Japan. J. Math.

4, 1–25 (2009)
DOI: 10.1007/s11537-009-0852-x

Holonomic -modules and positive characteristic

Maxim Kontsevich
Received: 8 December 2008 / Revised: 6 January 2009 / Accepted: 11 February 2009
Published online: ?????
c The Mathematical Society of Japan and Springer 2009

Communicated by: Toshiyuki Kobayashi

Abstract. We discuss a hypothetical correspondence between holonomic -modules on an al-

gebraic variety X defined over a field of zero characteristic, and certain families of Lagrangian
subvarieties in the cotangent bundle to X . The correspondence is based on the reduction to posi-
tive characteristic.
Keywords and phrases: rings of differential operators and their modules, characteristic p methods
Mathematics Subject Classification (2000): 13N10, 13A35

1. Quantization and -modules

1.1. Quantization

Let us recall the heuristic dictionary of quantization.

symplectic C∞ -manifold (Y, ω ) ←→ complex Hilbert space H

complex-valued functions on Y ←→ operators in H
Lagrangian submanifolds L ⊂ Y ←→ vectors v ∈ H

In the basic example of the cotangent bundle Y = T ∗ X the space H is L2 (X),

functions on Y which are polynomial along fibers of the projection Y = T ∗ X →
X correspond to differential operators, and with the Lagrangian manifold L ⊂ Y
 This article is based on the 5th Takagi Lectures that the author delivered at the University of
Tokyo on October 4 and 5, 2008.
M. KONTSEVICH
Institut des Hautes Études Scientifiques, 35 route de Chartres, Bures-sur-Yvette 91440, France
(e-mail: [email protected])
2 M. Kontsevich

of the form L = graph dF for some function F ∈ C∞ (Y ) we associate (approxi-

mately) vector exp(iF/) where  → 0 is a small parameter (“Planck constant”).
Our goal in these lectures is to give an evidence for a hypothetical analog of
the quantization in the algebraic case, based on the reduction to positive char-
acteristic, briefly mentioned at the end of [3]. The idea to study quantization in
positive characteristic was also used in [14], and in the fundamental article [12].
It turns out that the correspondence between the classical and the quantum in
the algebraic case is not one-to-one, but only between certain natural families of
Lagrangian manifolds (or cycles) and of holonomic -modules, closely related
to integrable systems. We formulate a series of conjectures about this correspon-
dence. In the case of one variable (which in a sense contain all the keys to the
general case) one can make all the constructions explicit and elementary.

1.2. Reminder about holonomic -modules

In the algebraic setting there is no obvious analog of the Hilbert space, even
in the case of the cotangent bundle. A possible replacement for the notion of a
function is the one of a holonomic -module. Here we recall the definition and
several basic and well-known facts (the standard reference is [4]).
Let X be a smooth affine algebraic variety over field k of zero characteristic,
dim X = n. The ring (X) of differential operators is k-algebra of operators
acting on  (X), generated by functions and derivations:
f → g f , f → ξ ( f ), g ∈  (X), ξ ∈ Γ(X, TX/Spec k ).

Algebra (X) carries the filtration (X) = k≥0 ≤k (X) by the degree of op-
erators, the associated graded algebra is canonically isomorphic to the algebra
of functions on T ∗ X. In geometric terms, the grading comes from the dilation
by Gm along the fibers of the cotangent bundle.
Let M be a finitely generated (X)-module, and choose a finite-dimensional
subspace V ⊂ M generating M. Then consider the filtration
M≤k := ≤k (X) ·V ⊂ M, k ≥ 0.
The associated graded module gr(M) is a finitely generated  (T ∗ X)-module.
By theorem of Gabber, its support (which is a reduced conical subscheme in
T ∗ X)
supp (gr(M)) ⊂ T ∗ X
is coisotropic. In particular, the dimension of any irreducible component is at
least n = dim X. The support does not depend on the choice of the generating
space V , and is denoted by supp (M).
A finitely generated module M is called holonomic if and only if the dimen-
sion of its support is exactly n. Let (Li )i∈I be the set of irreducible components
Holonomic -modules and positive characteristic 3

of supp (M). Each Li is the generic point of a conical Lagrangian subvariety of

T ∗ X. One can show that the multiplicity mi ≥ 1 of gr(M) at Li does not depend
on the choice of generators as well. The notion of support with multiplicities is
covariant with respect to automorphisms of X (or, more generally, contravariant
with respect to étale maps).
For general smooth X, not necessarily affine, we have a sheaf (in Zariski
topology) X of algebras of differential operators. Then one can define X -
modules (and also holonomic modules) and their support. Finitely generated
X -modules form a noetherian abelian category. For any two holonomic X -
modules M1 , M2 we have
dim Exti (M1 , M2 ) < ∞ ∀i ∈ Z,
Exti (M1 , M2 ) = 0 for i < 0 or i > n = dim X.
Any algebraic vector bundle (a locally trivial sheaf of X -modules)  ∈ Coh(X)
endowed with a flat connection ∇ has a natural structure of a holonomic X -
module. Its support (with multiplicities) is the zero section of T ∗ X taken with
the multiplicity equal to rank( ). For any holonomic X -module M there exists
a non-empty Zariski open subset U ⊂ X such that the restriction of M is a bundle
with a flat connection.
In the case of the affine space X = Ank one can use another filtration called
the Bernstein filtration. Namely, the algebra (Ank ) is the n-th Weyl algebra An,k
over1 k, i.e., it has the presentation
kx̂1 , . . . , x̂2n /([x̂i , x̂ j ] = ωi j , ∀i, j 1 ≤ i, j ≤ 2n),
where
ωi j = δi,n+ j − δ j,n+i .
The generators are realized as
x̂i = xi , x̂n+i = ∂ /∂ xi , 1 ≤ i ≤ n.
The filtration is given by the degree in generators. In this case the notion of a
holonomic module is the same, but the support is now a Lagrangian cone in
A2n , invariant with respect to the total dilation in A2n . Notice that the notion of
a holonomic module over An,k and its support is covariant with respect to the
action of the symplectic group Sp(2n, k).

1 Same formulas give the definition of algebra An,R for arbitrary commutative ring R.
4 M. Kontsevich

1.3. Lagrangian cycles

Let (Y, ω ) be a symplectic smooth quasi-projective algebraic variety over a field

k of arbitrary characteristic.
Definition 1. A Lagrangian subvariety in Y is the generic point of a smooth
(not necessarily closed) irreducible Lagrangian submanifold L ⊂ Y . An effective
Lagrangian cycle is a formal linear combination with coefficients in Z≥0 of
Lagrangian subvarieties.
Effective Lagrangian cycles form a submonoid in the abelian group of algebraic
cycles Z n (Y ) where n := dimY /2.
Let us assume that Y is quasi-projective, fix a projective embedding i : Y →
Pk and a constant d > 0. Then the set of effective Lagrangian cycles of de-
N

gree ≤ d with respect to i has a natural structure of the set of k-points of a

constructible set defined over k. If we do not bound the degree, we obtain
an ind-constructible set over k which we will denote by ELC(Y ). In the case
char(k) = 0 it is not inconceivable that ELC(Y ) is in fact an ind-scheme, not
just merely an ind-constructible set. Also, the same could be true for Lagrangian
cycles of a bounded degree when char(k) is large enough. In Section 3.1 we
present some evidence for it.
In the case char(k) = 0 the support (with multiplicities) of a holonomic X -
module is a k-point of ELC(T ∗ X) invariant under the dilation by Gm along fibers
of the bundle T ∗ X → X. In the case X = Ank and of the Bernstein filtration the
support is invariant under the total dilation.

