Journal of Pure and Applied Algebra 167 (2002) 15–35

www.elsevier.com/locate/jpaa

On t-structures and torsion theories induced by

compact objects
Mitsuo Hoshinoa , Yoshiaki Katoa , Jun-Ichi Miyachib; ∗
a Institute
of Mathematics, University of Tsukuba, Ibaraki 305-8571, Japan
b Department of Mathematics, Tokyo Gakugei University, Koganei-shi, Tokyo 184-8501, Japan

Received 22 May 2000; received in revised form 13 November 2000

Communicated by C.A. Weibel

Abstract
First, we show that a compact object C in a triangulated category, which satis5es suitable
conditions, induces a t-structure. Second, in an abelian category we show that a complex P · of
small projective objects of term length two, which satis5es suitable conditions, induces a torsion
theory. In the case of module categories, using a torsion theory, we give equivalent conditions
for P · to be a tilting complex. Finally, in the case of artin algebras, we give a one-to-one
correspondence between tilting complexes of term length two and torsion theories with certain
conditions.
c 2002 Elsevier Science B.V. All rights reserved.

MSC: Primary: 18E30, 16S90; secondary: 18E40, 16G99

0. Introduction

In the representation theory of 5nite dimensional algebras, torsion theories were stud-
ied by several authors in connection with classical tilting modules. For these torsion
theories, there are equivalences between torsion (resp., torsion-free) classes and torsion-
free (resp., torsion) classes, which is known as Theorem of Brenner and Butler ([5]).
One of the authors gave a one-to-one correspondence between classical tilting modules
and torsion theories with certain conditions [7,8]. But in the case of a self-injective
algebra A, tilting modules are essentially isomorphic to A. In [12], Rickard introduced
the notion of tilting complexes as a generalization of tilting modules, and showed that

∗Corresponding author.
E-mail addresses: [email protected] (M. Hoshino), [email protected] (Y. Kato),
[email protected] (J.-I. Miyachi).

0022-4049/02/$ - see front matter  c 2002 Elsevier Science B.V. All rights reserved.
PII: S 0 0 2 2 - 4 0 4 9 ( 0 1 ) 0 0 0 1 2 - 3
16 M. Hoshino et al. / Journal of Pure and Applied Algebra 167 (2002) 15–35

these complexes induce equivalences between derived categories of module categories.

Tilting complexes of term length two are often studied in the case of self-injective
algebras (e.g. [6,9]). On the other hand, for triangulated categories, Beilinson et al. [2]
introduced the notions of t-structures and admissible abelian subcategories, and studied
relationships between them. In this paper, we 5rst deal with a compact object C in
a triangulated category, and study a t-structure induced by C. Second, in an abelian
category A we deal with a complex P · of small projective objects of term length two
and study a torsion theory induced by P ·.
In Section 1, we show that a compact object C in a triangulated category T, which
satis5es suitable conditions, induces a t-structure (T60 (C); T¿0 (C)), and its core
T0 (C) is equivalent to the category Mod B of left B-modules, where B = End T (C)op
(Theorem 1.3). In Section 2, we de5ne subcategories X(P ·), Y(P ·) of an abelian
category A satisfying the condition Ab4, and show when (X(P ·); Y(P ·)) is a tor-
sion theory (Theorem 2.10). Furthermore, we show that if P · induces a torsion the-
ory (X(P ·); Y(P ·)) for A, then the core D(A)0 (P ·) is admissible abelian, and then
there is a torsion theory (Y(P ·)[1]; X(P ·)) for D(A)0 (P ·) (Theorem 2.15). In Sec-
tion 3, we apply results of Section 2 to module categories. We characterize a torsion
theory for the category Mod A of left A-modules, and for its core D(Mod A)0 (P ·)
(Theorems 3.5 and 3.8). Furthermore, using a torsion theory, we give equivalent
conditions for P · to be a tilting complex (Corollary 3.6). In Section 4, we show
that if P · is a tilting complex, then it induces equivalences between torsion the-
ories for Mod A and for Mod B, where B = End D(Mod A) (P ·)op (Theorem 4.4). In
Section 5, in the case of artin algebras, if a torsion theory (X; Y) satis5es certain
conditions, then there exists a tilting complex P · of term length two such that a tor-
sion theory (X; Y) coincides with (X(P ·); Y(P ·)) (Theorem 5.8). As a consequence,
we have a one-to-one correspondence between tilting complexes of term length two
and torsion theories with certain conditions (Corollary 3.7, Propositions 5.5, 5.7 and
Theorem 5.8).

1. t-structures induced by compact objects

In this section, we deal with a triangulated category T and its full subcategory C.
We will call C admissible abelian provided that HomT (C; C[n])=0 for n ¡ 0, and that
all morphisms in C are C-admissible in the sense of [2, 1.2.3]. In this case, according
to [2, Proposition 1:2:4] C is an abelian category. A triangulated category T is said to
contain direct sums if direct sums of objects indexed by any set exist in T. An object
C of T is called compact if HomT (C; −) commutes with direct sums. Furthermore, a
collection S of compact objects of T is called a generating set provided that X = 0
whenever HomT (S; X )=0, and that S is stable under suspension (see [10] for details).
For an object C ∈ T and an integer n, we denote by T¿n (C) (resp., T6n (C)) the
full subcategory of T consisting of X ∈ T with HomT (C; X [i]) = 0 for i ¡ n (resp.,
i ¿ n), and set T0 (C) = T60 (C) ∩ T¿0 (C).
 M. Hoshino et al. / Journal of Pure and Applied Algebra 167 (2002) 15–35 17

For an abelian category A, we denote by C(A) the category of complexes of

A, and denote by D(A) (resp., D+ (A), D− (A), Db (A)) the derived category of
complexes of A (resp., complexes of A with bounded below homologies, complexes
of A with bounded above homologies, complexes of A with bounded homologies).
For an additive category B, we denote by K(B) (resp., K− (B), Kb (B)) the homotopy
category of complexes of B (resp., bounded above complexes of B, bounded complexes
of B) (see [11] for details).

