Anda di halaman 1dari 25

a

r
X
i
v
:
1
3
0
4
.
0
6
6
7
v
2


[
m
a
t
h
.
R
T
]


1
5

A
p
r

2
0
1
3
CLASSIFYING -TILTING MODULES OVER PREPROJECTIVE
ALGEBRAS OF DYNKIN TYPE
YUYA MIZUNO
Abstract. We study support -tilting modules over preprojective algebras of Dynkin
type. We classify basic support -tilting modules by giving a bijection with elements
in the corresponding Weyl groups. Moreover we show that they are in bijection with
the set of torsion classes, the set of torsion-free classes and other important objects in
representation theory of preprojective algebras. We also study g-matrices of support -
tilting modules, which show terms of minimal projective presentations of indecomposable
direct summands. We give an explicit description of g-matrices and prove that cones
given by g-matrices coincide with chambers of the associated root systems.
Contents
Introduction 1
1. Preliminaries 4
1.1. Preprojective algebras 4
1.2. Support -tilting modules 4
1.3. Torsion classes 6
2. Support -tilting ideals and the Weyl group 7
2.1. Support -tilting ideals 7
2.2. Bijection between support -tilting modules and the Weyl group 12
2.3. Partial orders of support -tilting modules and the Weyl group 15
3. g-matrices and cones 17
4. Further connections 22
References 24
Introduction
Preprojective algebras rst appeared in the work of Gelfand-Ponomarev [GP]. Since
then, they have been one of the important objects in the representation theory of algebras.
Preprojective algebras appear also in many branches of mathematics. One of the notable
examples is the study of quantum groups, where preprojective algebras play central roles in
Lusztigs Lagrangian construction of semicanonical basis [L1, L2] and subsequent Geiss-
Leclerc-Schr oers construction of cluster monomials of certain types of cluster algebras
[GLS1, GLS3]. In this paper we study preprojective algebras from a new perspective.
In [BIRS] (see also [IR]), the authors studied preprojective algebras of non-Dynkin
quivers via tilting theory. As an application, they succeeded to construct a large class of
2-Calabi-Yau categories which have close connections with cluster algebras. On the other
hand, in Dynkin cases (i.e. the underlying graph of a quiver is A
n
(n 1), D
n
(n 4) and
The author is supported by Grant-in-Aid for JSPS Fellowships No.23.5593.
1
2 YUYA MIZUNO
E
n
(n = 6, 7, 8)), the preprojective algebras are selnjective, so that all tilting modules are
trivial (i.e. projective). In this paper, instead of tilting modules, we will study support
-tilting modules (Denition 1.3) over the preprojective algebras.
The notion of support -tilting modules was introduced in [AIR], as a generalization
of tilting modules. It is shown that there are deep connections between support -tilting
modules, torsion classes, silting complexes and cluster tilting objects. Support -tilting
modules also have nicer mutation properties than tilting modules (Denition-Theorem
1.5). Moreover, certain support -tilting modules over selnjective algebras provide tilt-
ing complexes [M]. It is therefore fruitful to investigate these remarkable modules for
preprojective algebras and they are also interesting in their own right. To explain our
results, we give the following set-up.
Let Q be a nite connected acyclic quiver with the set Q
0
= 1, . . . , n of vertices,
the preprojective algebra of Q and I
i
the two-sided ideal of generated by 1 e
i
, where
e
i
is an idempotent of corresponding to i Q
0
. If Q is non-Dynkin, then I
i
is a tilting
module. These ideals I
i
are quite useful as shown in [IR, BIRS, AIRT]. On the other
hand, if Q is Dynkin, then I
i
is never a tilting module. Instead, we show that I
i
is a
support -tilting module (Lemma 2.3) and this observation is crucial in this paper. We
study ideals of of the form I
i
1
I
i
2
I
i
k
for some k 0 and i
1
, . . . , i
k
Q
0
, and denote
by I
1
, . . . , I
n
the set of such ideals.
Our main goal is to give the following bijections which provide a complete description
of support -tilting modules over preprojective algebras of Dynkin type.
Theorem 0.1 (Theorem 2.21). Let Q be a Dynkin quiver with the set Q
0
= 1, . . . , n of
vertices and the preprojective algebra of Q. There are bijections between the following
objects.
(a) The Weyl group W
Q
associated with Q.
(b) The set I
1
, . . . , I
n
.
(c) The set s-tilt of isomorphism classes of basic support -tilting -modules.
(d) The set s-tilt
op
of isomorphism classes of basic support -tilting
op
-modules.
It is known that support -tilting modules have a natural structure of partially ordered
set (Denition 1.4). On the other hand, it is well-known that the Weyl group also has a
structure of partial orders. We relate partial orders of these two objects as follows.
Theorem 0.2 (Theorem 2.30). The bijection W
Q
s-tilt in Theorem 0.1 gives an
isomorphism of partially ordered sets (W
Q
,
L
) and (s-tilt, )
op
, where
L
is the left
order on W
Q
(Denition 2.25).
We also study the g-matrix (Denition 3.1) of support -tilting module T, which shows
terms of minimal projective presentations of each indecomposable direct summand of
T. By g-matrices, we dene cones, which admit geometric realizations of the simplicial
complex of support -tilting modules. We give a complete description of g-matrices in
terms of reection transformations and show that their cones coincide with chambers of
the associated root system. The results are summarized as follows, where e
1
, . . . , e
n
is the
canonical basis of Z
n
, C
0
:= a
1
e
1
+ +a
n
e
n
[ a
i
R
>0
and

: W
Q
GL
n
(Z) is the
contragradient of the geometric representation (Denition 3.5).
Theorem 0.3 (Theorem 3.9). Let Q be a Dynkin quiver with the set Q
0
= 1, . . . , n of
vertices and be the preprojective algebra of Q.
(a) The set of g-matrices of support -tilting -modules coincides with

(W
Q
).
CLASSIFYING -TILTING MODULES OVER PREPROJECTIVE ALGEBRAS OF DYNKIN TYPE 3
(b) For any w W
Q
, we have
C(w) =

w
(C
0
),
where C(w) denotes the cone of the support -tilting module I
w
(Denition 3.1)
and

w
:=

(w). In particular, cones of basic support -tilting -modules give


chambers of the associated root system.
The above results give several important applications. Among others we conclude that
there are only nitely many basic support -tilting -modules. This fact yields bijections
between basic support -tilting -modules, torsion classes in mod and torsion-free classes
in mod by a result in [IJ]. This seems to be interesting by itself since almost all prepro-
jective algebras are of wild representation type. Then we can give complete descriptions
of these two classes in terms of W
Q
(Proposition 4.2).
As another application by combining with results in [AIR], [KY, BY] and [ORT], we
extend bijections of Theorem 0.1 as follows.
Corollary 0.4 (Corollary 4.1). Objects of Theorem 0.1 are in bijection with the following
objects.
(e) The set of torsion classes in mod.
(f) The set of torsion-free classes in mod.
(g) The set of isomorphism classes of basic two-term silting complexes in K
b
(proj).
(h) The set of intermediate bounded co-t-structures in K
b
(proj) with respect to the
standard co-t-structure.
(i) The set of intermediate bounded t-structures in D
b
(mod) with length heart with
respect to the standard t-structure.
(j) The set of isomorphism classes of two-term simple-minded collections in D
b
(mod).
(k) The set of quotient closed subcategories in modKQ.
(l) The set of subclosed subcategories in modKQ.
We now describe the organization of this paper.
In section 1, we recall denitions of preprojective algebras, support -tilting modules
and torsion classes.
In section 2.1, we introduce support -tilting ideals (Denition 2.1), which play key
roles in this paper. Moreover, we prove that any element of I
1
, . . . , I
n
is a basic support
-tilting ideal (Theorem 2.2). In section 2.2, we prove that W
Q
correspond bijectively with
support -tilting modules (Theorem 2.21). We give a description of mutations of support
-tilting modules in terms of W
Q
. We also determine the annihilators of all support -
tilting modules. In section 2.3, we compare partial orders of s-tilt and W
Q
. We give a
close connection between a partial order of W
Q
and that of s-tilt (Theorem 2.30).
In section 3, we deal with g-matrices of support -tilting modules. We give an explicit
description of them in terms of linear transformations. Furthermore, we study cones of
support -tilting modules and show that they coincide with chambers of the associated
root system.
In section 4, we show Corollary 0.4. In particular, we give complete descriptions of
torsion classes and torsion-free classes (Proposition 4.2). Furthermore, we explain a rela-
tionship between our results and other works.
Notations. Let K be an algebraically closed eld and we denote by D := Hom
K
(, K).
By a nite dimensional algebra , we mean a basic nite dimensional algebra over K. By
a module, we mean a right module unless stated otherwise. We denote by mod the
4 YUYA MIZUNO
category of nitely generated -modules and by proj the category of nitely generated
projective -modules. The composition gf means rst f, then g. We denote by Q
0
the
set of vertices and by Q
1
the set of arrows of a quiver Q. For a quiver a Q
1
, we denote
by s(a) and e(a) the start and end vertices of a respectively. For X mod, we denote
by SubX (respectively, FacX) the subcategory of mod whose objects are submodules
(respectively, factor modules) of nite direct sums of copies of X. We denote by addM
the subcategory of mod consisting of direct summands of nite direct sums of copies of
M mod.
Acknowledgement. First and foremost, the author would like to thank Osamu Iyama
for his support and patient guidance. He would like to express his gratitude to Steen
Oppermann, who kindly explain results of his paper. He is grateful to Joseph Grant,
Laurent Demonet and Dong Yang for answering questions and helpful comments. He
thanks Kota Yamaura, Takahide Adachi and Gustavo Jasso for their help and stimulating
discussions.
1. Preliminaries
1.1. Preprojective algebras. In this subsection, we recall denitions and some proper-
ties of preprojective algebras. We refer to [BGL, DR1, DR2, Ri] for basic properties and
background information.
Denition 1.1. Let Q be a nite connected acyclic quiver with vertices Q
0
= 1, . . . , n.
The preprojective algebra associated to Q is the algebra
= KQ/

