{g}))]\to K
.
We provide analogues of the constructions given in Construction
4.3
.
Definition 7.1
.
Let
\mathbb{G}
be a linear algebraic group with Lie algebra
\mathfrak{g}
, equipped with a morphism
\mathcal{E}:\mathbb{G}_{a}\times\mathcal{N}_{p}(\mathfrak{g})\quad\to\quad\mathbb{G}
such that
(\mathbb{G}_{(r)},\ \mathcal{E}_{|r})
constitutes an affine group scheme of exponential
type of height
r
for each
r>0
.
Here,
\mathcal{E}_{|r}:\mathbb{G}_{a(r)}\times\mathcal{N}_{p}(\mathfrak{g})\to\mathbb{G}_{(r)}
is the restriction of
\mathcal{E}
.
We say that
\mathbb{G}
is of exponential type of infinite height.
For any (geometric) point
B
of
\mathcal{N}_{p}(\mathfrak{g})
, we denote by
\mathcal{E}_{|r,B}:k(B)\times\mathbb{G}_{a(r)}\ \to\ k(B)\otimes\mathbb{G}
the restriction of
\mathcal{E}_{|r}
along
\operatorname{Spec}\nolimits k(B)\to\mathcal{N}_{p}(\mathfrak{g})
.
We define
\mathcal{E}_{\infty|r}\ =\ \mathcal{E}_{|r}\circ(1\times pr_{s}):\mathbb{G}_{a(r)}\times\mathcal{C}_{\infty}(\mathcal{N}_{p}(\mathfrak{g}))\ \to\ \mathbb{G}_{a(r)}\times\mathcal{N}_{p}(\mathfrak{g})\ \to\ \mathbb{G}_{(r)}
and set
\mathcal{U}_{\mathfrak{g},\infty|r}\ \equiv\ \mathcal{E}_{\infty|r}\times pr_{\mathcal{C}_{\infty}(\mathcal{N}_{p}(\mathfrak{g}))}:\mathbb{G}_{a(r)}\times\mathcal{C}_{\infty}(\mathcal{N}_{p}(\mathfrak{g}))\ \to\ \mathbb{G}_{(r)}\times\mathcal{C}_{\infty}(\mathcal{N}_{p}(\mathfrak{g})).
We define the operator (a formal limit of well defined partial sums)
(7.1.1)
\Theta_{\mathfrak{g},\infty}^{exp}\ \equiv\ \sum_{s\geq 0}(\mathcal{U}_{\mathfrak{g},\infty|s})_{*}(1\otimes u_{s})\ =\sum_{s\geq 0}\in\ k[\mathcal{C}_{\infty}(\mathcal{N}_{p}(\mathfrak{g}))]\otimes k\mathbb{G}.
As verified in
[
12
, Prop 2.6]
, for any linear algebraic group
(\mathbb{G},\mathcal{E})
of exponential
type of infinite height, any finite dimensional
\mathbb{G}
-module
M
,
any element
m\in M
, and any
B\in\mathcal{N}_{p}(\mathfrak{g})
, the element
(\mathcal{E}_{|r,B_{s}})_{*}(1\otimes u_{s})
in
k[\mathcal{C}_{\infty}(\mathcal{N}_{p}(\mathfrak{g}))]\otimes k\mathbb{G}
acts trivially on
m
for all sufficiently large
s
.
This justifies the following proposition which should be compared with Definition
4.16
.
Proposition 7.2
.
For any finite dimensional
\mathbb{G}
-module
M
, the formal sum
\Theta_{\mathfrak{g},\infty}^{exp}
of (
7.1.1
)
determines a well defined
p
-nilpotent
k[\mathcal{C}_{\infty}(\mathfrak{g})]
-linear operator
\Theta_{\mathfrak{g},\infty}^{exp}:k[\mathcal{C}_{\infty}(\mathcal{N}_{p}(\mathfrak{g}))]\otimes M\ \to\ k[\mathcal{C}_{\infty}(\mathcal{N}_{p}(\mathfrak{g}))]\otimes M.