Proposition 1.1. Let T be a triangulated category which contains direct sums; C a

compact object satisfying HomT (C; C[n]) = 0 for n ¿ 0. Then for any r ∈ Z and any
object X ∈ T; there exists an object X ¿r ∈ T¿r (C) and a morphism ¿r : X → X ¿r
in T such that
(i) for any i ¿ r; HomT (C; ¿r [i]) is an isomorphism;
(ii) for every object Y ∈ T¿r (C); HomT (¿r ; Y ) is an isomorphism.

Proof. Let X0 = X . For n ¿ 1, by induction we construct a distinguished triangle

gn n h
Cn [n − r]−→Xn−1 −→X n →

as follows. If HomT (C; Xn−1 [r − n]) = 0, then we set Cn = 0. Otherwise, we take

a direct sum Cn of copies of C and a morphism gn : Cn → Xn−1 [r − n] such that
HomT (C; gn ) is an epimorphism, and let gn = gn [n − r]. Then, by easy calculation, we
have the following:
(a) HomT (C; Xn [i]) = 0 for r − n 6 i ¡ r,
(b) HomT (C; hn [i]) is an isomorphism for any n and i ¿ r.
Let X ¿r be a homotopy colimit hocolim Xn and ¿r : X → X ¿r a structural mor-
−−−−−→
phism X0 → hocolim Xn . According to [10, Lemma 2:8] the conditions (a), (b) imply
−−−−−→
that X ¿r belongs to T¿r (C) and satisfy the statement (i). For an object Y ∈ T¿r (C),
we have an exact sequence
HomT (Cn [n − r]; Y [j − 1]) → HomT (Xn ; Y [j]) →
HomT (Xn−1 ; Y [j]) → HomT (Cn [n − r]; Y [j]):
Since HomT (C[i]; Y [j]) = 0 for j − i ¡ r, HomT (hn ; Y [j]) is an isomorphism for any
n ¿ 1 and j 6 0. Then, we have an epimorphism

1-shift


HomT (Xn ; Y [j]) −−→ HomT (Xn ; Y [j])
n n
for any j 6 0. Therefore, we have an exact sequence

1-shift


0 → HomT (X ¿r ; Y ) → HomT (Xn ; Y ) −−→ HomT (Xn ; Y ) → 0:
n n
Hence we have

HomT (X ¿r ; Y ) ∼
= lim HomT (Xn ; Y )
←−
∼
= HomT (X; Y ):
18 M. Hoshino et al. / Journal of Pure and Applied Algebra 167 (2002) 15–35

Denition 1.2 (Beilinson et al. [2]). Let T be a triangulated category. For full sub-
categories T60 and T¿0 , (T60 ; T¿0 ) is called a t-structure on T provided that
(i) HomT (T60 ; T¿1 ) = 0,
(ii) T60 ⊂ T61 and T¿0 ⊃ T¿1 ,
(iii) for any X ∈ T, there exists a distinguished triangle

X  → X → X  →
with X  ∈ T60 and X  ∈ T¿1 ,
where T6n = T60 [ − n] and T¿n = T¿0 [ − n].

A t-structure (T60 ; T¿0 ) on T is called non-degenerate if n∈Z T
6n
=
 ¿n
n∈Z T = {0}.

Theorem 1.3. Let T be a triangulated category which contains direct sums; C a com-
pact object satisfying HomT (C; C[n]) = 0 for n ¿ 0; and B = End T (C)op . If {C[i] | i ∈
Z} is a generating set; then the following hold:
(1) (T60 (C); T¿0 (C)) is a non-degenerate t-structure on T.
(2) T0 (C) is admissible abelian.
(3) The functor

HomT (C; −) : T0 (C) → Mod B

is an equivalence.

Proof. (1) For any object X ∈ T60 (C), we take an object X ¿1 ∈ T¿1 (C) and a
morphism ¿1 : X → X ¿1 satisfying the conditions of Proposition 1.1. Then for any
Y ∈ T¿1 (C), by Proposition 1.1(ii), we have

HomT (X ¿1 ; Y ) ∼
= HomT (X; Y ):
By Proposition 1.1(i), X ∈ T60 (C) implies that HomT (C; X ¿1 [i]) = 0 for all i ∈ Z.
Since {C[i] | i ∈ Z} is a generating set, we have X ¿1 = 0, and hence HomT (X; Y ) = 0.
It is easy to see that T60 (C) ⊂ T61 (C) and T¿0 (C) ⊃ T¿1 (C). For any object
Z ∈ T, we take an object Z ¿1 ∈ T¿1 (C) and a morphism ¿1 : Z → Z ¿1 satisfying
the conditions of Proposition 1.1, and embed ¿1 in a distinguished triangle

Z  → Z → Z ¿1 → :
Applying HomT (C; −) to the above triangle, by Proposition 1.1(i), we have Z  ∈
T60 (C). Since {C[i] | i ∈ Z} is a generating set, it is easy to see that (T60 (C);
T¿0 (C)) is non-degenerate.
(2) Since T0 (C) is the core of the t-structure (T60 (C); T¿0 (C)), the assertion
follows by [2, Theorem 1:3:6].
(3) Step 1: According to [2, Proposition 1:2:2], the short exact sequences in T0 (C)
are just the distinguished triangles