aQ
1
(aa

a)
where Q is the double quiver of Q, which is obtained from Q by adding for each arrow
a : i j in Q
1
an arrow a

: i j pointing in the opposite direction.


If Q is a non-Dynkin quiver, the bounded derived category D
b
(fd ) of the category
fd of nite dimensional modules are 2-Calabi-Yau (2-CY for short), that is, there is a
functorial isomorphism
DHom
D
b
(fd )
(X, Y )

= Hom
D
b
(fd )
(Y, X[2]).
We refer to [AR1, Boc, CB] and see [GLS2] for the detailed proofs. On the other hand,
the algebra is nite-dimensional selnjective if Q is a Dynkin quiver.
In this paper, we use the two-sided ideal I
i
of generated by 1 e
i
, where e
i
is a
primitive idempotent of for i Q
0
. It is easy to see that the ideal has the following
property, where we denote by S
i
:= /I
i
.
Lemma 1.2. Let Q be a nite connected acyclic quiver, the preprojective algebra of
Q and X mod. Then XI
i
is maximal amongst submodules Y of X such that any
composition factor of X/Y is isomorphic to a simple module S
i
.
1.2. Support -tilting modules. We give denitions of support -tilting modules. We
refer to [AIR] for the details about support -tilting modules, and see [ARS] for denitions
and properties of the functor .
Let be a nite dimensional algebra.
Denition 1.3. (a) We call X in mod -rigid if Hom

(X, X) = 0.
CLASSIFYING -TILTING MODULES OVER PREPROJECTIVE ALGEBRAS OF DYNKIN TYPE 5
(b) We call X in mod -tilting (respectively, almost complete -tilting) if X is -
rigid and [X[ = [[ (respectively, [X[ = [[ 1), where [X[ denotes the number of
non-isomorphic indecomposable direct summands of X.
(c) We call X in mod support -tilting if there exists an idempotent e of such that
X is a -tilting (/e)-module.
(d) We call a pair (X, P) of X mod and P proj -rigid if X is -rigid and
Hom

(P, X) = 0.
(e) We call a -rigid pair (X, P) a support -tilting (respectively, almost complete
support -tilting) pair if [X[ +[P[ = [[ (respectively, [X[ +[P[ = [[ 1).
Note that (X, P) is a -rigid (respectively, support -tilting) pair for if and only
if X is a -rigid (respectively, -tilting) (/e)-module, where e is an idempotent of
such that addP = adde [AIR, Proposition 2.3]. Moreover, if (X, P) and (X, P

) are
support -tilting pairs for , then we get addP = addP

. Thus, support -tilting module


X determines support -tilting pair (X, P) uniquely and we can identify support -tilting
modules with support -tilting pairs.
We call (X, P) basic if X and P are basic, and we say that (X, P) is a direct summand
of (X

, P

) if X is a direct summand of X

and P is a direct summand of P

. We denote
by s-tilt the set of isomorphism classes of basic support -tilting -modules.
Next we see some properties of support -tilting modules. The set of support -tilting
modules has a natural partial order dened as follows.
Denition 1.4. Let be a nite dimensional algebra. For T, T

s-tilt, we write
T

T
if FacT

FacT. Then gives a partial order on s-tilt [AIR, Theorem 2.18].


Then we give the following results, which play signicant roles in this paper.
Denition-Theorem 1.5. [AIR, Theorem 2.18, 2.28 and 2.30] Let be a nite dimen-
sional algebra. Then
(i) any basic almost support -tilting pair (U, Q) is a direct summand of exactly two
basic support -tilting pairs (T, P) and (T

, P

). Moreover we have T > T

or
T < T

.
Under the above setting, let X be an indecomposable -module satisfying either T =
U X or P = Q X. We write (T

, P

) =
(X,0)
(T, P) if X is a direct summand of
T and (T

, P

) =
(0,X)
(T, P) if X is a direct summand of P and we say that (T

, P

) is
a mutation of (T, P). In particular, we say that (T

, P

) is a left mutation (respectively,


right mutation) of (T, P) if T > T

(respectively, if T < T

) and write

(respectively,

+
). By (i), exactly one of the left mutation or right mutation occurs.
Now, assume that X is a direct summand of T and T = U X. In this case, for
simplicity, we write a left mutation T

X
(T) and a right mutation T

=
+
X
(T).
Moreover the following conditions are equivalent.
(ii) (a) T > T

.
(b) X / FacU.
Furthermore, if T is a -tilting -module, then the following property holds.
(iii) Let X
f
U

Y 0 be an exact sequence, where f is a minimal left (addU)-


approximation. Then one of the following holds.
(a) Y = 0 and

X
(T)

= U is a basic support -tilting -module.
6 YUYA MIZUNO
(b) Y ,= 0 and Y is a direct sum of copies of an indecomposable -module Y
1
,
which is not in addT, and

X
(T)

= Y
1
U is a basic -tilting -module.
Finally, we dene the following useful quiver.
Denition 1.6. Let be a nite dimensional algebra. We dene the support -tilting
quiver 1(s-tilt) as follows.
The set of vertices is s-tilt.
Draw an arrow from T to T

if T

is a left mutation of T.
The following theorem relates 1(s-tilt) with partially orders of s-tilt.
Theorem 1.7. [AIR, Corollary 2.34] The support -tilting quiver 1(s-tilt) is the Hasse
quiver of the partially ordered set s-tilt.
1.3. Torsion classes. The notion of torsion classes is quite important in the representa-
tion theory. As we will see below, support -tilting modules have a close relationship with
torsion classes.
Denition 1.8. Let be a nite dimensional algebra and T a full subcategory in mod.
(a) We call T torsion class (respectively, torsion-free class) if it is closed under factor
modules (respectively, submodules) and extensions.
(b) We call T a contravariantly nite subcategory in mod if for each X mod,
there is a map f : T X with T T such that Hom

(T

, T)
f
Hom

(T

, X) is
surjective for all T

T . Dually, a covariantly nite subcategory is dened. We


call T a functorially nite subcategory if it is contravariantly and covariantly nite.
We denote by f-tors the set of functorially nite torsion classes in mod.
Then, we have the following result, where we denote by P(T ) the direct sum of one copy
of each of the indecomposable Ext-projective objects in T (i.e. X mod [ Ext
1

(X, T ) =
0) up to isomorphism.
Theorem 1.9. [AIR] For a nite dimensional algebra , there are mutually inverse bi-
jections
s-tilt f-tors
given by s-tilt T FacT f-tors and f-tors T P(T ) s-tilt.
This explains why support -tilting modules are useful to investigate functorially nite
torsion classes. We have the following much stronger result under the assumption that
[s-tilt[ < .
Theorem 1.10. [IJ] For a nite dimensional algebra , the following conditions are
equivalent.
(a) [s-tilt[ < .
(b) Any torsion class in mod is functorially nite.
(c) Any torsion-free class in mod is functorially nite.
Therefore, if there are only nitely many basic support -tilting modules, then Theorem
1.9 give a bijection between s-tilt and torsion classes of mod.
CLASSIFYING -TILTING MODULES OVER PREPROJECTIVE ALGEBRAS OF DYNKIN TYPE 7
2. Support -tilting ideals and the Weyl group
The aim of this section is to give a complete description of all support -tilting modules
over preprojective algebras of Dynkin type and study their several properties.
Throughout this section, unless otherwise specied, let Q be a Dynkin quiver with
Q
0
= 1, . . . , n, the preprojective algebra of Q and I
i
:= (1 e
i
) for i Q
0
. We
denote by I
1
, . . . , I
n
the set of ideals of which can be written as
I
i
1
I
i
2
I
i
k
for some k 0 and i
1
, . . . , i
k
Q
0
.
2.1. Support -tilting ideals. In this subsection, we introduce support -tilting ideals.
We will show that elements of I
1
, . . . , I
n
are support -tilting ideals. This fact plays a
key role in this paper.
We start with the following denition.
Denition 2.1. We call a two-sided ideal I of support -tilting if I is a left support
-tilting -module and a right support -tilting -module.
The aim of this subsection is to prove the following result.
Theorem 2.2. Any T I
1
, . . . , I
n
is a basic support -tilting ideal of .
First we recall some exact sequences for later use (see [BBK]). For any i Q
0
, we can
take a minimal projective resolution of S
i
:= /I
i
as follows
e
i

p
2

aQ
1
,e(a)=i
e
s(a)

p
1

e
i

p
0

S
i

0.
Since Imp
1
= e
i
I
i
, we have the following exact sequences
(1) 0

e
i
I
i

e
i

p
0

S
i

0.
(2) e
i

p
2

aQ
1
,e(a)=i
e
s(a)

e
i
I
i

0.
As a direct sum of (1) and 0 (1 e
i
)
id
(1 e
i
) 0, we have
(3) 0

I
i

0
0 id


(p
0
0)