2. Reduction of -modules to positive characteristic

2.1. Differential operators in positive characteristic

Let X be a smooth affine scheme over an arbitrary commutative ring R. We

define the ring (X) of differential operators on X as an R-linear associative
algebra generated additively by  (X) and by T (X) := Γ(X, TX/Spec R ) subject to
the following relations:
f · f = f f , f ·ξ = fξ, ξ · f − f · ξ = ξ ( f ), ξ · ξ − ξ · ξ = [ξ , ξ ]
where f , f ∈  (X), ξ , ξ ∈ T (X).
This definition is equivalent to the usual one if R = k is a field of charac-
teristic zero. In general, (X) maps to the algebra of differential operators in
the sense of Grothendieck, but the map is neither surjective, nor injective. The
surjectivity fails e.g. for
R = Z, X = A1R ,  (X) = Z[x].
Holonomic -modules and positive characteristic 5

In this case the divided powers of d/dx acting on  (X)

 
(d/dx)n m m m−n
: x → x
n! n
do not belong to the image of (X) for n ≥ 2.
The injectivity fails in positive characteristic:

R = Z/pZ, Y = A1R , (d/dx) p → 0 ∈ End( (X)).

In general, for R in characteristic p > 0, i.e., p · 1R = 0, the algebra (X)

has a big center:
Center( (X))   (T ∗ X )
where X /SpecR is the pullback of X under the Frobenius map

FrR : Spec R → Spec R, Fr∗R (a) = a p .

Moreover, the algebra (X) is an Azumaya algebra of its center, it is a twisted

form of the matrix algebra Mat(pn × pn ,  (T ∗ X )) where n = dim X. In the
basic example of the Weyl algebra An,R := (AnR ), the center is the algebra of
polynomials R[x̂1p , . . . , x̂2n
p
].

2.2. Infinitely large prime

It will be convenient to introduce the following notation (“reduction modulo

infinitely large prime”) for an arbitrary commutative ring R:
  
R∞ := lim 
∏ (R ⊗ Z/pZ) 
(R ⊗ Z/pZ) .
−→

f .g. R⊂R primes p primes p

Here the inductive limit is taken over the filtered system consisting of all
finitely generated subrings R ⊂ R, and the index p runs over primes 2, 3, 5, . . ..
It is easy to see that the ring R∞ is defined over Q (all primes are invertible in
R∞ ), and the obvious map R → R∞ gives an inclusion R ⊗ Q → R∞ . Also, there
is a universal Frobenius endomorphism given by

Fr∗R∞ : R∞ → R∞ , Fr∗R∞ (a p )primes p := (a pp )primes p .

Now we discuss a related notion. Let R = k be a field of characteristic zero,

and S/k be a constructible set. We define the set of “twisted points modulo large
primes” as

tw
S∞ := lim ∏ (C.S. : 
S
→ Spec R ⊗ Z/pZ) ,
−→ 
R,p

f .g. R⊂k, S primes p
6 M. Kontsevich

where the limit is taken over pairs consisting of a finitely generated subring
R ⊂ k and an affine scheme of finite type S/ R endowed with a constructible
over k bijection between S ×Spec R Spec k and S. The scheme SR,p

 is defined
as the pullback under Fr  of the scheme S × Spec Z/pZ. Finally,
Spec (R⊗Z/pZ),p
the abbreviation C.S. means the set of constructible sections of a morphism of
schemes of finite type.
Notice that we automatically identify collections of constructible sections
which differ only at a finite set of primes. The reason is that R = k contains Q
and we can always add to R inverses of any finite set of primes.
In the special case S = A1k we have an embedding
k∞ → (A1k )tw
∞.
For an ind-constructible set over k represented as a countable limit of con-
structible sets S = lim Si we define S∞
tw as the inductive limit of sets (S )tw .
i ∞
−→

2.3. Reduction of finitely generated -modules

Let X be a smooth affine2 variety over field k, char(k) = 0 and M be a finitely
generated (X)-module. We also choose a projective embedding i : T ∗ X →
PNk . Noetherianity of (X) implies that M is the cokernel of a morphism of
free finitely generated (X)-modules. Therefore, there exists a finitely gener-
ated ring R ⊂ k such that variety X, embedding i and module M have models
XR , iR , MR over Spec R. We assume that XR is a smooth affine variety over Spec R
and MR is a finitely presented (XR )-module.
For any prime p we obtain a finitely generated module MR ⊗ Z/pZ over non-
commutative ring (XR ) ⊗ Z/pZ which is a finitely generated module over its
center. In particular, we can consider MR ⊗ Z/pZ as a module over the center (it
is again finitely generated). Hence, for any prime p and for any point v ∈ Spec R p
(here R p := R⊗Z/pZ), with the residue field kv , we obtain a reduced subscheme
over kv
supp p,v (MR ) ⊂ T ∗ XR p ×Spec R p Spec kv .
Here XR p is the Frobenius pullback of XR p := X ×Spec R Spec R p . Then one has
the following easy result (the proof is omitted here).
Proposition 1. The dimension of supp p,v (MR ) coincides with the dimension of
supp (M) for large enough prime (and all points v ∈ Spec R p ).
Projective embedding iR : TX∗R → PN induces an embedding i R p : TX∗ → PN .
Rp
Therefore, one can speak about the degree of supp p,v (MR ) via the embedding
i R p , as the sum of degrees of closures in PNkv of generic points of top-dimensional
components of supp p,v (MR ). It seems that following holds:
2 All the considerations here extend to the case of not necessarily affine X .
Holonomic -modules and positive characteristic 7

Conjecture 1. The degree of supp p,v (MR ) is bounded above by const · pr where
r = dim supp (M) − dim X.
We have a good evidence for this conjecture in the case of cyclic (X)-modules
of the form M = (X)/ (X)·P where P = 0 is a non-zero differential operator,
see section 3.2 for the special case X = A1k and section 5.1 for X = Ank , n > 1.
In the case of holonomic -modules the conjecture implies that the degree
is uniformly bounded. Also, one can expect an analog of Gabber theorem:
Conjecture 2. In the above notation, for holonomic M the support supp p,v MR is
Lagrangian for sufficiently large p and any v.
Let us assume the above conjecture, and let Li be the generic point of an irre-
ducible component of supp p,v (MR ). Then the length of
Mv := MR ⊗R kv
at Li is divisible by pdim X . Hence, we have an effective algebraic cycle on T ∗ Xv
where Xv := XR p ×Spec R p Spec kv given by

lengthLi Mv
p,v (MR ) := ∑
suppnum [Li ].
i pdim X

2.4. Arithmetic support

Let us use the notation form section 2.1, and assume that module M is holo-
nomic. Let us assume3 conjectures 1 and 2. Then there exists prime p0 such
that for any p ≥ p0 and any v ∈ Spec R p we have an effective Lagrangian cycle
∗
p,v (MR ) in T Xv . Let us replace R by its localization obtained by inverting
suppnum
all primes < p0 .