X →Y →Z →
 M. Hoshino et al. / Journal of Pure and Applied Algebra 167 (2002) 15–35 19

with X; Y and Z belonging to T0 (C). It follows that HomT (C; −) : T0 (C) → Mod B
is exact. Let M ∈ Mod B and take a free presentation P1 → P0 → M → 0. We take
C  = C ¿0 ∈ T0 (C) and  = ¿0 : C → C  satisfying the conditions of Proposition 1.1.
Then there exist sets I; J and a collection of morphisms hij : C  → C  such that
P1 −−−−−−−−−−−−−−→ P0
 
  
Hom(C;hij )
HomT (C; C  )(J ) −−−−−−−→ HomT (C; C  )(I )
ij

is commutative, where the vertical arrows are isomorphisms. We take an exact sequence
in T0 (C)

hij
C (J ) −−→C (I ) → X → 0:
ij

Since C is compact, by the exactness of HomT (C; −), we have HomT (C; X ) ∼
= M.
Step 2: We show that HomT (C; −) reJects isomorphisms. Let
u
X →Y → Z →
be a distinguished triangle in T with X; Y ∈ T0 (C) and with HomT (C; u) an isomor-
phism. Then, by applying HomT (C; −), we get HomT (C; Z[n]) = 0 for all n ∈ Z, and
hence Z = 0. It follows that u is an isomorphism.
Next, we show that HomT (C; −) is faithful. Let v : X → Y be a morphism in
T0 (C) with HomT (C; v) = 0. By the exactness of HomT (C; −), HomT (C; Im v) ∼ =
Im HomT (C; v) = 0. Since Im v ∈ T0 (C), we have HomT (C; Im v[n]) = 0 for all
n ∈ Z, and hence Im v = 0 and v = 0.
Let M be the full subcategory of T0 (C) consisting of objects X such that there
exists an exact sequence C1 → C0 → X → 0 in T0 (C), where C0 ; C1 are direct sums
of copies of C  . Since HomT (C; −) is faithful, by the same technique as in Step 1,
it is not hard to see that HomT (C; −)|M is full dense, and hence an equivalence. It
remains to show that M = T0 (C). For an object X ∈ T0 (C), we have a commutative
diagram
(J ) Hom(C;f) Hom(C;g)
HomT (C; (C; C (I ) ) −−−−→ HomT (C; X ) −−→ 0
 C ) −−−−→ HomT
 
 
 HomT (C;1 )  HomT (C;0 )
 
Hom(C;f
)
HomT (C; C (J ) ) −−−−−→ HomT (C; C (I ) )

with the top row being exact and with the vertical arrows being isomorphisms. And
we have a commutative diagram in T
f g
C(J ) −−→ C(I ) −−→ X
 
 1  0
f

C (J ) −−→ C (I )
with gf = 0. By Proposition 1.1(ii), there exists h : C (I ) → X such that g = h0 . Since
HomT (C; hf ) = 0, we have hf = 0. Then there exists w : Cok f → X such that
20 M. Hoshino et al. / Journal of Pure and Applied Algebra 167 (2002) 15–35

g = wg 0 , where g : C (I ) → Cok f is a canonical morphism. Then HomT (C; w) is

an isomorphism, and hence w is an isomorphism and X ∼ = Cokf ∈ M.

Remark 1.4. Under the condition of Theorem 1.3, according to [2, Proposition 1:3:3]
there exists a functor (−)¿n : T → T¿n (C) (resp., (−)6n : T → T6n (C)) which is
the right (resp., left) adjoint of the natural embedding functor T¿n (C) → T (resp.,
T6n (C) → T).

For an object C in a triangulated category T and integers s 6 t, let T[s] (C) =

 
T0 (C)[ − s], T[s; t] (C) = T6t (C) ∩ T¿s (C), and Tb (C) = ( n∈Z T6n (C)) ∩ ( n∈Z
T¿n (C)). An object M of an abelian category A is called small provided that
HomA (M; −) commutes with direct sums in A.

Corollary 1.5. Let A be an abelian category satisfying the condition Ab4 (i.e. direct
sums of exact sequences are exact) and T · a bounded complex of small projective
objects of A satisfying
(i) {T ·[i] | i ∈ Z} is a generating set for D(A);
(ii) HomD(A) (T ·; T ·[i]) = 0 for i = 0.
If either of the following conditions (1) or (2) is satis;ed; then we have an equivalence
of triangulated categories

D(A)b (T ·) ∼
= Db (Mod B);
where B = End D(A) (T ·)op .
(1) A has enough projectives.
(2) A has enough injectives and D(A)¿0 (T ·) ⊂ D+ (A).
Moreover; if D(A)0 (T ·) ⊂ Db (A); then we have an equivalence

Db (A) ∼
= Db (Mod B):

Proof. According to [3, Corollary 1:7] D(A) contains direct sums. Since T · is a
bounded complex of small projective objects of A, T · is a compact object in D(A). By
Theorem 1.3 D(A) has a t-structure (D(A)60 (T ·); D(A)¿0 (T ·)), and
HomD(A) (T ·; −) : D(A)0 (T ·) → Mod B is an equivalence.
(1) By the construction of X ¿r in Proposition 1.1, D− (A) also has a t-structure
(D− (A)60 (T ·); D− (A)¿0 (T ·)) and hence by Theorem 1.3(3) we have D− (A)0
(T ·)=D(A)0 (T ·). According to [12, Proposition 10:1] we have a fully faithful @-functor
F  : D− (Mod B) → D− (A). Also, since F  (B) = T ·, F  sends B-modules to objects in
D(A)0 (T ·). Then we have a fully faithful @-functor