S
i

0.
Moreover, we have Ker p
2
= S
(i)
, where : Q
0
Q
0
is a Nakayama permutation of
(i.e. D(e
(i)
)

= e
i
) [BBK, Proposition 4.2]. Therefore, we have the following exact
sequences
(4) 0

S
(i)

e
i

p
2

aQ
1
,e(a)=i
e
s(a)
.
Then we give the following lemma.
Lemma 2.3. I
i
is a basic support -tilting ideal of for any i Q
0
.
8 YUYA MIZUNO
Proof. It is clear that I
i
is basic. We will show that I
i
is a right support -tilting -
module; the proof for a left -module is similar. If e
i
is a simple module, then is a
simple algebra since is a preprojective algebra. Thus, we have I
i
= 0 and it is a support
-tilting module.
Assume that e
i
is not a simple module and hence e
i
I
i
,= 0. Since is selnjective, e
j

has a simple socle for any j Q


0
and these simple modules are mutually non-isomorphic.
Therefore, e
i
I
i
is not isomorphic to e
j
for any j Q
0
. Hence, we obtain [I
i
[ = [[. Thus,
it is enough to show that Hom

(I
i
, (I
i
)) = 0.
Applying the functor := DHom

(, ) to (2), we have the following exact sequence


(5) 0

(e
i
I
i
)

(e
i
)
p
2

aQ
1
,e(a)=i
(e
s(a)
)
by the denition of the AR translation. On the other hand, applying the functor to
(4), we have the following exact sequence
(6) 0

(S
(i)
)

(e
i
)
p
2

aQ
1
,e(a)=i
(e
s(a)
).
Since is selnjective, we have (S
(i)
)

= (Soc(e
i
))

= Top(e
i
)

= S
i
. Therefore,
comparing (5) with (6), we get (I
i
)

= (e
i
I
i
)

= S
i
. Since S
i
/ add(Top(I
i
)), we have
Hom

(I
i
, (I
i
)) = 0.
Next we show the following property.
Lemma 2.4. Let T be a support -tilting -module. For a simple
op
-module S, at least
one of the statements T

S = 0 and Tor

1
(T, S) = 0 holds.
Proof. By [AIR, Proposition 2.5], we can take a minimal projective presentation P
1

P
0
T 0 such that P
0
and P
1
do not have a common summand. Thus, at least one of
the statement P
0

S = 0 and P
1

S = 0 holds. Since we have Tor

1
(T, S)

= P
1

S
and T

S

= P
0

S, the assertion follows.


From the above lemma, we have the following corollary.
Corollary 2.5. Let T be a support -tilting -module. If TI
i
,= T, then we have T

I
i

=
TI
i
.
Proof. By the assumption TI
i
,= T and Lemma 1.2, we have T

S
i
,= 0. Then, by
Lemma 2.4, we obtain Tor

1
(T, S
i
) = 0. By applying the functor T

to the exact
sequence (3), we get the following exact sequence
0 = Tor

1
(T, S
i
) T

I
i
T

= T.
Hence we have T

I
i

= Im(T

I
i
T) = TI
i
.
Next we give the following easy observation.
Lemma 2.6. Let I be a two-sided ideal of . For a primitive idempotent e
i
with i Q
0
,
e
i
I is either indecomposable or zero.
Proof. Since is selnjective, e
i
has a simple socle. Thus the submodule e
i
I of e
i
has
a simple socle or zero.
CLASSIFYING -TILTING MODULES OVER PREPROJECTIVE ALGEBRAS OF DYNKIN TYPE 9
Moreover, we need the following lemmas.
Lemma 2.7. Let I be a two-sided ideal of . Then there exists a surjective algebra map
End

(I), (I x x I).
In particular, for an idempotent e
i
, i Q
0
, we have a surjective map
e
i
(1 e
i
) Hom

((1 e
i
)I, e
i
I).
Proof. We have to show that, for any f End

(I), there exists such that f = () :


I I. Since is selnjective, we have the following commutative diagram
0

I

0

I

,
where is the canonical inclusion and h : is the -module map. Put h(1) = .
Then h is given by left multiplication with . Thus, we obtain f(x) = h(x) = x for any
x I and we have proved our claim. The second statement easily follows from the rst
one.
Lemma 2.8. Let T be a support -tilting ideal of . If I
i
T ,= T, then we have e
i
T /
Fac((1 e
i
)T). In particular, T has a left mutation

e
i
T
(T).
Proof. By Lemma 2.6 and the assumption, e
i
T ,= 0 is indecomposable. We assume that
e
i
T Fac((1 e
i
)T). Then, there exist an integer d > 0 and a surjective map
(1 e
i
)T
d
e
i
T.
By Lemma 2.7, there exist
m
(1 m d) such that
d

m=1
e
i

m
(1 e
i
) (1 e
i
)T = e
i
T.
On the other hand, since I
i
= (1 e
i
), we get
e
i
(1 e
i
) (1 e
i
)T = e
i
I
i
T.
Thus we have
e
i
T =
d

m=1
e
i

m
(1 e
i
)T e
i
I
i
T,
which is a contradiction. The second statement follows from Denition-Theorem 1.5 (ii).

Moreover we need to prove the following.


Lemma 2.9. Let T be a support -tilting ideal of . For the map p
2
in the sequence (2),
the map
e
i

T
p
2

aQ
1
,e(a)=i
e
s(a)

T
is a left (add((1 e
i
)T))-approximation.
10 YUYA MIZUNO
Proof. We have to show that
Hom

aQ
1
,e(a)=i
e
s(a)

T, (1 e
i
)T)
(p
2

T,(1e
i
)T)
Hom

(e
i

T, (1 e
i
)T) 0.
Let E := End

(T). We have
Hom

(e
i

T, (1 e
i
)T)

= Hom

(e
i
, Hom

(T, (1 e
i
)T))

= Hom

(e
i
, (1 e
i
)E),
and similarly
Hom

aQ
1
,e(a)=i
e
s(a)

T, (1 e
i
)T)

= Hom

aQ
1
,e(a)=i
e
s(a)
, (1 e
i
)E).
Then, from their functoriality, we write the above surjection by
Hom

aQ
1
,e(a)=i
e
s(a)
, (1 e
i
)E)
(p
2
,(1e
i
)E)
Hom

(e
i
, (1 e
i
)E) 0.
On the other hand, by Lemma 2.7, there exists a surjective algebra map
g : E,
which admits -module map. Then, we have the following commutative diagram
Hom

aQ
1
,e(a)=i
e
s(a)
, (1 e
i
))
(p
2
,(1e
i
))

(

aQ
1
,e(a)=i
e
s(a)
, g)

Hom

(e
i
, (1 e
i
))
(e
i
, g)

Hom

aQ
1
,e(a)=i
e
s(a)
, (1 e
i
)E)
(p
2
,(1e
i
)E)

Hom

(e
i
, (1 e
i
)E).
It is clear that Hom

aQ
1
,e(a)=i
e
s(a)
, g) and Hom

(e
i
, g) are surjective since
e
s(a)
and e
i
are projective. Moreover, by (4), Hom

(p
2
, (1 e
i
)) is also surjective.
Consequently, Hom

(p
2
, (1e
i
)E) is also surjective. Then the statement follows from the
additivity of the functors.
Now we apply the above results to the following key proposition.
Proposition 2.10. Let T I
1
, . . . , I
n
and assume that T is a basic support -tilting
ideal of . Then I
i
T is a basic support -tilting -module.
Proof. There is nothing to show if I
i
T = T. Assume that I
i
T ,= T. Then, by Lemma 2.8,
there exists a left mutation

e
i
T
(T).
Now let e be an idempotent of such that T is a -tilting (/e)-module. By applying
the functor

T to (2), we get exact sequence


e
i

T
p
2

aQ
1
,e(a)=i
e
s(a)