Definition 2. The arithmetic support suppnum arith (M) of M is an element of

∗
ELC(T X)∞ uniquely specified by the condition that for a model MR , XR over a
tw

finitely generated ring R ⊂ k as above, it is given by the collection of Lagrangian

p,v (MR ).
cycles suppnum

It is easy to see that the definition is consistent, i.e., that for any model the
collection of cycles comes from a collection of constructible maps for all suffi-
ciently large primes, and that models form a filtered system.
The main advantage of the arithmetic support is that it gives a more elaborate
signature of a holonomic module, and the Lagrangian cycle is no longer conical
3 In fact, conjectures 1, 2 are very plausible statements which should not be hard to prove. On
the contrary, all the further conjectures made in the paper seem to really deserve their name, and
need some new ideas.
8 M. Kontsevich

in general. In the next subsection we will give explicit non-conical examples of

arithmetic supports.
The usual support describes only the limiting behavior at infinity (along
fibers of the cotangent bundle T ∗ X → X) of the arithmetic one. More precisely,
there is a projection ELC(T ∗Y ) → ELC(T ∗Y ) which associates with every La-
grangian cycle its limit under the contraction by λ ∈ Gm , λ → 0. This limit is
automatically conical. We expect that the conical limit of the arithmetic support
coincides with the pullback by the Frobenius of the usual support. The same can
be said about the Bernstein filtration and the corresponding support in the case
X = Ank .
The arithmetic support is covariant with respect to automorphisms of X (and
also contravariant for étale maps), as well as under symplectic affine transfor-
mations of A2n ∗ n
k = T Ak in the case X = Ak . Hence, we see that it naturally
n

generalizes two classical types of supports.

Obviously, the arithmetic support behaves additively for extensions of X -
modules, hence it is sufficient to study it only for simple holonomic modules.
p,v (MR ) for infinitely many pairs (p, v) with p → ∞ is just one
Also, if suppnum
irreducible Lagrangian subvariety taken with multiplicity one, then M is simple.
Morally, we should think about the arithmetic support as about Lagrangian
cycle defined over k∞ , this idea is elaborated further in section 3.1. The version
with constructible maps presented here is a surrogate for the “right” version in
section 3.1.

2.5. Examples: p-curvature, exponents, fractional powers and Gauss–Manin

connections

Let M be holonomic X -module corresponding to a vector bundle  over X/k

with flat connection ∇. Let us choose a model of (X,  , ∇) over a finitely gener-
ated ring R ⊂ k. Then for each prime p we obtain a bundle  p with flat con-
nection over a smooth scheme X p /Spec R p in characteristic p, where R p :=
R ⊗ Z/pZ. The p-curvature of such a connection is a p-linear map
TXp /Spec R p →  nd  p , ξ → (∇ξ ) p − ∇ξ p .
Moreover, the image of this map consists of commuting operators, hence we can
interpret p-curvature as the Higgs bundle structure on  . More precisely, it is
a coherent sheaf Higgs on T ∗ X p where X p /SpecR p is the pullback of X p under
the Frobenius map FrR p ,p : Spec R p → Spec R p , together with an isomorphism of
coherent sheaves on X p :
(prT ∗ Xp →Xp )∗ Higgs  (prXp →X )∗  .
It follows directly from definitions that the arithmetic support of M (at prime p)
is the same as the support of Higgs (compare with [12]).
Holonomic -modules and positive characteristic 9

There are two cases when one can easily calculate the arithmetic support.
First, for any F ∈  (X) we have an associated holonomic X -module given by
the trivial line bundle X endowed with the flat connection

∇ = d + (dF ∧ ·).

One can think about this X -module as exp(F) · X . We claim that the corre-
sponding arithmetic support is the pullback by the universal Frobenius Frk∞ of
the graph of differential form −dF. This follows easily from the identity
d dG p  d p  dG p
+ = +
dx dx dx dx
which is held in (R[x]) for any ring R over Z/pZ and any element G ∈ R[x]
(see proposition 3 in [3]). In particular, for the case X = Ank and F polynomial of
degree ≤ 2, the support is the pullback by the Frobenius of the affine Lagrangian
∗
subspace in A2n k = T X corresponding to F.
One can also calculate the arithmetic support for the connection on the trivial
bundle corresponding to a closed but not exact 1-form. For example, for X =
Spec k[x, x−1 ] and for the connection on  := X given by 1-form λ dx/x for
some λ ∈ k, the arithmetic support is the curve in T ∗ X given by the equation

x(p) y(p) = λ p − λ (mod p).

Here x(p) , y(p) are coordinates on T ∗ X p ⊂ T ∗ A1k with symplectic form dx(p) ∧
dy(p) . In the case λ ∈ Q the expression (λ p − λ ) (mod p) vanishes for all suf-
ficiently large p, hence the arithmetic support of this X -module is just the zero
section of T ∗ X p . If λ ∈
/ Q then the arithmetic support is not equal to the zero
section, as follows from Chebotarev density theorem in the case when λ is al-
gebraic, and by elementary reasons when λ is transcendental.
Finally, let M be a holonomic X -module corresponding to the vector bundle
 on X endowed with a flat connection ∇ of Gauss–Manin type (for variations
of pure motives). This means that ( , ∇) is a subquotient of the natural con-
nection of the bundle of de Rham cohomology of fibers of a smooth projective
morphism Y → X. Then by a classical result of N. Katz (see [8]) the p-curvature
is nilpotent. Hence the support is the zero section of the cotangent bundle, taken
with the multiplicity equal to rank  .

3. One-dimensional case

3.1. From higher-dimensional case to A1k

Here we sketch a geometric construction which reduces the study of arithmetic

supports in higher dimensions to the case of A1k .
10 M. Kontsevich

First of all, the arithmetic support is compatible with localization, hence

we can assume that we consider holonomic (X)-modules for a smooth affine
variety X. Let us choose a closed embedding j : X → ANk for some N ∈ Z≥0 .
Then any holonomic -module M gives a holonomic -module j∗ M on ANk .
The behavior of arithmetic supports for embeddings is the natural one, given by
the Lagrangian correspondence in T ∗ X × T ∗ ANk equal to the conormal bundle to
graph( j).
Now, we describe a way to put a structure of a reduced ind-scheme on the
ind-constructible set ELC(A2N k ), similar to the Chow scheme (see [1]) param-
etrizing algebraic cycles of given dimension and degree in a smooth projective
∗ N
scheme. Here A2N k = T Ak is considered as a symplectic manifold.
Let us consider the variety V /k parametrizing triples (B, b1 , b2 ) where B ⊂
Ak is a coisotropic affine subspace in A2N
2N
k , of dimension N + 1, and b1 , b2 are
two points in the symplectic affine plane B  A2 which is obtained by factor-
ization of B along the kernel of the natural Poisson structure on B. We claim
that every effective Lagrangian cycle C = ∑i∈I mi Li gives a non-zero rational
function φC on V .
Indeed, for generic B the intersection of all subvarieties Li with B is one-
dimensional and transversal at the generic point of every component. Moreover,
its projection to B is a plane curve. Hence, taking the sum over i we obtain an
effective divisor CB on plane B (a collection of curves with positive multiplic-
ities). There exist a unique up to scalar non-zero polynomial FB on B whose
divisor of zeroes is CB . Then, for generic b1 , b2 ∈ B the ratio FB (b1 )/FB (b2 ) is
canonically defined and is not zero. We set

φC (B, b1 , b2 ) := FB (b1 )/FB (b2 )

for generic (B, b1 , b2 ).