The specialization at a scheme-theoretic point
\underline{B}\in\mathcal{C}_{\infty}(\mathcal{N}_{p}(\mathfrak{g}))
of
\Theta_{\mathfrak{g},\infty}^{exp}
determines
the operator
\Theta_{\mathfrak{g},\infty,\underline{B}}^{exp}\ =\ \sum_{s\geq 0}(\mathcal{E}_{|s,B_{s}})_{*}(1\otimes u_{s-1}):k[\underline{B}]\otimes M\ \to\ k[\underline{B}]\otimes M,
(denoted
\sum_{s\geq 0}(\mathcal{E}_{B_{s}})_{*}(u_{s})
in
[
12
]
).
The following definition of
JT_{\mathbb{G},\infty,M}(\underline{B})
is essentially that of
[
12
, Defn 4.4]
.
Definition 7.3
.
For any finite dimensional
\mathbb{G}
-module
M
,
we define the Jordan type of
M
at
\underline{B}\in\mathcal{C}_{\infty}(\mathfrak{g})
to be
(7.3.1)
JT_{\mathfrak{g},\infty,M}^{exp}(\underline{B})\quad\equiv\quad JT(\sum_{s\geq 0}(\mathcal{E}_{|s,B_{s}})_{*}(1\otimes u_{s}),k(\underline{B})\otimes M),
the Jordan type of the
p
-nilpotent action of
k(\underline{B})\otimes_{k[\mathcal{C}_{\infty}(\mathcal{N}_{p}(\mathfrak{g}))]}\Theta_{\mathfrak{g},\infty}^{exp}\in k(\underline{B})\otimes k\mathbb{G}
acting on
k(\underline{B})\otimes M
.
Thus, (
7.3.1
) determines the Jordan type function
JT_{\mathfrak{g},\infty,M}^{exp}(-):\ \mathcal{C}_{\infty}(\mathcal{N}_{p}(\mathfrak{g}))\quad\to\quad\mathcal{Y}.
We point out that
[
12
, Prop 4.8]
is wrong as stated. The following proposition provides a correction.
The proof of Proposition
7.4
is an immediate consequence of the observation
above that
(\mathcal{E}_{|r,B_{s}})_{*}(1\otimes u_{s})
in
k[\mathcal{C}_{\infty}(\mathfrak{g})]\otimes k\mathbb{G}
acts trivially on
m\in M
for all sufficiently large
s
.
Proposition 7.4
.
Adopt the notation and hypotheses of Definition
7.3
. If
M
is finite dimensional,
then
JT_{\mathbb{G},\infty,M}^{exp}(-)\ =\ JT_{\mathbb{G},r,M_{|\mathbb{G}_{(r)}}}^{exp}(-)\circ\Lambda^{r}:\ \mathcal{C}_{\infty}(\mathcal{N}_{p}(\mathfrak{g}))\ \to\ \mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))\ \to\ \mathcal{Y},\quad r\gg 0,
where
\Lambda^{r}:\mathcal{C}_{\infty}(\mathcal{N}_{p}(\mathfrak{g}))\to\mathcal{C}_{r}(\mathcal{N}_{p}(\mathfrak{g}))
is given by
sending
(B_{0},\ldots,B_{n},\ldots)
to
(B_{r-1},\ldots,B_{0})
.
In particular,
JT_{\mathbb{G},\infty,M}(-)
is continuous.
One can interpret the Proposition
7.4
as saying for
M
a finite dimensional
\mathbb{G}
-module that the Jordan type function
JT_{\mathbb{G},\infty,M}^{exp}(-)
is a twisted stabilization
(with respect to increasing
r
) of the exponential Jordan type functions
JT_{\mathbb{G},r,M_{|\mathbb{G}_{(r)}}}^{exp}(-)
of Definition
4.16
.