T
e
i
I
i

T 0
CLASSIFYING -TILTING MODULES OVER PREPROJECTIVE ALGEBRAS OF DYNKIN TYPE 11
of (/e)-modules. By applying Corollary 2.5 to the support -tilting left -module T,
we have e
i
I
i

T

= e
i
I
i
T. Moreover, Lemma 2.6 implies that e
i
I
i
T is indecomposable.
On the other hand, from Lemma 2.9, p
2

T is a left (add((1 e
i
)T))-approximation.
If e
i
I
i
T = 0, then p
2

T is clearly left minimal. Assume e


i
I
i
T ,= 0. Since e
j
T has a
simple socle for any j Q
0
and these simple modules are mutually non-isomorphic,

T
is a radical map and hence p
2

T is left minimal. Then, applying Denition-Theorem


1.5 (iii), we conclude that

e
i
T
(T)

= e
i
I
i
T (1 e
i
)T = I
i
T is a basic support -tilting
(/e)-module.
Now we are ready to prove Theorem 2.2.
Proof of Theorem 2.2. We use induction on the number of products of the ideal I
i
, i Q
0
.
By Lemma 2.3, I
i
is a basic support -tilting ideal of for any i Q
0
. Assume that
I
i
1
I
i
k
is a basic support -tilting ideal for i
1
, . . . , i
k
Q
0
and k > 1. By Proposition
2.10, I
i
0
(I
i
1
I
i
k
) is a basic support -tilting -module for i
0
Q
0
. Similarly we can
show that it is a basic support -tilting left -module.
At the end of this subsection, we give the following lemma for later use.
Lemma 2.11. Let T I
1
, . . . , I
n
. If I
i
T ,= T, then there is a left mutation of T :

e
i
T
(T)

= I
i
T.
Proof. This follows from Lemma 2.8 and Proposition 2.10.
Now we show examples of support -tilting modules and their mutations.
Example 2.12. (a) Let be the preprojective algebra of type A
2
. In this case, 1(s-tilt)
is given as follows.
1
2
1
I
1

0 1
I
2

1
2
2
1
I
2

I
1

0 0
2
2
1
I
2

2 0
I
1

Here we represent modules by their radical ltrations and we write a direct sum XY
by X Y . For example,
1
2
1 denotes the support -tilting pair (e
1
S
1
, 0) and 0 1 denotes
(S
1
, P
2
). Note that I
i
is given by left multiplications.
(b) Let be the preprojective algebra of type A
3
. In this case, 1(s-tilt) is given as
follows
12 YUYA MIZUNO
1 2
0 2 1

1
2 1 2
3 2 1

1
0 2 1

1
2 1
3 2 1

1 2
2 31 2
3 2 1

1
2 31
3 2 1

2
0 2 1

0 0 1

2 2
3 2 1

1 2 3
2 31 2
3 2 1

1 3
2 31 2
3 2 1

2
2 13 2
3 2 1

3 0 1

0 2 0

0 0 0
31
3 2 1

2 3
2 13 2
3 2 1

3
31 2
3 2 1

2
3 2 0

3 0 0

2 3
3 2 0

3
2 3 2
3 2 1

3
3 2 0

3
3 2
3 2 1

In these two examples, 1(s-tilt) consists of a nite connected component. We will


show that this is the case for preprojective algebras of Dynkin type in the sequel. Thus,
all support -tilting modules can be obtained by mutations from .
2.2. Bijection between support -tilting modules and the Weyl group. In the
previous subsection, we have shown that elements of I
1
, . . . , I
n
are basic support -tilting
ideals. In this subsection, we will show that any basic support -tilting -module is given
as an element of I
1
, . . . , I
n
. Moreover, we show that there exists a bijection between
the Weyl group and basic support -tilting -modules, which implies that there are only
nitely many basic support -tilting -modules.
First we recall the following denition.
Denition 2.13. Let Q be a nite connected acyclic quiver with vertices Q
0
= 1, . . . , n.
The Coxeter group W
Q
associated to Qis dened by the generators s
1
, . . . , s
n
and relations.
s
2
i
= 1,
s
i
s
j
= s
j
s
i
if there is no arrow between i and j in Q,
s
i
s
j
s
i
= s
j
s
i
s
j
if there is precisely one arrow between i and j in Q.
Each element w W
Q
can be written in the form w = s
i
1
s
i
k
. If k is minimal among
all such expressions for w, then k is called the length of w and we denote by l(w) = k. In
this case, we call s
i
1
s
i
k
a reduced expression of w.
CLASSIFYING -TILTING MODULES OVER PREPROJECTIVE ALGEBRAS OF DYNKIN TYPE 13
Note that W
Q
is the Weyl group if Q is Dynkin. Then, we give the following important
result.
Theorem 2.14. There exists a bijection W
Q
I
1
, . . . , I
n
. It is given by w I
w
=
I
i
1
I
i
2
I
i
k
for any reduced expression w = s
i
1
s
i
k
.
The statement is given in [BIRS, III.1.9] for non-Dynkin quivers. The same statement
also holds for Dynkin quivers. For the convenience of the reader, we give a complete proof
of Theorem 2.14 for Dynkin quivers.
For a proof, we provide the following set-up.
Notation 2.15. Let

Q be an extended Dynkin quiver obtained from Q by adding a vertex
0 (i.e.

Q
0
= 0 Q
0
) and the associated arrows. We denote by

the completion of the
associated preprojective algebra and

I
i
:=

(1 e
i
)

for i

Q. For each w W

Q
, let

I
w
:=

I
i
1

I
i
k
, where w = s
i
1
s
i
k
is a reduced expression (this is well-dened [BIRS]).
Note that we have
=

/e
0
, I
i
=

I
i
/e
0
.
Proof of Theorem 2.14. One can check that the map
W
Q
I
1
, . . . , I
n
, w = s
i
1
s
i
k
I
w
:= I
i
1
I
i
2
I
i
k
is well-dened and surjective by [BIRS, Theorem III.1.9], where the assumption that Q is
non-Dynkin is not used. We only have to show the injectivity.
Since i
1
, , i
k
Q
0
, we have s
i
j
,= s
0
for any 1 j k. Thus we have
I
w
= I
i
1
I
i
k
= (

I
i
1
/e
0
) (

I
i
k
/e
0
) = (

I
i
1

I
i
k
)/e
0
=

I
w
/e
0
.
Therefore, if I
w
= I
w
for w, w

W
Q
, then we have

I
w
/e
0
=

I
w
/e
0
. Since s
0
does
not appear in reduced expressions of w and w

, we have e
0


I
w
and e
0


I
w
. Hence, we
obtain

I
w
=

I
w
and conclude w = w

by [BIRS, Theorem III.1.9].


Thus, for any w W
Q
, we have support -tilting module I
w
. Then we give an explicit
description of mutations using this property.
Let (I
w
, P
w
) be a basic support -tilting pair, where P
w
is a basic projective -module
which is determined from I
w
(subsection 1.2). Because e
i
has simple socle S
(i)
, where
: Q
0
Q
0
is a Nakayama permutation of , if e
i
I
w
= 0, then e
(i)
addP
w
and hence
P
w
=

iQ
0
,e
i
Iw=0
e
(i)
. Then, for i Q
0
, we write a mutation

i
(I
w
, P
w
) :=
_

(e
i
Iw,0)
(I
w
, P
w
) if e
i
I
w
,= 0

(0,e
(i)
)
(I
w
, P
w
) if e
i
I
w
= 0.
Then we have the following result.
Theorem 2.16. For any w W
Q
and i Q
0
, a mutation of basic support -tilting
-module I
w
is given as follows:

i
(I
w
, P
w
)

= (I
s
i
w
, P
s
i
w
).
In particular, W
Q
acts transitively and freely on s-tilt by s
i
(I
w
, P
w
) :=
i
(I
w
, P
w
).
Proof. By Theorem 2.2 and 2.14, I
w
and I
s
i
w
are basic support -tilting modules, which
are not isomorphic.
14 YUYA MIZUNO
Let w = s
i
1
s
i
k
be a reduced expression. If l(s
i
w) > l(w), then s
i
w = s
i
s
i
1
s
i
k
is
a reduced expression. Hence, we get (1 e
i
)I
s
i
w
= (1 e
i
)I
i
I
w
= (1 e
i
)I
w
. Therefore
we have
(I
s
i
w
, P
s
i
w
)

=
_
(e
i
I
i
I
w
(1 e
i
)I
w
, P
w
) if e
i
I
i
I
w
,= 0
((1 e
i
)I
w
, P
w
e
(i)
) if e
i
I
i
I
w
= 0.
Thus (I
w
, P
w
) and (I
s
i
w
, P
s
i
w
) have a common almost complete support -tilting pair
((1 e
i
)I
w
, P
w
) as a direct summand.
On the other hand, if l(s
i
w) < l(w), then u := s
i
w satises l(u) < l(s
i
u). Similarly, it
shows that (1 e
i
)I
s
i
w
= (1 e
i
)I
w
and (I
w
, P
w
) = (I
s
i
u
, P
s
i
u
) and (I
s
i
w
, P
s
i
w
) = (I
u
, P
u
)
have a common almost complete support -tilting pair as a direct summand. Therefore,
by Denition-Theorem 1.5 (i), the rst statement follows. The second one follows from
Theorem 2.14.
Using the above result, we give the following crucial lemma.
Lemma 2.17. The support -tilting quiver 1(s-tilt) has a nite connected component.
Proof. Let C be the connected component of 1(s-tilt) containing . By Theorem 2.2,
I
w
is a basic support -tilting -module for any w W
Q
. We will show that I
w
[ w W
Q

gives the set of vertices in C.