Let R/k be a finitely generated algebra without nilpotents. We define a fam-
ily over Spec R of effective Lagrangian cycles on A2N k to be a pair (U, φ ) where
U ⊂ Spec R ×Spec k V is an Zariski open subset which is dominant over Spec R,
and φ ∈  (U), φ = 0 is a function on U such that for every point x ∈ Spec R
the restriction of φ to the fiber over x coincides with the restriction of a rational
function associated with an effective Lagrangian cycle on A2N kx . We identify two

pairs (U, φ ) and (U , φ ) if and only if

φ|U∩U = φ|U∩U .

Thus, we have defined ELC(A2N k ) as a set-valued functor on finitely gener-

ated rings without nilpotents. One can check that this gives a structure of a re-
duced ind-scheme. The above definition (at least of a functor) also work without
the assumption that the ground field k has zero characteristic, and it generalizes
immediately to symplectic manifolds over arbitrary base.
Holonomic -modules and positive characteristic 11

Similar constructions can be performed for holonomic (Ank )-modules. The

analog of intersection with B and projection to B is given by the functor from
holonomic (Ank )-modules to holonomic (A1k )-modules determined by the
kernel corresponding to an affine Lagrangian subspace in T ∗ (Ank × A1k ). One
expects that the arithmetic supports for such functors behave as is prescribed
by the geometric construction from above. Hence, one get a reduction of the
problem of the description of the arithmetic support to the case X = A1k .
Moreover, it seems that there should exist an enhanced definition of the arith-
metic support of a holonomic X -module M as a k∞ -point of the ind-scheme
Fr∗k∞ ELC(T ∗ X).

3.2. Arithmetic support of a cyclic module

Let us consider the case X = A1k , char(k) = 0. The algebra (X) is the first Weyl
algebra A1,k , we denote its generators by x̂ = x and ŷ = d/dx. We consider holo-
nomic (A1k )-module which is a non-trivial cyclic module (A1k )/ (A1k ) · P,
where
P = ∑ ai j xi (d/dx) j
i+ j≤N
is a non-zero differential operator on X. Let us fix a finitely generated subring
R ⊂ k containing all coefficients ai j . The center of A1,R p (recall R p := R⊗Z/pZ)
is the polynomial algebra R p [x̂ p , ŷ p ]. We extend it by adding central variables x̃, ỹ
satisfying
x̃ p = x̂ p , ỹ p = ŷ p .
The resulting extension of A1,R p is isomorphic to the matrix algebra
Mat(p × p, R p [x̃, ỹ]).
Indeed, this extension is the algebra over R p [x̃, ỹ] generated by two elements x̂, ŷ
satisfying the relations
[ŷ, x̂] = 1, x̂ p = x̃ p , ŷ p = ỹ p .
Shifted generators (x̂ − x̃, ŷ − ỹ) satisfy the same relations as the operators x and
d/dx in the truncated polynomial ring Z/pZ [x]/(x p ):
x p = 0, (d/dx) p = 0, [d/dx, x] = 1.
For example, for p = 5 the corresponding matrices are
⎛ ⎞ ⎛ ⎞
00000 01000
⎜1 0 0 0 0⎟ ⎜0 0 2 0 0⎟
⎜ ⎟ ⎜ ⎟
X p = ⎜ 0 1 0 0 0 ⎟ , Yp = ⎜
⎜ ⎟ ⎟
⎜0 0 0 3 0 ⎟.
⎝0 0 1 0 0⎠ ⎝0 0 0 0 4⎠
00010 00000
12 M. Kontsevich

Then one get the following description of the arithmetic support of M =

(A1k )/ (A1k ) · P. Namely, for a given prime p let us consider the polynomial

 (p) := det ∑ ai j (X p + x̃ · 1p )i · (Yp + ỹ · 1p ) j ∈ R p [x̃, ỹ]
D P
i+ j≤N

where 1p is the identity matrix of size p × p. Obviously D  (p) has degree ≤ N · p

P
in x̃, ỹ. We claim that it is in fact a polynomial of degree ≤ N in x̃ p , ỹ p . This can
be seen by general reasons, and as well by a direct check. Namely, the property
of a polynomial in characteristic p to depend only on p-th powers of variables
is equivalent to the vanishing of its partial derivatives:
∂  (p) ∂  (p)
DP = D = 0.
∂ x̃ ∂ ỹ P
The vanishing of, say, the derivative with respect to x̃ can be proved as follows.
Taking this derivative is equivalent to taking the derivative of the determinant
of the matrix from above under the infinitesimal conjugation by 1p + ε Yp where
ε is a small parameter, ε 2 = 0. The invariance of the determinants under the
conjugation proves the result.
Therefore, we can write
 (p) (x̃, ỹ) = D(p) (x̃ p , ỹ p )
D P P
(p)
where DP is a polynomial in two variables of degree ≤ N:
(p)
DP ∈ R p [x̃ p , ỹ p ] = R p [x̂ p , ŷ p ] = Center(A1,R p ).
Proposition 2. The arithmetic support of module M at pair (p, v) where p is a
prime and v ∈ Spec R p , is the effective one cycle on plane A2kv = Fr∗kv A2kv given
(p)
as the divisor of zeroes of the image of polynomial DP in kv [x̃ p , ỹ p ].
This proposition follows directly from the definitions, and from the obvious
equivalence
Mat(p × p, k)/ Mat(p × p, k) · T = 0 ⇐⇒ det(T ) = 0
for any matrix T ∈ Mat(p × p, k) and any field k.
(p)
The free term of the polynomial DP is

det p (P) := det ∑ ai j X pi ·Ypj ∈ R p .
i+ j≤N

This expression we will call the p-determinant of a polynomial differential op-

erator in one variable. One can treat coefficients (ai j )i+ j≤N as independent vari-
ables, hence we have a universal p-determinant
det≤N
p ∈ Z/pZ[(ai j )i+ j≤N ]
Holonomic -modules and positive characteristic 13

which is a homogeneous polynomial of degree p in (N+1)(N+2) 2 variables with

coefficients in Z/pZ.
(p)
The calculation of other coefficients of DP can be reduced to the calcula-
tion of finitely many p-determinants of differential operators. Indeed, any poly-
nomial of a bounded degree can be reconstructed by the Lagrange interpolation
formula from its values at finitely many points.

3.3. Determinant formulas

One can calculate p-determinants effectively using a well-known formula. The

algorithm runs very fast, linearly in prime p. Notice that the matrix

Mp := ∑ ai j X pi ·Ypj
i+ j≤N

contains non-zero terms only at distance at most N from the main diagonal. We
are interested in its determinant for p  N.
Consider the general situation: let M be a square matrix of size L × L (with
coefficients in a commutative ring) such that Mi j = 0 if |i − j| > N for some
N < L/2. For every integer i ∈ [N + 1, L] denote A(i) the square matrix of size
2N × 2N given by
⎧
⎨Mi−N,i if j1 = j2 − 1
(i)
A j1 , j2 = −Mi−N,i−2N+ j2 −1 if j1 = 2N
⎩
0 otherwise

Here we set Mi j := 0 for j ≤ 0. Also introduce rectangular matrices

B ∈ Mat(N × 2N), B j1 , j2 = M j1 +L−N, j2 +L−2N , 1 ≤ j1 ≤ N, 1 ≤ j2 ≤ 2N,

B ∈ Mat(2N × N), B j1 , j2 = δ j1 , j2 −N , 1 ≤ j1 ≤ 2N, 1 ≤ j2 ≤ N.