F : Db (Mod B) → D(A);

which sends B-modules to objects in D(A)0 (T ·). For any X ∈ D(A)b (T ·), there exist
integers m 6 n such that X ∈ D(A)[m; n] (T ·). Let l = n − m. If l = 0, then there exist,
 M. Hoshino et al. / Journal of Pure and Applied Algebra 167 (2002) 15–35 21

obviously, a B-module M and an integer s such that X ∼

= F(M [s]). If l ¿ 0, then we
have a distinguished triangle

X 6n−1 → X → X ¿n →
with X ¿n ∈ D(A)[n] (T ·) and X 6n−1 ∈ D(A)[m; n−1] (T ·). Since F is full, by induction
on l, there exists U ·∈ Db (Mod B) such that X ∼= F(U ·).
(2) By the assumption, D (A) also has a t-structure (D+ (A)60 (T ·); D+ (A)¿0
+

(T ·)). Thus D+ (A)b (T ·) = D(A)b (T ·), and hence D+ (A)0 (T ·) = D(A)0 (T ·). By [2,
Section 3] we have a @-functor real : Db (D(A)0 (T ·)) → D+ (A), and then we have
a @-functor

F : Db (Mod B) → D(A);
which sends B-modules to objects in D(A)0 (T ·). Let f ∈ HomD(A) (X ·; Y ·[n]) with
X ·; Y · ∈ D(A)0 (T ·) and n ¿ 0. Take a distinguished triangle in D+ (A)

X1· → V ·→ X · →
t

such that V · is a direct sum of copies of T · and HomD(A) (T ·; t) is an epimorphism. By

easy calculation, X1· ∈ D(A)0 (T ·), and hence we get an exact sequence in D(A)0 (T ·)

0 → X1· → V ·→ X · → 0:
t

Since HomD(A) (T ·; Y ·[n]) = 0, we have ft = 0, i.e. t eLaces f. Thus the epimorphic

version of eLacibility in [2, Proposition 3:1:16] can be applied.
Finally, it is easy to see that D(A)0 (T ·) ⊂ Db (A) implies Db (A) = D(A)b (T ·).

2. Torsion theories for abelian categories

Throughout this section, we 5x the following notation. Let A be an abelian category

satisfying the condition Ab4, and let d−1
P : P
−1
→ P 0 be a morphism in A with the
P i being small projective objects of A, and denote by P · the mapping cone of d−1 P .
We set C(P ·) = D(A)0 (P ·), B = End D(A) (P ·)op , and de5ne a pair of full subcategories
of A

X(P ·) = {X ∈ A | HomD(A) (P ·; X [1]) = 0};

Y(P ·) = {X ∈ A | HomD(A) (P ·; X ) = 0}:

For any X ∈ A, we de5ne a subobject of X

$(X ) = Im f
f∈HomA (H0 (P · );X )

and an exact sequence in A

jX
(eX ) : 0 → $(X ) → X → &(X ) → 0:
22 M. Hoshino et al. / Journal of Pure and Applied Algebra 167 (2002) 15–35

Remark 2.1. It is easy to see that P · is a compact object of D(A), and we have
X(P ·) = D(A)60 (P ·) ∩ A and Y(P ·) = D(A)¿1 (P ·) ∩ A.

Lemma 2.2. For any X ∈ A; the following hold:

(1) Ker(HomA (d−1 ∼ ·
P ; X )) = Hom D(A) (P ; X ).
(2) Cok(HomA (dP ; X )) = HomD(A) (P ·; X [1]).
−1 ∼

Lemma 2.3. For any X ∈ A; the following hold:

(1) HomD(A) (P ·; X [n]) = 0 for n ¿ 1 and n ¡ 0.
(2) HomD(A) (P ·; X ) ∼
= HomA (H0 (P ·); X ).

Lemma 2.4. The following hold:

(1) X(P ·) is closed under factor objects and direct sums.
(2) Y(P ·) is closed under subobjects.
(3) For any X ∈ A; HomA (H0 (P ·); jX ) is an isomorphism.

Lemma 2.5. For any X · ∈ D(A) and n ∈ Z; we have a functorial exact sequence

0 → HomD(A) (P ·; Hn−1 (X ·)[1]) → HomD(A) (P ·; X ·[n])

→ HomD(A) (P ·; Hn (X ·)) → 0:
Moreover; the above short exact sequence commutes with direct sums.

Proof. For X ·[n] ∈ D(A), applying HomD(A) (−; X ·[n]) to a distinguished triangle
d−1
P −1 →
P
P 0 → P · →;
we have a short exact sequence

0 → Cok(HomD(A) (d−1 · · ·
P ; X [n − 1])) → Hom D(A) (P ; X [n])

→ Ker(HomD(A) (d−1 ·
P ; X [n])) → 0:

Also, by Lemma 2.2 we get

Ker(HomD(A) (d−1 · ∼ −1 n ·
P ; X [n])) = Ker(Hom A (dP ; H (X )))

= HomD(A) (P ·; Hn (X ·));
∼

Cok(HomD(A) (d−1 · ∼ −1 n−1

(X ·)))
P ; X [n − 1])) = Cok(Hom A (dP ; H

= HomD(A) (P ·; Hn−1 (X ·)[1]):

∼

Since the P i are small objects, the above short exact sequence commutes with direct
sums.
 M. Hoshino et al. / Journal of Pure and Applied Algebra 167 (2002) 15–35 23

Lemma 2.6. The following are equivalent:

(1) {P ·[i] | i ∈ Z} is a generating set for D(A).
(2) X(P ·) ∩ Y(P ·) = {0}.