We show that every I
w
belongs to C by induction on the length of w W
Q
. Assume
that I
w
[ w W
Q
with l(w) k belong to C. Take w W
Q
with l(w) = k + 1 and
decompose w = s
i
w

, where l(w

) = k. Then, we have I
w
= I
i
I
w
,= I
w
by Theorem 2.14
and we get

e
i
I
w

(I
w
)

= I
i
I
w
by Lemma 2.11. Thus, I
w
belongs to C from Theorem 1.7
and hence our claim follows inductively.
Moreover we show that any vertex in C has a form I
w
for some w W
Q
. This follows
the fact that the neighbours of I
w
are given as I
s
i
w
for i Q by Theorem 2.16. Therefore
we obtain the assertion.
Now we recall the following useful result.
Proposition 2.18. [AIR, Corollary 2.38] If 1(s-tilt) has a nite connected component
C, then C = 1(s-tilt).
As a conclusion, we can obtain the following statement.
Theorem 2.19. Any basic support -tilting -module is isomorphic to an element of
I
1
, . . . , I
n
. In particular, it is isomorphic to a support -tilting ideal of .
Proof. By Lemma 2.17, 1(s-tilt) has a nite connected component, and Corollary 2.18
implies that it coincides with 1(s-tilt). Thus, any support -tilting -module is given
by I
w
for some w W
Q
. In particular, it is a support -tilting ideal from Theorem 2.2.
Furthermore, we give the following lemma.
Lemma 2.20. If right ideals T and U are isomorphic as -modules, then T = U.
Proof. Assume that f : T U is an isomorphism of -modules. Since is selnjective,
we have the following commutative diagram
0

T

0

U

,
CLASSIFYING -TILTING MODULES OVER PREPROJECTIVE ALGEBRAS OF DYNKIN TYPE 15
where is the canonical inclusion and h : is a -module map. Put h(1) = . By
the commutative diagram, we have f(T) = h(T) and hence we get U = h(T) = T T.
By the similar argument for f
1
, we can show that T U. Thus we conclude that
T = U.
Now we are ready to prove one of the main result of this paper.
Theorem 2.21. There exist bijections between the isomorphism classes of basic support
-tilting -modules, basic support -tilting
op
-modules and the elements of W
Q
.
Proof. We will give a bijection between s-tilt and W
Q
. A bijection s-tilt
op
and W
Q
can be given similarly.
By Theorem 2.2 and 2.14, we have a map W
Q
w I
w
s-tilt. This map is
surjective since any support -tilting -module is isomorphic to I
w
for some w W
Q
by
Theorem 2.19. Moreover it is injective by Theorem 2.14 and Lemma 2.20. Thus we get
the conclusion.
Remark 2.22. It is known that the order of W
Q
is nite and the explicit number is
determined by the type of underlying graph of Q as follows [H, Section 2.11].

Q A
n
D
n
E
6
E
7
E
8
(n + 1)! 2
n1
n! 51840 2903040 696729600
Finally, we give an application related to annihilators of s-tilt using the above result.
Note that tilting modules are nothing but faithful support -tilting modules [AIR, Propo-
sition 2.2]. Hence, it is important to investigate annihilators of support -tilting modules.
We can completely determine them for preprojective algebras of Dynkin type. Here, we
denote by annX := [ X = 0 the annihilator of X mod.
Corollary 2.23. We have ann I
w
= I
w
1
w
0
, where w
0
is the longest element in W
Q
.
Thus, we have a bijection
s-tilt s-tilt, T annT.
For a proof, we recall the following useful result.
Proposition 2.24. [ORT, Proposition 6.4] Let w
0
be the longest element in W
Q
. Then
we have the following isomorphisms
(a) DI
w

= /I
w
1
w
0
in mod
op
.
(b) DI
w

= /I
w
0
w
1 in mod.
Then we give a proof of Corollary 2.23.
Proof of Corollary 2.23. It is clear that the right annihilator of I
w
coincides with the left
annihilator of DI
w
. On the other hand, we have isomorphism DI
w

= /I
w
1
w
0
of
op
-
modules by Proposition 2.24. Thus, we get the rst statement. The second one follows
from Theorem 2.14.
2.3. Partial orders of support -tilting modules and the Weyl group. In this
subsection, we study a relationship between a partial order of support -tilting modules
and that of W
Q
. In particular, we show that the bijection W
Q
s-tilt given in Theorem
2.21 induces an isomorphism as partially ordered sets.
16 YUYA MIZUNO
Denition 2.25. Let u, w W
Q
. We write u
L
w if there exist s
i
k
, . . . , s
i
1
such that
w = s
i
k
. . . s
i
1
u and l(s
i
j
. . . s
i
1
u) = l(u) + j for 0 j k. Clearly
L
gives a partial
order on W
Q
, and we call
L
the left order (it is also called weak order). We denote by
1(W
Q
,
L
) the Hasse quiver of left order on W
Q
.
The following assertion is immediate from the denition of the left order.
Lemma 2.26. An arrow in 1(W
Q
,
L
) is given by
w s
i
w (if l(w) > l(s
i
w)),
s
i
w w (if l(w) < l(s
i
w))
for any w W
Q
and i Q
0
.
Example 2.27. (a) Let Q be a quiver of type A
2
. Then 1(W
Q
,
L
) is given as follows.
132
.

312
123 321

213

231
(b) Let Q be a quiver of type A
3
. Then 1(W
Q
,
L
) is given as follows.
2341
.

2314
.

3241
.

3214

2134
/

3124
.

2431
.

3421
.

2413
.

1234 1324

2143

3412
.

4231

4321

3142
.

1243

1342

4213
.

4312

4123
.

1423

4132

1432

Next we give the following observation.


Lemma 2.28. Let w W
Q
and i Q
0
.
CLASSIFYING -TILTING MODULES OVER PREPROJECTIVE ALGEBRAS OF DYNKIN TYPE 17
(i) If l(w) < l(s
i
w), then I
i
I
w
= I
s
i
w
I
w
and we have a left mutation

i
(I
w
, P
w
).
(ii) If l(w) > l(s
i
w), then I
i
I
w
= I
w
I
s
i
w
and we have a right mutation
+
i
(I
w
, P
w
).
Proof. Let w = s
i
1
s
i
k
be a reduced expression. If l(s
i
w) > l(w), then s
i
w = s
i
s
i
1
s
i
k
is a reduced expression. Thus, we have I
i
I
w
= I
s
i
w
I
w
by Theorem 2.14 and, by Lemma
2.11, we get a left mutation

i
(I
w
, P
w
).
On the other hand, if l(s
i
w) < l(w), then u := s
i
w satises l(u) < l(s
i
u). Then,
similarly, we have I
i
I
u
= I
s
i
u
= I
w
I
u
= I
s
i
w
. In particular, we have I
i
I
w
= I
2
i
I
u
=
I
i
I
u
= I
w
and left mutation

i
(I
u
, P
u
) = (I
w
, P
w
), or equivalently
+
i
(I
w
, P
w
) = (I
u
, P
u
) =
(I
s
i
w
, P
s
i
w
).
Summarizing previous results, we give the following equivalent conditions that a left (or
right) mutation occurs.
Corollary 2.29. Let w W
Q
and i Q
0
.
(a) The following conditions are equivalent :
(i) l(w) < l(s
i
w).
(ii) I
i
I
w
,= I
w
.
(iii) We have a left mutation

i
(I
w
, P
w
).
(b) The following conditions are equivalent :
(i) l(w) > l(s
i
w).
(ii) I
i
I
w
= I
w
.
(iii) We have a right mutation
+
i
(I
w
, P
w
).
Proof. It is enough to show (a).
By Lemma 2.11 and 2.28, (i) implies (ii) and (ii) implies (iii). Assume that we have a
left mutation