Proposition 3. In the above notation one has

 L N−1
∏ Mi−N,i · det(M) = ± det(B · (A(L) A(L−1) · · · A(N+1) ) · B ).
i=N+1

The idea of the proof. Suppose that the matrix M is degenerate and all elements
Mi−N,i are non-zero for i ∈ [N + 1, L]. Hence the left hand side of the above
identity vanishes, and we want to prove that the right hand side vanishes, too.
Let us consider the sequence

(v j ) j=1,N+L := (0, . . . , 0, x1 , . . . , xL )
  
N times
14 M. Kontsevich

where (x1 , . . . , xL ) is a non-zero vector in the kernel of M. Then its subsequences

of length 2N

U ( j) := (v j , v j+1 , . . . , v j+2L−1 ), 1 ≤ j ≤ L−N +1

satisfy the relations

U (1) ∈ Im(B )
U (2) = M1,N+1
−1
· A(N+1) U (1)
U (3) = M2,N+2
−1
· A(N+2)U (2)
···
U (L−N+1) = ML−N,L
−1
· A(L) U (L−N)
0 = BU (L−N) .

Hence we conclude that

det(B · (A(L) A(L−1) · · · A(N+1) ) · B ) = 0.


The above proposition allows to calculate p-determinants up to a simple
factor which has the form
p−N
prime p → ∏ f ( j) (mod p),
j=1

where f = f (x) ∈ R[x] is as polynomial in one variable with coefficients in a

finitely generated ring R ⊂ k. Such factors vanish sometimes (e.g. when f has a
root in Q), in this case one modify it by replacing f by f + c where c is a new
independent constant.
We see that p-determinants (up to factors discussed above) belong to the
following class of expressions

prime p → Tr(F(1) · F(2) · · · · · F(p − k) · G) (mod p), if p ≥ k

where F ∈ Mat(K × K, R)[x], G ∈ Mat(K × K, R) for some K, k ∈ Z≥1 .

3.4. Logarithmic families of planar curves

Let k be an algebraically closed field of characteristic zero. The set of planar

curves CurvesA2 understood as effective divisors in A2k , is the same as the quo-
k
tient of the set of non-zero polynomials P ∈ k[x, y] \ {0} modulo multiplicative
Holonomic -modules and positive characteristic 15

constant. Hence it carries a natural structure of an ind-scheme, it is an infinite-

dimensional projective space
(d+1)(d+2)
CurvesA2 = P∞ (k) = P(k[x, y]) = lim P 2 −1
(k).
k −→
The group AutA2k acts by automorphisms of this ind-scheme.
Our goal here is to introduce an equivalence relation on CurvesA2 invariant
k
under Aut A2k such that all equivalence classes will be sets of k-points of con-
structible sets, and for any cyclic (A1k )-module M its arithmetic support will
belong to one such an equivalence class. These equivalence classes we will call
logarithmic families because they have a characterization in terms logarithmic
divergence of certain integrals.
First, we introduce certain set Jk∞ associated with A2k . It can be thought as
truncated jets of algebraic curves in P2k intersecting the projective line at infinity
P1k = P2k \ A2k . The truncation means that we do not specify the terms in Puiseux
series which change the germ of the curve by another germ such that the area
with respect to 2-form dx ∧ dy in a segment bounded by germs is finite. Here is
the precise definition for the special case of germs intersecting P1k at the infinite
point on the x-axis given by y = 0 in A2k = Spec k[x, y]. The germ is given by an
integer d ≥ 1 and a sequence of numbers ai ∈ k, −d < i < d such that

g.c.d.({d} ∪ {i | ai = 0}) = 1.

We identify such data factorizing by the free action of the group μd of roots 1
of order d
ξ ∈ k, ξ d = 1 acts as ai → ξ i ai .
The interpretation of the pair d, (ai )−d<i<d is as the truncated germ of a curve
d−1
x = x(t) = t d , y = y(t) = ∑ ait i + O(t −d ), t→∞
1−d

Alternatively, one can write y as a Puiseux series in x:

+ · · · + a1−d x−
d−1 d−1
y = ad−1 x d d + O(1/x).

Conversely, any Puiseux series

y∼ ∑ cλ xλ
λ ∈Q, λ <1

which is ∼ o(x) as x → ∞, has a unique representation as above (mod O(1/x)).

Namely, we define d as the minimal integer ≥ 1 such that λ ∈ (1/d) · Z for all
λ ∈ (−1, 1) such that cλ = 0, and set ai := ci/d .
16 M. Kontsevich

Acting by the group GL(2, k) we obtain the description of the whole set Jk∞ .
There is a Z≥1 -valued function deg on Jk∞ , with the value equal to d in the above
notation.
Any planar curve C ∈ CurvesA2 gives a function νC : Jk∞ → Z≥0 with finite
k
support. Namely, we count with multiplicities all the germs of C intersecting P1k
at infinity. The degree of C coincides with the sum over Jk∞ of the product of νC
with deg.
We define a logarithmic family to be the set of all curves C with a given
function νC .
Here are examples in small degrees. First, we have the logarithmic family
consisting of the empty curve C with νC = 0. Next, we have one-point loga-
rithmic families each of which consists of a line in A2k . The simplest non-trivial
example is the family of hyperbolas xy = t where t ∈ k, including the degenerate
case t = 0.
It is easy to see that there is a natural action of Aut A2k on Jk∞ , and the decom-
position by logarithmic families is Aut A2k -equivariant. An intrinsic definition of
Jk∞ is as the inductive limit of the set of divisors where the volume form dx ∧ dy
has pole of order one, over the partially ordered set of smooth compactifications
of A2 on which the volume form does not vanish at infinity (i.e., it is a Poisson
compactification, compare with [9]). For any curve C ∈ CurvesA2 there exists
k
a Poisson compactification of A2k such that C intersects only those divisors at
infinity where the form has logarithmic pole. Also, If C1 and C2 are two curves
such that νC1 · νC2 = 0 (i.e., functions νC1 and νC2 have disjoint support), then
the intersection C1 ∩ C2 is finite and the intersection number [C1 ] ∩ [C2 ] can be
determined entirely in terms of functions νC1 and νC2 .
For a non-algebraically closed field k of zero characteristic we define log-
arithmic families using the embedding k → k to the algebraic closure. Also,
logarithmic families for curves of a given bounded degree can be defined for
positive characteristic if it is large enough.
There is an alternative meaning of Jk∞ in terms of singularities of holonomic
(Ak )-modules. Let M be such a module. We will associate with M a Z≥0 -
valued function νM on Jk∞ with finite support.
First, there exists a finite set S ⊂ k such that M|A1 \S is a vector bundle  with
k
connection ∇. It can have irregular singularities at S and at ∞.
It is well-known (see e.g. [11]) that the category of bundles with connections
over the field of Laurent series k((z)) is decomposed into the direct sum of
blocks corresponding to Puiseux polynomials in negative powers of z:
F(z) = ∑ bλ zλ , bλ ∈ k, bλ = 0 for almost all λ ,
λ ∈Q<0

defined modulo the action of μd where d is l.c.m. of all denominators of λ with

bλ = 0. The basic -module in such a block is exp(F) · k((z)). These blocks for
Holonomic -modules and positive characteristic 17