Proof. (1) ⇒ (2). For any X ∈ X(P ·) ∩ Y(P ·), by Lemma 2.3(1), HomD(A) (P ·;
X [n]) = 0 for all n ∈ Z and hence X = 0.
(2) ⇒ (1). Let X · ∈ D(A) with HomD(A) (P ·; X ·[n]) = 0 for all n ∈ Z. Then by
Lemma 2.5, Hn (X ·) ∈ X(P ·) ∩ Y(P ·) = {0}.

Lemma 2.7. The following hold:

(1) H0 (P ·) ∈ X(P ·) if and only if HomD(A) (P ·; P ·[i]) = 0 for all i ¿ 0.
(2) H−1 (P ·) ∈ Y(P ·) if and only if HomD(A) (P ·; P ·[i]) = 0 for all i ¡ 0.

Proof. By Lemma 2.5.

Denition 2.8. A pair (X; Y) of full subcategories X; Y in an abelian category A is

called a torsion theory for A provided that the following conditions are satis5ed (see
e.g. [4] for details):
(i) X ∩ Y = {0};
(ii) X is closed under factor objects;
(iii) Y is closed under subobjects;
(iv) For any object X of A, there exists an exact sequence 0 → X  → X → X  → 0
in A with X  ∈ X and X  ∈ Y.

Remark 2.9. Let A be an abelian category and (X; Y) a torsion theory for A. Then
for any Z ∈ A, the following hold:
(1) Z ∈ X if and only if HomA (Z; Y) = 0.
(2) Z ∈ Y if and only if HomA (X; Z) = 0.

d−1
Theorem 2.10. The following are equivalent for a complex P · : P −1 −→
P
P 0 with the
i
P being small projective objects of A:
(1) {P ·[i] | i ∈ Z} is a generating set for D(A) and HomD(A) (P ·; P ·[i]) = 0 for all
i ¿ 0.
(2) X(P ·) ∩ Y(P ·) = {0} and H0 (P ·) ∈ X(P ·).
(3) X(P ·) ∩ Y(P ·) = {0} and $(X ) ∈ X(P ·); &(X ) ∈ Y(P ·) for all X ∈ A.
(4) (X(P ·); Y(P ·)) is a torsion theory for A.

Proof. (1) ⇔ (2). By Lemmas 2:6 and 2:7(1).

(2) ⇒ (3). Let X ∈ A. Since H0 (P ·) ∈ X(P ·), it follows that $(X ) ∈ X(P ·).
Next, apply HomD(A) (P ·; −) to the canonical exact sequence (eX ). It then follows by
Lemmas 2:3(2) and 2:4(3) that HomD(A) (P ·; jX ) is an isomorphism. Thus HomD(A)
(P ·; &(X )) = 0 and hence &(X ) ∈ Y(P ·).
24 M. Hoshino et al. / Journal of Pure and Applied Algebra 167 (2002) 15–35

(3) ⇒ (4). Obvious.

(4) ⇒ (1). By Lemmas 2:3(2), 2:6, 2:7(1) and Remark 2.9(1).

Denition 2.11. For a complex X · = (X i ; di ), we de5ne the following truncations:

'¿n (X ·) : : : : → 0 → Im dn → X n+1 → X n+2 → : : : ;

'6n (X ·) : : : : → X n−2 → X n−1 → Ker dn → 0 → : : : ;


'¿n (X ·) : : : : → 0 → Cok dn−1 → X n+1 → X n+2 → : : : ;

'¡n (X ·) : : : : → X n−2 → X n−1 → Imdn−1 → 0 → : : : :

Lemma 2.12. For any X · ∈ D(A) with Hn (X ·) = 0 for n ¿ 0 and n ¡ − 1; there

exists a distinguished triangle in D(A) of the form

H−1 (X ·)[1] → X · → H0 (X ·) → :

Proof. We have exact sequences in C(A)

0 → '6−1 (X ·) → X · → '¿−1 (X ·) → 0;

0 → '¡0 ('¿−1 (X ·)) → '¿−1 (X ·) → '¿0

(X ·) → 0:

Also, '6−1 (X ·) ∼
= H−1 (X ·)[1]; '¡0

('¿−1 (X ·)) ∼ 
= 0 and '¿0 (X ·) ∼
= H0 (X ·) in
D(A). Thus we get a desired distinguished triangle in D(A).

Lemma 2.13. Assume X(P ·) ∩ Y(P ·) = {0}. Then for any X · ∈ D(A); the following
are equivalent:
(1) X · ∈ C(P ·).
(2) Hn (X ·) = 0 for n ¿ 0 and n ¡ − 1; H0 (X ·) ∈ X(P ·) and H−1 (X ·) ∈ Y(P ·).

Proof. By Lemma 2.5.

Remark 2.14. Let A be an abelian category and X; Y full subcategories of A. Then

the pair (X; Y) is a torsion theory for A if and only if the following two conditions
are satis5ed:
(i) HomA (X; Y) = 0;
(ii) for any object X in A, there exists an exact sequence 0 → X  → X → X  → 0
in A with X  ∈ X and X  ∈ Y.
−1
dP
Theorem 2.15. Let P · be a complex P −1 → P 0 with the P i being small projective
· ·
objects of A. Assume X(P ) ∩ Y(P ) = {0} and H0 (P ·) ∈ X(P ·). Then the following
hold:
(1) C(P ·) is admissible abelian.
 M. Hoshino et al. / Journal of Pure and Applied Algebra 167 (2002) 15–35 25

(2) The functor

HomD(A) (P ·; −) : C(P ·) → Mod B

is an equivalence.
(3) (Y(P ·)[1]; X(P ·)) is a torsion theory for C(P ·).