i
(I
w
, P
w
). If l(w) > l(s
i
w), then we get a right mutation
+
i
(I
w
, P
w
) by
Lemma 2.28, which contradicts Denition-Theorem 1.5. Thus (iii) implies (i).
Finally we give the following results.
Theorem 2.30. The bijection in W
Q
s-tilt in Theorem 2.21 gives an isomorphism
of partially ordered sets (W
Q
,
L
) and (s-tilt, )
op
.
Proof. It is enough to show that 1(W
Q
,
L
) coincides with 1(s-tilt)
op
since the Hasse
quivers determine the partial orders.
By Theorem 1.7 and Corollary 2.29, arrows starting from I
w
are given by I
w
I
i
I
w
in
1(s-tilt), where i Q
0
satises l(w) < l(s
i
w).
On the other hand, Lemma 2.26 implies that arrows to w are given by s
i
w w
in 1(W
Q
,
L
), where s
i
satises l(w) < l(s
i
w). Therefore 1(s-tilt)
op
coincides with
1(W
Q
,
L
) and the statement follows.
We remark that the Bruhat order on W
Q
coincides with the reverse inclusion relation
on I
1
, I
n
[ORT, Lemma 6.5].
3. g-matrices and cones
In this section, we study g-vectors [DK] (which is also called index [AR2, P]) and g-
matrices of support -tilting modules. We give a complete description of g-matrices in
terms of reection transformations. Moreover, we study cones dened by g-matrices, which
give geometric realizations of support -tilting modules, and we show that they coincide
with chambers of the associated root system.
18 YUYA MIZUNO
Throughout this section, unless otherwise specied, let Q be a Dynkin quiver with
vertices Q
0
= 1, . . . , n, be the preprojective algebra of Q and I
i
= (1 e
i
) for
i Q
0
.
First, we dene g-vectors in our setting. We refer to [AIR, section 5] for a background
of this notion (see also section 4).
Denition 3.1. By Krull-Schmidt theorem, we identify the set of isomorphism classes of
projective -modules with submonoid Z
n
0
of the free abelian group Z
n
.
For a -module X, take a minimal projective presentation
P
1
(X)

P
0
(X)

X

0
and let g(X) = (g
1
(X), , g
n
(X))
t
:= P
0
(X)P
1
(X) Z
n
, where t denotes the transpose
matrix. Then, for w W
Q
and i Q
0
, we dene a g-vector by
Z
n
g
i
(w) =
_
g(e
i
I
w
) if e
i
I
w
,= 0
(0, . . . , 0,
(i)
1, 0, , 0)
t
if e
i
I
w
= 0,
where : Q
0
Q
0
is a Nakayama permutation of (i.e. D(e
(i)
)

= e
i
)). It is shown in
[AIR, Theorem 5.1] that g
1
(w), , g
n
(w) form a basis of Z
n
. Then we dene a g-matrix
of support -tilting -module I
w
, or simply of w, by
g(w) := (g
1
(w), , g
n
(w)) GL
n
(Z)
and its cone by
C(w) := a
1
g
1
(w) + +a
n
g
n
(w) [ a
i
R
>0
.
Next, we dene g-vectors for extended Dynkin cases. We follow Notation 2.15. By
Theorem 2.14, there exists a bijection
W

I
0
,

I
1
, . . . ,

I
n
, w

I
w
=

I
i
1

I
i
2

I
i
k
,
where w = s
i
1
s
i
k
is a reduced expression of w. Note that

I
w
is a tilting

-modules of
projective dimension at most one [IR] (hence e
i

I
w
,= 0). For w W

Q
and i

Q
0
, take a
minimal projective presentation of the

-module e
i

I
w
P
1
(e
i

I
w
)

P
0
(e
i

I
w
)

e
i

I
w

0.
Then we dene a g-vector by
g
i
(w) := (g
0
(e
i

I
w
), , g
n
(e
i

I
w
))
t
:= P
0
(e
i

I
w
) P
1
(e
i

I
w
) Z
n+1
,
and we dene a g-matrix by
g(w) := ( g
0
(w), , g
n
(w)) GL
n+1
(Z).
We use these notations in this section and, for simplicity, denote by () :=

.
Then, we give the following results.
Proposition 3.2. Let w W
Q
and i Q
0
. Take a minimal projective presentation of
the

-module e
i

I
w
(7) 0

P
1
f
1

P
0
f
0

e
i

I
w

0.
Then, applying the functor

to (7), we have the following exact sequence


0


1
(e
i
/e
i
I
w
)

P
1
f
1

P
0
f
0

e
i
I
w

0 ,
CLASSIFYING -TILTING MODULES OVER PREPROJECTIVE ALGEBRAS OF DYNKIN TYPE 19
where
1
:= Hom

(D, ). Moreover, if e
i
I
w
,= 0, then the sequence
P
1
f
1

P
0
f
0

e
i
I
w

0
is a minimal projective presentation of the -module e
i
I
w
.
Proof. Since e
i
,= e
0
and =

/e
0
, we have e
i

I
w



= e
i

I
w
/e
i

I
w
e
0


= e
i

I
w
/e
i
e
0


=
e
i
I
w
. Then, by applying the functor

to (7), we have the following exact sequence


0

Tor

1
(e
i

I
w
, )
g

P
1
f
1

P
0
f
0

e
i
I
w

0 .
On the other hand, we have the following exact sequence
0

e
i

I
w

e
i


e
i

/e
i

I
w

0.
Then, we obtain -module isomorphisms
Tor

1
(e
i

I
w
, )

= DExt
1

(e
i

I
w
, D) ([CE])

= DExt
2

(e
i

/e
i

I
w
, D)

= Hom

(D, e
i

/e
i

I
w
) (2-CY property)

= Hom

(D, e
i
/e
i
I
w
) (e
i

/e
i

I
w

= e
i
/e
i
I
w
)

=
1
(e
i
/e
i
I
w
).
Now assume that e
i
I
w
,= 0. Then, clearly
1
(e
i
/e
i
I
w
) is not projective and hence g
is a radical map. Moreover, it is clear that f
0
is a projective cover by the assumption of
f
0
. Thus we obtain the conclusion.
Remark 3.3. It is known that, for every non-projetive indecomposable module X over
a preprojective algebra of Dynkin type, its third syzygy
3
X is isomorphic to
1
(X)
[AR1, BBK]. This property appears in the above proposition.
Using the above proposition, we give the following key result.
Proposition 3.4. For any i Q
0
and w W
Q
, the g-vector g
i
(w) equals the image of
g
i
(w) under the projection Z
n+1
Z
n
with respect to the i-th entries (i = 1, , n).
Proof. Take a minimal projective presentation of the

-module e
i

I
w
(8) 0

P
1
f
1

P
0
f
0

e
i

I
w

0.
We decompose P
0
= P

0
(e
0

and P
1
= P

1
(e
0

)
m
, where (e
0

(respectively,
(e
0

)
m
) is a maximal direct summand of P
0
(respectively, P
1
) which belongs to add(e
0

).
For g
i
(w) = (g
0
(e
i

I
w
), g
1
(e
i

I
w
), , g
n
(e
i

I
w
))
t
, it is clear that (g
1
(e
i

I
w
), . . . , g
n
(e
i

I
w
))
t
is
determined by P

0
and P

1
. We will show that it coincides with g
i
(w).
By Proposition 3.2, applying

to (8), we have the following exact sequence


(9) 0


1
(e
i
/e
i
I
w
)

P

1
f
1

0
f
0

e
i
I
w

0.
(i) Assume that P

0
,= 0. It implies that e
i
I
w
,= 0 and, by Proposition 3.2, the se-
quence (9) gives a minimal projective presentation of e
i
I
w
. Thus g
i
(w) coincides with
(g
1
(e
i

I
w
), . . . , g
n
(e
i

I
w
))
t
.
20 YUYA MIZUNO
(ii) Assume that P

0
= 0. It implies that e
i
I
w
= 0 and g
i
(w) = (0, . . . , 0,
(i)
1, 0, , 0)
t
by the denition. On the other hand, by (9), we have P

=
1
(e
i
)

= e
(i)
and
hence P

= e
(i)

. Thus g
i
(w) = (g
0
(e
i

I
w
), 0, . . . , 0,
(i)
1, 0, , 0)
t
and we obtain the
conclusion.
Next we introduce the following notations [Bou, BB].
Denition 3.5. Let Q be a nite connected acyclic quiver with vertices Q
0
= 1, . . . , n.
For i, j Q
0
, let m
ij
:= a Q
1
[ s(a) = i, e(a) = j +a Q
1
[ s(a) = j, e(a) = i. Let
e
1
, . . . , e
n
be the canonical basis of Z
n
.
The geometric representation : W
Q
GL
n
(Z) of W
Q
is well-dened by
(s
i
)(e
j
) =
i
(e
j
) := e
j
+ (m
ij
2
ij
)e
i
,
where
ij
denotes the Kronecker delta. On the other hand, the contragradient

: W
Q

GL
n
(Z) of the geometric representation is dened by

(s
i
)(e
j
) =

i
(e
j
) =
_
_
_
e
j
i ,= j
e
j
+

aQ
1
,s(a)=i
e
e(a)
i = j.
For an arbitrary expression w = s
i
1
. . . s
i
k
of w, we dene (w) =
w
:=
i
k
. . .
i
1
and