F = 0 correspond exactly to elements in Jk∞ corresponding to truncated germs

intersecting the divisor at infinity at point (0, ∞) in the compactification
A2k = A1k × A1k ⊂ P1k × P1k ,
with the exception of the germ at infinity of the vertical line z = 0. The corre-
spondence is given by
F → germ of the curve (z, F (z)) at z → 0.
In this manner we will associate to any holonomic (A1k )-module M multi-
plicities νM at all points of Jk∞ except truncated germs of lines x = x0 for x0 ∈ S.
In order to get multiplicities at these points one can apply an automorphism of
the Weyl algebra A1,k corresponding to a non-trivial matrix in SL(2, k), e.g. the
Fourier transform. We notice also that the complicated formulas from [11] relat-
ing irregular singularities ofM with  the ones of its Fourier transform, translate
0 1
just to the action of matrix on Jk∞ .
−1 0
A holonomic (A1k )-module M has only regular singularities if and only if
the multiplicities νM vanish at all points of Jk∞ except the germs of lines given
by equations
y = 0, or x = x0 for some x0 ∈ k.

Theorem 1. For any non-zero differential operator P ∈ A1,k = (A1k ) with coef-
ficients in a finitely generated ring R ⊂ k, char(k) = 0, for all sufficiently large
p and for any point v ∈ Spec R over p, the multiplicities νM at point v where
M = A1,k /A1,k · P coincides with the (pullback by Frobenius of) multiplicities of
(p)
the curve given by the equation DP = 0 at point v.

The idea of the proof. First of all, it is easy to identify contributions of germs
of the line y = 0. Namely, in the case of a curve given by equation H(x, y) =
0, H = ∑i, j Hi j xi y j = 0 ∈ k[x, y], this multiplicity can be read from the Newton
polygon of P. Namely, the multiplicity is equal to
max{ j | Hi j = 0, ∀(i , j ) Hi j = 0 =⇒ (i − j) ≥ (i − j )}
A similar description works for cyclic A1 -modules. For the multiplicities at
k
other points of Jk∞ one can apply automorphisms of the Weyl algebra, and also
take the tensor product with -modules corresponding to exponents of Puiseux
polynomials. 
Finally, one can show that for two holonomic A1 -modules M1 , M2 such that
k
supports of νM1 and νM2 are disjoint, there is no non-trivial homomorphisms
from M1 to M2 , and the dimension of Ext1 (M1 , M2 ) coincides with the intersec-
tion number of the corresponding curves.
18 M. Kontsevich

4. Correspondence between classical and quantum families

4.1. Rough picture for Lagrangian cycles

We expect that in the case dim X > 1 also there exists a notion of a logarithmic
family of effective Lagrangian cycles in T ∗ X, and the arithmetic support should
always belong to such a family. In the special case when a Lagrangian cycle is
a smooth closed Lagrangian variety L ⊂ T ∗ X (taken with multiplicity one) we
expect a more clearer picture of what is the logarithmic family:
Definition 3. A smooth logarithmic family of smooth Lagrangian subvarieties
in T ∗ X is a pair (S,  ) where S is a smooth variety over k and  ⊂ T ∗ X × S
is a smooth closed submanifold such that its projection to S is smooth, all fibers
s, s ∈ S are Lagrangian, and the following property holds. For any s ∈ S the
natural map
Ts S → Γ(s , (TX )|s /Ts ) = Γ(s , T∗ s )
identifies Ts S with the space of 1-forms on s with logarithmic singularities4 .

Conjecture 3. For a smooth closed Lagrangian L ⊂ T ∗ X there exists a smooth

logarithmic family (S,  ) with base point s0 ∈ S such that s0 = L. Also, any
two such families coincide with each other in the vicinity of s0 .

4.2. Isosingular families of holonomic -modules

By analogy with the geometry of logarithmic families, we expect that something

similar should happen for holonomic X -modules as well. We say (to a first
approximation) that two holonomic modules are isosingular if and only if the
corresponding arithmetic supports belong to the same logarithmic family. In
the case X = A1 the precise definition is the coincidence of multiplicities, as
explained in section 3.4.
Namely, for any holonomic module M we expect that there exists a natural
moduli stack ModM parametrizing holonomic X -modules which looks locally
like the quotient of a scheme of finite type by a group whose connected compo-
nent of identity is a finite-dimensional affine algebraic group. The tangent com-
plex of ModM at the base point corresponding to M should have cohomology
Hom(M, M) in degree −1 and Ext1 (M, M) in degree 0. Connected components
of stacks ModM we will call isosingular families of holonomic X -modules.
One of motivations is that the abelian category of holonomic X -modules
has the following finiteness property: for any two objects M1 , M2 we have
dim Hom(M1 , M2 ) < ∞, dim Ext1 (M1 , M2 ) < ∞.
4 All such forms are automatically closed.
Holonomic -modules and positive characteristic 19

The same property is shared by the category of finite-dimensional modules over

a finitely generated associative algebra A/k. In the latter case we have naturally
defined moduli stacks of objects.
Also, in the case X = A1k any holonomic X -module M belongs to the
abelian category consisting of all holonomic modules M such that

supp νM ⊂ supp νM .

In the case k = C using the Riemann–Hilbert correspondence (for irregular

singularities) we can identify the above abelian category with the category of
finite-dimensional representations of a finitely generated algebra. Hence, we get
moduli stacks of the form described above, but with a “wrong” algebraic (but
“correct” complex analytic) structure on the moduli stack.
If the arithmetic support of a simple holonomic X -module M is a family
of effective Lagrangian cycles such that the generic representative L of such
a family is a smooth connected non-empty closed subvariety with multiplicity
one, then we expect that

dim Hom(M, M) = dim H 0 (L) = 1, dim Ext1 (M, M) = dim HdR

1
(L).

We see that the dimension of an isosingular family is greater than the dimension
of the corresponding logarithmic family. The difference between two dimen-
sions is the dimension of the Picard variety of any smooth compactification L of
L.
Informally speaking, holonomic X -modules correspond to Lagrangian sub-
manifolds in T ∗ X together with a line bundle on L (or something like that5 ).
More precisely, there should be the decomposition of the space of holonomic
modules into isosingular families, and the decomposition of Lagrangian sub-
manifolds with line bundles into logarithmic families, such that there is a canon-
ical one-to-one correspondence between families (equivalence classes) of both
kinds, and dimensions of the corresponding families coincide.