Proof. (1) and (2) According to Theorem 2.10, Theorem 1.3 can be applied.
(3) Note 5rst that by Lemma 2.13 we have X(P ·) ⊂ C(P ·) and Y(P ·)[1] ⊂
C(P ·). Also, it is trivial that HomD(A) (Y(P ·)[1]; X(P ·)) = 0. Let X · ∈ C(P ·). Then
by Lemmas 2.12 and 2.13 we have a distinguished triangle in D(A) of the form

H−1 (X ·)[1] → X · → H0 (X ·) → :

It follows that the sequence in C(P ·)

0 → H−1 (X ·)[1] → X · → H0 (X ·) → 0

is exact. Thus by Remark 2.14 (Y(P ·)[1]; X(P ·)) is a torsion theory for
C(P ·).

Proposition 2.16. Assume P · satis;es the conditions

(i) {P ·[i] | i ∈ Z} is a generating set for D(A);
(ii) HomD(A) (P ·; P ·[i]) = 0 for i = 0.
If A has either enough projectives or enough injectives; then we have an equivalence
of triangulated categories

Db (A) ∼
= Db (Mod B):

Proof. Let X · ∈ D(A). According to Lemma 2.5 and Theorem 2.10, it is easy to see
that if X · belongs to D(A)¿0 (P ·) (resp., C(P ·)), then Hn (X ·) = 0 for n ¡ − 1 (resp.,
n ¡ − 1 and n ¿ 0). Thus we have

D(A)¿0 (P ·) ⊂ D+ (A) and C(P ·) ⊂ Db (A);

so that Corollary 1.5 can be applied.

3. Torsion theories for module categories

In this section, we apply results of Section 2 to the case of module categories. In

and after this section, R is a commutative ring and I is an injective cogenerator in the
category of R-modules. We set D = HomR (−; I ). Let A be an R-algebra and denote by
Proj A (resp., proj A) the full additive subcategory of Mod A consisting of projective
(resp., 5nitely generated projective) modules. We denote by Aop the opposite ring of
A and consider right A-modules as left Aop -modules. Also, we denote by (−)∗ both
26 M. Hoshino et al. / Journal of Pure and Applied Algebra 167 (2002) 15–35

the A-dual functors HomA (−; A) and set * = D ◦ (−)∗ . Throughtout this section, P · is
d−1
a complex P −1 →P
P 0 with the P i being 5nitely generated projective A-modules.
It is well known that, in a module category, the small projective objects are just the
5nitely generated projective modules. In the following, we deal with the case where
A = Mod A and use the same notation as in Section 2.

Lemma 3.1. For any X ∈ Mod A; we have

HomD(Mod A) (P ·; X [1]) ∼
= H1 ((P ·)∗ ) ⊗A X:

Proof. We have

HomD(Mod A) (P ·; X [1]) ∼
= HomK(Mod A) (P ·; X [1])

= H1 (HomA· (P ·; X ))
∼

= H1 ((P ·)∗ ⊗A· X )

∼

= H1 ((P ·)∗ ) ⊗A X:
∼

Lemma 3.2. The following hold:

(1) X(P ·) = Ker(H1 ((P ·)∗ ) ⊗A −).
(2) Y(P ·) = Ker(HomA (H0 (P ·); −)).

Proof. By Lemmas 2:3(2) and 3:1.

Lemma 3.3. The following hold:

(1) D(H1 ((P ·)∗ )) ∼= H−1 (*(P ·)).
·
(2) X(P ) = Ker(HomA (−; H−1 (*(P ·)))) and hence H0 (P ·) ∈ X(P ·) if and only if
H−1 (*(P ·)) ∈ Y(P ·).
(3) Ker(Tor A1 (H1 ((P ·)∗ ); −)) = Ker(Ext 1A (−; H−1 (*(P ·)))).

Proof. We have D(H1 ((P ·)∗ )) ∼

= H−1 (D((P ·)∗ ))=H−1 (*(P ·)) and for any X ∈ Mod A
we have

D(H1 ((P ·)∗ ) ⊗A X ) ∼

= HomA (X; H−1 (*(P ·)));

D(Tor A1 ((H1 ((P ·)∗ ); X ))) ∼

= Ext1A (X; H−1 (*(P ·))):

Lemma 3.4. The following hold:

(1) X(P ·) ⊂ Ker(Ext1A (H0 (P ·); −)).
(2) Y(P ·) ⊂ Ker(Tor A1 (H1 ((P ·)∗ ); −)).
 M. Hoshino et al. / Journal of Pure and Applied Algebra 167 (2002) 15–35 27

Proof. This is due essentially to Auslander [1]. We have an exact sequence in Mod A

0 → H−1 (P ·) → P −1 → P 0 → H0 (P ·) → 0

with the P i 5nitely generated projective, and an exact sequence in Mod Aop

0 → H0 (P ·)∗ → P 0∗ → P −1∗ → H1 ((P ·)∗ ) → 0

with the P i∗ 5nitely generated projective.