(w) =

w
:=

i
k
. . .

i
1
.
We refer to [Bou] for the following properties.
Proposition 3.6. (a) The maps : W
Q
GL
n
(Z) and

: W
Q
GL
n
(Z) are
injective.
(b) We have (
i
)
2
= id = (

i
)
2
for any i Q
0
.
Then, we give the following important result, where we denote the set of g-matrices of
support -tilting -modules by g-mat().
Proposition 3.7. For any w W
Q
, we have g(w) =

(w). In particular, the map


W
Q
g-mat(), w g(w) is injective.
Proof. By Proposition 3.4, the g-matrix g(w) = (g
1
(w), , g
n
(w)) is given by 1-st to n-th
rows and columns of g(w) = ( g
0
(w), g
1
(w), , g
n
(w)).
On the other hand, the g-vector g
i
(w) Z
n+1
gives the class of e
i

I
w
in the Grothendieck
group of mod

for any i

Q
0
since

I
w
has the projective dimension at most one. Hence, by
[IR, Theorem 6.6], we have g(w) =

(w), where

is the contragradient W

Q
GL
n+1
(Z)
of the geometric representation.
For w = s
i
1
. . . s
i
k
, the 0-th column of

i
j
is (1, 0, . . . , 0)
t
by the assumption of s
i
j
,= s
0
for any 1 j k. Because 1-st to n-th rows and columns of

i
j
is equal to

i
j
by the
denition, we conclude that g(w) =

i
k
. . .

i
1
=

(w). The second statement follows


from Proposition 3.6.
Finally we consider the following settings.
Denition 3.8. An element of :=
w
(e
i
)
iQ
0
,wW
Q
is called a root. For a, b R
n
,
we write a natural pairing a, b := a b
t
. To each root x , we dene a hyperplane
H
x
:= y R
n
[ y, x = 0.
CLASSIFYING -TILTING MODULES OVER PREPROJECTIVE ALGEBRAS OF DYNKIN TYPE 21
A chamber of is a connected component of R
n

x
H
x
. We denote by C
0
the chamber
a
1
e
1
+ +a
n
e
n
[ a
i
R
>0
. Note that chambers are given by

wW
Q

w
(C
0
) and there
is a bijection between W
Q
and chambers via w

w
(C
0
) [Bou].
Then we obtain the following conclusion.
Theorem 3.9. (a) The set of g-matrices of support -tilting -modules coincides with
the subgroup

1
, . . . ,

n
of GL(Z
n
) generated by

i
for all i Q
0
.
(b) For w W
Q
, we have
C(w) =

w
(C
0
),
where C(w) denotes the cone of I
w
. In particular, cones of basic support -tilting
-modules give chambers of the associated root system .
Proof. From Proposition 3.7, (a) follows. Moreover, we have
C(w) = a
1
g
1
(w) + +a
n
g
n
(w) [ a
i
R
>0

= a
1

w
(e
1
) + +a
n

w
(e
n
) [ a
i
R
>0

w
(C
0
).
Thus, (b) follows.
Example 3.10. (a) Let be the preprojective algebra of type A
2
. In this case, the
g-matrices of Example 2.12 (a) are given as follows, where

1
=
_
1 0
1 1
_
and

2
=
_
1 1
0 1
_
.
1 1
0 1

0 1
1 1

1 0
0 1

0 1
1 0
1 0
1 1

1 1
1 0

Moreover, their chambers are given as follows.


1
1
1
0
1
0
1
0
0
1

1
1

(b) Let be the preprojective algebra of type A


3
. In this case, the g-matrices of
Example 2.12 (b) are given as follows.
22 YUYA MIZUNO
0 1 0
0 0 1
1 1 1

1 1 0
0 0 1
0 1 1

0 1 1
0 0 1
1 1 0

1 1 1
0 0 1
0 1 0

1 0 0
0 1 1
0 0 1

1 1 1
0 1 1
0 1 0

0 1 0
0 1 1
1 1 1

0 0 1
0 1 1
1 0 0

1 1 0
1 1 1
0 1 1

1 0 0
0 1 0
0 0 1

1 1 0
0 1 0
0 1 1

1 0 0
1 1 1
0 0 1

0 0 1
1 1 1
1 0 0

0 1 1
0 1 0
1 1 0

0 0 1
0 1 0
1 0 0
0 1 1
1 1 1
1 1 0

1 0 0
1 1 0
0 0 1

0 1 0
1 1 0
1 1 1

1 1 1
1 1 0
0 1 0

0 0 1
1 1 0
1 0 0

1 1 1
1 0 0
0 1 0

1 1 0
1 0 0
0 1 1

0 1 1
1 0 0
1 1 0

0 1 0
1 0 0
1 1 1

Remark 3.11. In the above examples, the rows of each matrix are sign-coherent, that
is, all elements are non-positive or non-negtative. This follows from the fact that, for a
minimal projective presentation of a support -tilting module T
P
1

P
0

T

0,
we have addP
0
addP
1
= 0 [AIR, Proposition 2.5]. Note that a similar observation was
made by [DK, section 2.4] for cluster-tilting objects in 2-CY triangulated categories.
4. Further connections
In this section, we extend our bijections by combining with other works, and explain
some relationships with other results.
We have the following bijections (We refer to [BY] for the denitions (g), (h), (i) and
(j)).
Theorem 4.1. Let Q be a Dynkin quiver with vertices Q
0
= 1, . . . , n and the prepro-
jective algebra of Q. There are bijections between the following objects.
(a) The elements of the Weyl group W
Q
.
(b) The set I
1
, . . . , I
n
.
CLASSIFYING -TILTING MODULES OVER PREPROJECTIVE ALGEBRAS OF DYNKIN TYPE 23
(c) The set of isomorphism classes of basic support -tilting -modules.
(d) The set of isomorphism classes of basic support -tilting
op
-modules.
(e) The set of torsion classes in mod.
(f) The set of torsion-free classes in mod.
(g) The set of isomorphism classes of basic two-term silting complexes in K
b
(proj).
(h) The set of intermediate bounded co-t-structures in K
b
(proj) with respect to the
standard co-t-structure.
(i) The set of intermediate bounded t-structures in D
b
(mod) with length heart with
respect to the standard t-structure.
(j) The set of isomorphism classes of two-term simple-minded collections in D
b
(mod).
(k) The set of quotient closed subcategories in modKQ.
(l) The set of subclosed subcategories in modKQ.
We have given bijections between (a), (b), (c) and (d). Bijections between (g), (h), (i)
and (j) are the restriction of [KY] and it is given in [BY, Corollary 4.3] (it is stated for
Jacobian algebras, but it holds for any nite dimensional algebra as they point out). Note
that compatibilities of mutations and partial orders of these objects are shown in [KY].
We will give bijections between (a), (e) and (f) by showing the following statement,
which provides complete descriptions of torsion classes and torsion-free classes in mod.
Proposition 4.2. (i) For any torsion class T in mod, there exists w W
Q
such
that T = FacI
w
.
(ii) For any torsion-free class T in mod, there exists w W
Q
such that T =
Sub/I
w
.
(iii) There exist bijections between W
Q
, torsion classes and torsion-free classes in mod.
Proof. Since there exists only nitely many support -tilting modules by Theorem 2.21,
Theorem 1.10 shows that torsion classes and torsion-free classes are functorially nite.
Thus, Theorem 1.9 gives a bijection between basic support -tilting -modules and torsion
classes of mod.
On the other hand, since T is functorially nite, there exists X mod such that
T = SubX by [S]. Then DSubX = FacDX is a functorially nite torsion class in mod
op
.
Similarly, via the bijection of Theorem 2.21, there exists w W
Q
such that DX = I
w

mod
op
. By Proposition 2.24, we have X

= DI
w

= /I
w
0
w
1, where w
0
is the longest
element in W
Q
. From the above arguments, (iii) is clear.
Remark 4.3. It is shown that objects FacI
w
and Sub/I
w
have several nice properties.
For example, FacI
w
and Sub/I
w
are Frobenius and, moreover, stable 2-CY categories
which have cluster-tilting objects. They also play important roles in the study of cluster
algebra structures for a coordinate ring of the unipotent cell associated with w (refer to
[BIRS, GLS3]).
It is known that a left order of W
Q
is a lattice (i.e. any two elements x, y W
Q
have a
greatest lower bound, called meet x
_
y and a least upper bound, called join x
_
y). From
the viewpoint of Theorem 4.2, it is natural to consider what is a join and meet in torsion
classes. We give an answer to the question as follows.
Proposition 4.4. Let T and T

be torsion classes of mod. Then we have T


_
T

=
T

and T
_
T

= TorsT , T

, where TorsT , T

is the smallest torsion class in


mod containing T and T

.
Proof. It is clear that T

and TorsT , T

are torsion classes. Since any torsion class


in mod is functorially nite by Theorem 1.10 and 2.21, the statement follows.
24 YUYA MIZUNO
Next we will see bijections between (a), (k) and (l), which are given by [ORT]. We
briey see their results and explain a relationship with our results.
Let X be a -module. We denote by X
KQ
the KQ-module by the restriction, that is,
we forget the action of the arrows a