4.3. Constants for the arithmetic support

Let k be a field of characteristic zero, and consider the subring Pk ⊂ k∞ gener-

ated by p-determinants of differential operators in one variable, with coefficients
in k. Considerations from sections 3.1 and 3.2 lead to the following question.
What is the structure of Pk ?
5 In the picture with reduction modulo prime we get not only a Lagrangian submanifold but
also a module over an Azumaya algebra, of minimal rank. Two such modules differ by a line
bundle.
20 M. Kontsevich

The importance of the ring Pk is that the arithmetic supports should be effective
Lagrangian cycles parametrized by Spec Pk . For countable k (e.g. for k = Q or
k = Q) the ring Pk is a countable subring in uncountable ring k∞ , similar to the
subalgebra of periods in C for algebraic varieties over Q.
The result from section 3.3 indicates that Pk should be related with the dif-
ferential Galois group of difference equations with coefficients in k.
Also, there are strong indications that Pk should be related somehow to an-
other group, the motivic Galois group of k. Indeed, for holonomic X -module
corresponding to the flat connection on the trivial bundle given by a closed but
not exact 1-form, the support is controlled by the Cartier operator, which is re-
lated to the the comparison of de Rham and crystalline cohomology of X (in
degree 1). Roughly speaking, one can expect that the motivic Galois group of k
in de Rham realization acts on the category of holonomic X -modules. The cor-
respondence between holonomic modules and their arithmetic support could be
related to torsors comparing de Rham and crystalline cohomology (and maybe
Hodge realization which is the associated graded with respect to the Hodge
filtration on de Rham cohomology). Fixed points of the motivic Galois group
should correspond to motivic (or, more generally, motivic-exponential) holo-
nomic X -modules. This fits well with the fact that supports of motivic and
exponential X -modules are defined over Fr∗k∞ (k) ⊂ k∞ .
For a given field k, char(k) = 0 we define extended motivic-exponential X -
modules on smooth algebraic varieties over k as the minimal class which is
closed under extensions, subquotients, pushforwards and pullbacks, and contain
all X -modules of type exp(F) · X for F ∈  (X).

Conjecture 4. The arithmetic support of a holonomic X -module M is the pull-

back by the universal Frobenius Frk∞ of an effective Lagrangian cycle in T ∗ X
defined over k ⊂ k∞ if and only if M is extended motivic-exponential.

4.4. Isolated points

One can ask what are the “most canonical” holonomic modules and correspond-
ing Lagrangian varieties6 .

Conjecture 5. For any smooth closed connected Lagrangian subvariety L in T ∗ X

over k = C such that H 1 (L(C), Z) = 0 there exists a unique holonomic X -
module M = ML with the arithmetic support equal to L taken with multiplicity
1. Moreover, Ext1 (M, M) = 0.

The reason for the condition H 1 (L(C), Z) = 0 is that it guarantees that there is
no non-trivial local system of rank 1 over L. This condition can be reformulated
6 In a recent preprint [2] a related but different question was studied for X = A1k .
Holonomic -modules and positive characteristic 21

in form which makes sense for arbitrary field k with char(k) = 0:

Hét1 (L ×Spec k Spec k, Z/lZ) = 0
for any prime l. In this case we will write simply H 1 (L) = 0.
Presumably, one can weaken the condition on smoothness of L, e.g. it is
definitely sufficient to assume that the codimension of singularities is ≥ 3.

Conjecture 6. Any holonomic X -module M with Ext1 (M, M) = 0 is of the ex-

tended motivic-exponential type.

A corollary of conjecture 5 is one of conjectures discussed in [3], which says

that for any polynomial symplectomorphism φ of A2n k there exists a canonically
associated to it a bimodule Mφ over the Weyl algebra An,k giving Morita self-
equivalence of the category of An,k -modules. Indeed, the graph of φ is a smooth
Lagrangian subvariety in A4n k . Moreover, this subvariety is simply connected
being isomorphic to Ak . Hence it should give a canonical holonomic module
2n

Mφ over A2n,k which can be interpreted as a bimodule over An,k .

Also, conjecture 6 implies that Mφ is of the extended motivic-exponential
type.

5. Relation with integrable systems

5.1. Arithmetic support of a non-holonomic module

Let us consider a typical non-holonomic module M which is just a cyclic mod-

ule (X)/ (X) · P for a non-zero differential operator P ∈ (X). Here X/k
is a smooth affine variety, of dimension n > 1. Then the support at prime p is
a hypersurface of degree which is bounded by const · pn−1 . This can be seen
most easily in the case X = Ank . The consideration similar to one from section
3.2 shows that after the adding formally p-th roots (x̃i )i=1,...,2n of central ele-
ments (x̂ip )i=1,...,2n , we identify (the pullback of) the n-th Weyl algebra with the
algebra of matrices of size (pn × pn ). Operator P gives a matrix MP,p with coef-
ficients being polynomials of a bounded degree independent on p in generators
(x̃i )i=1,...,2n . The support at prime p is the hypersurface given by the equation
det(MP,p ) = 0
and hence has degree const · pn in (x̃i )i=1,...,2n . This polynomial has zero deriva-
tive with respect to each variable, and therefore is in fact a polynomial in (x̃ip =
x̂ip )i=1,...,2n , of degree const · pn−1 .
The computer experiments leave no doubt that for “typical” P this polyno-
mial has indeed such a large degree, and also is indecomposable. Hence we can
not make a reasonable algebraic limit as p → ∞.
22 M. Kontsevich

Nevertheless, for certain operators P the resulting polynomial is pn−1 -st

power of a polynomial whose degree is uniformly bounded in p. This happens,
for example, for the Hamiltonian of the periodic Toda lattice
 y ∂ 2 yi+1
∑
i
+
i∈Z/nZ
∂ yi yi

which after the transcendental change of variables xi = log yi has a more familiar
form
 ∂ 2
∑ ∂ xi + exp(xi+1 − xi ) .
i∈Z/nZ

In general, we conjecture that such a situation is related to integrable sys-

tems:
Conjecture 7. For a differential operator P ∈ (X), the support at prime p of
(X)/ (X)·(P+ λ ·1) is divisible by pn−1 for generic constant λ if and only if
P belongs to a quantum integrable system, i.e., P belongs to a finitely generated
commutative k-subalgebra of (X) of Krull dimension n = dim X.
The modification of P by a generic additive constant is necessary in order to
exclude certain parasitic examples, e.g. P = P1 P2 where P1 and P2 belong to two
different integrable systems.

5.2. Donagi–Markman construction

In [6] a construction of integrable systems was proposed starting from any

smooth projective variety X. One considers the scheme B parametrizing smooth
projective Lagrangian submanifolds L ⊂ T ∗ X of a bounded degree. Scheme B is
smooth of dimension equal to dim Γ(L, TL∗ ). Now consider the bundle M over B
with the fiber over [L] equal to the Albanese variety of B (i.e., the abelian variety
dual to Pic0 (B)). Then M carries a natural symplectic structure, and fibers over
B are Lagrangian abelian varieties. Hence, we obtain an algebraic completely
integrable system. In the special case dim X = 1 one get a Zariski open part in
the Hitchin integrable systems for the group GL(N) for N := deg(L → X).
Smooth logarithmic families of non-compact Lagrangian submanifolds pro-
vide a natural generalization of Donagi–Markman construction. The Albanese
variety one can replace by an appropriate logarithmic version.
In [10], section 2, one can find a new notion of an algebraic integrable system
which includes simultaneously both the classical and the quantized cases. This
notion makes sense also over finite and local fields. The program in [10] was
proposed relating integrable systems and Langlands correspondence in the func-
tional field case. Roughly speaking, the spectrum of the maximal commutative
Holonomic -modules and positive characteristic 23

subalgebra in the quantized system should parametrize in certain sense holo-

nomic X -modules. Although the subject is not yet fully developed, it looks
certain that the integrable systems which should appear in the context of this
program are the same as generalized Donagi–Markman systems.
Finally, we should mention that logarithmic families of planar curves appear
naturally in matrix models. For example, one of the standard families in matrix
theory (see [5]) is the family of hyperelliptic curves
n−1
2
y = (W (x)) + f (x),
2
f (x) = ∑ ci xi
i=0

where W (x) is a fixed polynomial of degree (n + 1) and (ci )i≤n−1 are parameters
of the curve. The full solution of the matrix model is based on the associated
family of holonomic A1 -modules. Also, Seiberg–Witten curves for N = 2 pure
k
SU(2) gauge theory, given by the equation
x + 1/x = y2 + u
where u is a parameter, form a logarithmic family of curves in T ∗ Gm endowed
with coordinates x, y (x = 0) and the symplectic form dx ∧ dy/x.