(1) Let X ∈ Mod A. For any M ∈ Mod Aop , we have a functorial homomorphism

+M : M ⊗A X → HomA (M ∗ ; X ); m ⊗ x → (h → h(m)x);

which is an isomorphism if M is 5nitely generated projective. Since the P i are reJexive,

we have H0 (P ·) ∼
= H0 ((P ·)∗∗ ) and H−1 (P ·) ∼
= H1 ((P ·)∗ )∗ . We have a commutative
diagram

P 0∗ 
⊗A X −−→ P −1∗⊗A X −−→H1 ((P ·)∗ ) ⊗A X −−→ 0
  
+P0∗   +P−1∗ 
HomA (P 0∗∗ ; X ) −−→ HomA (P −1∗∗ ; X ) −−→ HomA (H−1 (P ·); X )

with the top row exact. Since the +Pi∗ are isomorphisms, Ext 1A (H0 (P ·); X ) is embedded
in H1 ((P ·)∗ ) ⊗A X . The assertion follows by Lemma 3.2.
(2) Let X ∈ Mod A. For any Y ∈ Mod A, we have a functorial homomorphism

-Y : Y ∗ ⊗A X → HomA (Y; X ); h ⊗ x → (y → h(y)x);

which is an isomorphism if Y is 5nitely generated projective. We have a commutative

diagram

H0 (P ·
)∗ ⊗A X −−→ P 0∗ ⊗A X −−→ P −1∗⊗A X
  
  -P 0  -P−1
0 ·
0 −−→ HomA (H (P ); X ) −−→ HomA (P ; X ) −−→ HomA (P −1 ; X )
0

with the bottom row exact. Since the -Pi are isomorphisms, Tor A1 (H1 ((P ·)∗ ); X ) is a
homomorphic image of HomA (H0 (P ·); X ). The assertion follows by Lemma 3.2.

d−1
Theorem 3.5. The following are equivalent for a complex P · : P −1 →
P
P 0 with the P i
being ;nitely generated projective A-modules:
(1) X(P ·) ∩ Y(P ·) = {0} and H0 (P ·) ∈ X(P ·).
(2) X(P ·) ∩ Y(P ·) = {0} and $(X ) ∈ X(P ·); &(X ) ∈ Y(P ·) for all X ∈ Mod A.
(3) (X(P ·); Y(P ·)) is a torsion theory for Mod A.
(4) X(P ·) consists of the modules generated by H0 (P ·) and Y(P ·) consists of the
modules cogenerated by H−1 (*(P ·)).
28 M. Hoshino et al. / Journal of Pure and Applied Algebra 167 (2002) 15–35

Proof. (1) ⇔ (2) ⇔ (3). By Theorem 2.10.

(3) ⇒ (4). Since HomA (H0 (P ·); −) vanishes on Y(P ·), H0 (P ·) ∈ X(P ·). Thus
X(P ·) contains the modules generated by H0 (P ·). Conversely, let X ∈ X(P ·). Then,
since (1) implies (2), &(X ) ∈ Y(P ·) and hence HomA (X; &(X )) = 0. Thus X =
$(X ), which is generated by H0 (P ·). Next, since by Lemma 3.3(2) H−1 (*(P ·)) ∈
Y(P ·), Y(P ·) contains the modules cogenerated by H−1 (*(P ·)). Conversely, let X ∈
Y(P ·). Take a set of generators {f/ }/∈0 for an R-module HomA (X; H−1 (*(P ·)))
and set

f : X → H−1 (*(P ·))0 ; x → (f/ (x))/∈0 :

It is obvious that HomA (f; H−1 (*(P ·))) is surjective. Also, by Lemmas 3:3(3) and
3:4(2) we have Ext 1A (Imf; H−1 (*(P ·))) = 0. Applying HomA (−; H−1 (*(P ·))) to the
canonical exact sequence

0 → Ker f → X → Im f → 0;

we get HomA (Ker f; H−1 (*(P ·)))=0. Thus Ker f ∈ X(P ·)∩Y(P ·) and hence Ker f =
0.
(4) ⇒ (1). By Lemma 3.3(2).

Corollary 3.6. The following are equivalent:

(1) P · is a tilting complex.
(2) X(P ·) ∩ Y(P ·) = {0}; H0 (P ·) ∈ X(P ·) and H−1 (P ·) ∈ Y(P ·).
(3) (X(P ·); Y(P ·)) is a torsion theory for Mod A and H−1 (P ·) ∈ Y(P ·).

Proof. By Lemmas 2:6, 2:7 and Theorem 3.5.

For an object X in an additive category B, we denote by add(X ) the full subcategory

of B consisting of objects which are direct summands of 5nite direct sums of copies
of X .

Corollary 3.7. For any tilting complexes P1· : P1−1 → P10 ; P2· : P2−1 → P20 for A of
term length two; the following are equivalent:
(1) (X(P1·); Y(P1·)) = (X(P2·); Y(P2·)).
(2) add(P1·) = add(P2·) in Kb (Proj A).

Proof. (1) ⇒ (2). It follows by Corollary 3.6 that Q · = P1· ⊕ P2· is a tilting complex
such that (X(Q ·); Y(Q ·))=(X(Pi·); Y(Pi·)) (i=1; 2). Let B=End D(Mod A) (Q ·)op and for
i=1; 2 denote by ei the composite of canonical homomorphisms Q · → Pi· → Q · . Then,
for i = 1; 2 we have an equivalence D− (Mod B) → D− (Mod ei Bei ) which sends Bei to
ei Bei , so that the Bei are tilting complexes for B, i.e. projective generators for Mod B.
It follows by Morita Theory that add B = add Bei in Mod B. Thus add(P1·) = add(P2·)
in Kb (Proj A).
 M. Hoshino et al. / Journal of Pure and Applied Algebra 167 (2002) 15–35 29

(2) ⇒ (1). It is obviously deduced that add(H−1 (*(P1·))) = add(H−1 (*(P2·))) and
add(H0 (P1·)) = add(H0 (P2·)).
−1
dP
Theorem 3.8. Let P · be a complex P −1 → P 0 with the P i being ;nitely generated
projective A-modules. Assume X(P ·) ∩ Y(P ·) = {0} and H0 (P ·) ∈ X(P ·). Then the
following hold:
(1) {P ·[i] | i ∈ Z} is a generating set for D(Mod A).
(2) C(P ·) is admissible abelian.
(3) (Y(P ·)[1]; X(P ·)) is a torsion theory for C(P ·).
(4) The functor
Hom D(Mod A) (P ·; −) : C(P ·) → Mod B
is an equivalence.