Q. Then we dene ()
KQ
: mod modKQ, X
(X)
KQ
:= addX
KQ
. Take I
w
for w W
Q
. In [ORT], the authors show that (I
w
)
KQ
(respec-
tively, (/I
w
)
KQ
) is a quotient closed subcategory (respectively, subclosed subcategory)
of modKQ, and any quotient closed subcategory (respectively, subclosed subcategory) is
given in this form.
Now, we can conclude that the functor ()
KQ
gives a bijection between (e) and (k)
(similarly (f) and (l)). Indeed, any torsion class is given by FacI
w
for some w W
Q
from
Proposition 4.2 and we have (FacI
w
)
KQ
= (I
w
)
KQ
by a result of [ORT].
Moreover, it is natural to consider when a quotient closed subcategory (I
w
)
KQ
is a tor-
sion class. The answer is given by [ORT, Proposition 10.4] and they show that elements
of W
Q
satisfying a certain condition dened by c-sortable [Re] are in bijection with tor-
sion classes of modKQ by this correspondence. We emphasize that there exist bijections
between torsion classes of modKQ and several important objects such as clusters in the
cluster algebra given by Q (we refer to [IT, Theorem 1.1] for more details). From this
observation, we conclude that there exists a surjection from objects of Theorem 4.1 to
objects of [IT, Theorem 1.1].
Finally, we provide a bijection between (c) and (g), which is given by [AIR, Theorem
3.2], and explain a denition of g-vectors. Let K
b
(proj) be the homotopy category of
bounded complexes of proj. For a -module X, take a minimal projective presentation
of X
P
1
(X)

P
0
(X)

X

0.
We denote by P
X
:= (
1
P
1
(X)
0
P
0
(X)) K
b
(proj). Then, for a support -tilting
pair (X, P) for , the map (X, P) P
X
P[1] K
b
(proj) gives a two-term silting
complex in K
b
(proj) and it provides a bijection between (c) and (g). Note that the g-
vector of (X, P) is given as a class of the corresponding silting complex P
X
P[1] in the
Grothendieck group K
0
(K
b
(proj)) [AIR, section 5]. Furthermore, we remark that, since
is selnjective, it is shown that a -stable support -tilting -module X (i.e X

= (X) for
:= DHom

(, )) gives a two-term tilting complex by this correspondence [M]. Thus,


we can also conclude that there exist only nitely many basic two-term tilting complexes
in K
b
(proj).
References
[AIR] T. Adachi, O. Iyama, I. Reiten, -tilting theory, arXiv: 1210.1036.
[AIRT] C. Amiot, O. Iyama, I. Reiten, G. Todorov, Preprojective algebras and c-sortable words, Proc.
Lond. Math. Soc. (3) 104 (2012), no. 3, 513539.
[AR1] M. Auslander, I. Reiten, DTr-periodic modules and functors, Representation theory of algebras
(Cocoyoc, 1994), 3950, CMS Conf. Proc., 18, Amer. Math. Soc., Providence, RI, 1996.
[AR2] M. Auslander, I. Reiten, Modules determined by their composition factors, Ill. J. of Math 29 (1985),
280301.
[ARS] M. Auslander, I. Reiten, S. O. Smal, Representation Theory of Artin Algebras, Cambridge Studies
in Advanced Mathematics, 36. Cambridge University Press, Cambridge, 1995.
[BGL] D. Baer, W. Geigle, H. Lenzing, The preprojective algebra of a tame hereditary Artin algebra, Comm.
Algebra 15 (1987), no. 1-2, 425457.
[Boc] R. Bocklandt, Graded Calabi Yau algebras of dimension 3, J. Pure Appl. Algebra 212 (2008), no. 1,
1432.
CLASSIFYING -TILTING MODULES OVER PREPROJECTIVE ALGEBRAS OF DYNKIN TYPE 25
[BB] A. Bj orner, F. Brenti, Combinatorics of Coxeter groups, Graduate Texts in Mathematics, vol. 231,
Springer, New York, 2005.
[Bou] N. Bourbaki, Lie groups and Lie algebras. Chapters 46, Elements of Mathematics (Berlin), Springer-
Verlag, Berlin, 2002, Translated from the 1968 French original by Andrew Pressley.
[BBK] S. Brenner, M. C. R. Butler, A. D. King, Periodic Algebras which are Almost Koszul, Algebr.
Represent.Theory 5 (2002), no. 4, 331367.
[BY] T. Br ustle, D. Yang, Ordered exchange graphs, arXiv:1302.6045.
[BIRS] A. B. Buan, O. Iyama, I. Reiten, J. Scott, Cluster structures for 2-Calabi-Yau categories and
unipotent groups, Compos. Math. 145 (2009), 10351079.
[CE] H. Cartan, S. Eilenberg, Homological algebra, Princeton University Press, Princeton, NJ, 1956.
[CB] W. Crawley-Boevey, On the exceptional bres of Kleinian singularities, Amer. J. Math. 122 (2000),
no. 5, 10271037.
[DK] R. Dehy, B. Keller, On the combinatorics of rigid objects in 2-Calabi-Yau categories, Int. Math. Res.
Not. IMRN 2008, no. 11, Art. ID rnn029.
[DR1] V. Dlab, C. M. Ringel, The preprojective algebra of a modulated graph, Representation theory, II
(Proc. Second Internat. Conf., Carleton Univ., Ottawa, Ont., 1979), pp. 216231, Lecture Notes in Math.,
832, Springer, Berlin-New York, 1980.
[DR2] V. Dlab, C. M. Ringel, The module theoretical approach to quasi-hereditary algebras, Representa-
tions of algebras and related topics (Kyoto, 1990), 200224, London Math. Soc. Lecture Note Ser., 168,
Cambridge Univ. Press, Cambridge, 1992.
[GLS1] C. Geiss, B. Leclerc, J. Schroer, Rigid modules over preprojective algebras, Invent. Math. 165
(2006), no. 3, 589632.
[GLS2] C. Geiss, B. Leclerc, J. Schroer, Semicanonical bases and preprojective algebras II: A multiplication
formula, Compos. Math. 143 (2007), no. 5, 13131334.
[GLS3] C. Geiss, B. Leclerc, J. Schroer, Kac-Moody groups and cluster algebras, Adv. Math. 228 (2011),
no. 1, 329433.
[GP] I. M. Gelfand, V. A. Ponomarev, Model algebras and representations of graphs, Funktsional. Anal. i
Prilozhen. 13 (1979), no. 3, 112.
[H] J. E. Humphreys, Reection groups and Coxeter groups, Cambridge Studies in Advanced Mathematics,
29. Cambridge University Press, Cambridge, 1990.
[IT] C. Ingalls, H. Thomas, Noncrossing partitions and representations of quivers, Compos. Math. 145
(2009), no. 6, 15331562.
[IJ] O. Iyama, G. Jasso, in preparation.
[IR] O. Iyama, I. Reiten, Fomin-Zelevinsky mutation and tilting modules over Calabi-Yau algebras, Amer.
J. Math. 130 (2008), no. 4, 10871149.
[KY] S. Koenig, D. Yang, Silting objects, simple-minded collections, t-structures and co-t-structures for
nite-dimensional algebras, arXiv:1203.5657.
[L1] G. Lusztig, Quivers, perverse sheaves, and quantized enveloping algebras, J. Amer. Math. Soc. 4
(1991), no. 2, 365421.
[L2] G. Lusztig, Semicanonical bases arising from enveloping algebras, Adv. Math. 151 (2000), no. 2,
129139.
[M] Y, Mizuno, -stable -tilting modules, arXiv:1210.8322.
[ORT] S. Oppermann, I. Reiten, H. Thomas, Quotient closed subcategories of quiver representations,
arXiv:1205.3268.
[P] Y. Palu, Cluster characters for 2-Calabi-Yau triangulated categories, Ann. Inst. Fourier (Grenoble) 58
(2008), no. 6, 22212248.
[Re] N. Reading, Clusters, Coxeter-sortable elements and noncrossing partitions, Trans. Amer. Math. Soc.,
359 (2007), no. 12, 59315958.
[Ri] C. M. Ringel, The preprojective algebra of a quiver, Algebras and modules, II (Geiranger, 1996),
467480, CMS Conf. Proc., 24, Amer. Math. Soc., Providence, RI, 1998.
[S] S. O. Smal, Torsion theory and tilting modules, Bull. London Math. Soc. 16(1984), 518522.
Graduate School of Mathematics, Nagoya University, Frocho, Chikusaku, Nagoya, 464-
8602, Japan
E-mail address: yuya.mizuno@math.nagoya-u.ac.jp

Anda mungkin juga menyukai