6. Trigonometric version

6.1. Quantum tori

The multiplicative (trigonometric) analog of the Weyl algebra An is the algebra

of functions on quantum torus:
ωi j
Kx̂±1 ±1
1 , . . . , x̂2n /(x̂i · x̂ j = q x̂ j · x̂i )

where ω = (ωi j ) is the standard skew-symmetric matrix. The ground field is

K := k(q),
the field of rational functions in variable q, with coefficients in an algebraically
closed field k = k, char(k) = 0. There exists a theory of holonomic modules
over quantum tori, see [13]. Examples of such modules are non-trivial cyclic
modules in the case n = 1, and also bimodules corresponding to automorphisms
of skew-fields of fractions. An analog of a prime here is a primitive root of 1.
Namely, for any N ≥ 1 and qN ∈ k a primitive N-th root of 1, the algebra
ω
kx̂±1 ±1
1 , . . . , x̂2n /(x̂i · x̂ j = qN x̂ j · x̂i )
ij

has a large center equal to

k[(x̂N1 )±1 , . . . , (x̂N2n )±1 ]
24 M. Kontsevich

and it is an Azumaya algebra over its center, a twisted form of the algebra of
matrices of size N n × N n .
In the case n = 1 any element
P= ∑ ai j x̂i1 x̂2j
i, j∈Z, |i|+| j|≤d

gives a sequence of N × N-matrices for qN ∈ k a primitive N-th root of 1 and

N large enough. Their determinants can be calculated similarly to section 3.3.
The resulting algebra of “periods” is close to the algebra of expressions which
appear e.g. as quantum Chern–Simons invariants of 3-dimensional manifolds.
Here is an analog for quantum tori of the main conjecture from [3].
Conjecture 8. There exists a homomorphism from the group BirSympln,k of bi-
rational symplectomorphisms the algebraic torus G2n m,k endowed with the stan-
−1 −1
dard symplectic form ∑i, j≤2n ωi j (xi dxi ) ∧ (x j dx j ), to the group of outer au-
tomorphisms of the skew field of fractions of the quantum torus. Also, the semi-
classical limit as q → 1 exists and gives the identity map from the group of bi-
rational symplectomorphisms the group of birational symplectomorphisms the
algebraic torus to itself.
Let us assume the conjecture. Then, taking the reduction at a root of one
q = qN , we obtain an outer birational automorphism of the algebra with a large
center, hence an automorphism of the center. In this way one should obtain some
mysterious “quantum Frobenius” endomorphism
FrN : BirSympln,k → BirSympln,k .
There is a class of birational transformations analogous to tame transformations
(see [3]) in the additive case. It is generated by Sp(2n, Z), by multiplicative
translations
M(λi )i≤2n : (xi )1≤i≤2n → (λi xi )1≤i≤2n
for some constants λi ∈ k∗ , 1 ≤ i ≤ 2n, and by the following non-trivial auto-
morphism:
T : x1 → x1 (1 − xn+1 ), xi → xi for i ≥ 2.
The miracle is that for the corresponding quantum automorphism
x̂1 → x̂1 (1 − x̂n+1 ), x̂i → x̂i for i ≥ 2
the induced transformation on the center for q = qN , is given by the same for-
mula:
x̂N1 → x̂N1 (1 − x̂Nn+1 ), x̂Ni → x̂Ni for i ≥ 2.
The quantum Frobenius FrN acts identically on Sp(2n, Z) and on transfor-
mation T , but on multiplicative translations it acts non-trivially
FrN : M(λi )i≤2n → M(λiN )i≤2n .
Holonomic -modules and positive characteristic 25

We see that graphs of elements of the group Γn generated by Sp(2n, Z) and T

are multiplicative analogs of isolated points (see section 4.4), i.e., could be con-
sidered as the “most canonical” Lagrangian subvarieties (and the corresponding
holonomic modules). The group Γn deserves further study, it contains both all
arithmetic groups and mapping class groups for large n, as follows e.g. from the
work of Goncharov and Fock on cluster transformations and generalizations of
Penner coordinates, see [7]. Finally, the above discussion has a generalization
to elliptic algebras.

References

[1] B. Angéniol, Familles de Cycles Algébriques—Schéma de Chow, Lecture Notes in Math.,

896, Springer-Verlag, 1981.
[2] D. Arinkin, Rigid irregular connections on P1 , e-print, arXiv:0808.0742.
[3] A. Belov-Kanel and M. Kontsevich, Automorphisms of the Weyl algebra, Lett. Math. Phys.,
74 (2005), 181–199.
[4] A. Borel, P.-P. Grivel, B. Kaup, A. Haefliger, B. Malgrange and F. Ehlers, Algebraic D-
modules, Perspect. Math., 2, Academic Press, Boston, MA, 1987.
[5] R. Dijkgraaf and C. Vafa, Matrix models, topological strings, and supersymmetric gauge
theories, Nuclear Phys. B, 644 (2002), 3–20.
[6] R. Donagi and E. Markman, Spectral covers, algebraically completely integrable, Hamilto-
nian systems, and moduli of bundles, In: Integrable Systems and Quantum Groups, Monte-
catini Terme, 1993, Lecture Notes in Math., 1620, Springer-Verlag, Berlin, 1996, pp. 1–119.
[7] V. Fock and A. Goncharov, Moduli spaces of local systems and higher Teichmüller theory,
Publ. Math. Inst. Hautes Études Sci., 103 (2006), 1–211.
[8] N. Katz, Nilpotent connections and the monodromy theorem: Applications of a result of
Turrittin, Inst. Hautes Études Sci. Publ. Math., 39 (1970), 175–232.
[9] M. Kontsevich, Deformation quantization of algebraic varieties, Lett. Math. Phys., 56
(2001), 271–294.
[10] M. Kontsevich, Notes on motives in finite characteristic, e-print, arXiv:math/0702206.
[11] B. Malgrange, Équations Différentielles à Coefficients Polynomiaux, Progr. Math., 96,
Birkhäuser Boston, Boston, MA, 1991.
[12] A. Ogus and V. Vologodsky, Nonabelian Hodge theory in characteristic p, Publ. Math. Inst.
Hautes Études Sci., 106 (2007), 1–138.
[13] C. Sabbah, Systèmes holonomes d’équations aux q-différences, In: D-modules and Mi-
crolocal Geometry, Lisbon, 1990, de Gruyter, Berlin, 1993, pp. 125–147.
[14] Y. Tsuchimoto, Preliminaries on Dixmier conjecture, Mem. Fac. Sci. Kochi Univ. Ser. A
Math., 24 (2003), 43–59.