Proof. By Lemma 2.6 and Theorem 2.15.

Remark 3.9. The following are equivalent:

(1) P · is a tilting complex.
(2) X(P ·) ∩ Y(P ·) = {0} and P · ∈ C(P ·).

Example 3.10 (cf. Hoshino and Kato [9]). Let A be a 5nite dimensional algebra over
a 5eld k given by a quiver

1 −−→
 2

 
3 4
4 −−→ 3
5
with relations 4 = 54 = 35 = 3 = 0. For each vertex i, we denote by S(i); P(i)
the corresponding simple and indecomposable projective left A-modules, respectively.
De5ne a complex P · as the mapping cone of the homomorphism
f 0 0 0
d−1
P = : P(2)2 ⊕ P(4)2 → P(1) ⊕ P(3);
0 0 g 0
where f and g denote the right multiplications of  and 5, respectively. Then P · is
not a tilting complex. However, P · satis5es the assumption of Theorem 3.8 and hence
we have an equivalence of abelian categories
Hom (P ·; −) : C(P ·) → Mod B;
D(Mod A)

where B = End D(Mod A) (P ·)op is a 5nite dimensional k-algebra given by a quiver

1←2 3 ← 4:
There exist exact sequences in C(P ·) of the form
0 → S(1) → S(2)[1] → P(1)[1] → 0; 0 → S(3) → S(4)[1] → P(3)[1] → 0;
and these objects and morphisms generate C(P ·).
30 M. Hoshino et al. / Journal of Pure and Applied Algebra 167 (2002) 15–35

4. Equivalences between torsion theories

−1
dP
Throughout this section, P · : P −1 → P 0 is assumed to be a tilting complex. Then
there exists an equivalence of triangulated categories
F : D− (Mod B) → D− (Mod A)
such that F(B) = P ·. Let G : D− (Mod A) → D− (Mod B) be a quasi-inverse of F. For
any n ∈ Z, we have ring homomorphisms
B → End A (Hn (P ·))op and B → End A (Hn ((P ·)∗ )):
In particular, H0 (P ·) is an A-B-bimodule and H1 ((P ·)∗ ) is a B-A-bimodule.

Lemma 4.1. The following hold:

(1) For any X · ∈ C(P ·); we have G(X ·) ∼
= HomD(Mod A) (P ·; X ·).
(2) We have an equivalence
HomD(Mod A) (P ·; −) : C(P ·) → Mod B
whose quasi-inverse is given by the restriction of F to Mod B.

Proof. See [12, Section 4].

Lemma 4.2. There exists a tilting complex Q · ∈ Kb (proj B) such that

(i) Q· ∼= G(A);
(ii) Qi = 0 for i ¿ 1 and i ¡ 0;
(iii) Hi (Q ·) ∼
= Hi ((P ·)∗ ) for 0 6 i 6 1;
(iv) H (HomB· (Q ·; B)) ∼
i
= Hi (P ·) for −1 6 i 6 0.

Proof. By [12, Proposition 6:3], there exists Q · ∈ Kb (proj B) satisfying Q · ∼

= G(A).
Since
Hi (Q ·) ∼
= HomD(Mod B) (B; Q ·[i])

= HomD(Mod A) (P ·; A[i])
∼

= Hi ((P ·)∗ );
∼

we have Q · ∼
= '61 (Q ·) in Kb (proj B). Also, since
Hi (HomB· (Q ·; B)) ∼
= HomD(Mod B) (Q ·; B[i])

= HomD(Mod A) (A; P ·[i])

∼

= Hi (P ·);
∼

we have HomB· (Q ·; B) ∼
= '60 (HomB· (Q ·; B)) in Kb (proj Bop ) and Q · ∼ 
= '¿0 (Q ·) in
b i
K (proj B). Thus, we can assume Q = 0 for i ¿ 1 and i ¡ 0.
 M. Hoshino et al. / Journal of Pure and Applied Algebra 167 (2002) 15–35 31

Lemma 4.3. For any M ∈ Mod B; the following hold:

(1) Hi (F(M )) = 0 for i ¿ 0 and i ¡ − 1.
= H0 (P ·) ⊗B M
(2) H0 (F(M )) ∼
(3) H (F(M )) ∼
−1
= HomB (H1 ((P ·)∗ ); M ).

Proof. For any i ∈ Z, we have

Hi (F(M )) ∼
= HomD(Mod A) (A; F(M )[i])
= HomD(Mod B) (Q ·; M [i]):
∼

Thus Hi (F(M )) = 0 for i ¿ 0 and i ¡ − 1. Also,

= HomD(Mod B) (Q ·; M )
H0 (F(M )) ∼
= H0 (HomB· (Q ·; M ))
∼
= H0 (Hom · (Q ·; B) ⊗ M )
∼
B B

= H0 (HomB· (Q ·; B)) ⊗B M
∼
= H0 (P ·) ⊗B M;
∼

= HomD(Mod B) (Q ·; M [ − 1])
H−1 (F(M )) ∼
= H−1 (Hom · (Q ·; M ))
∼
B
= HomB (H1 (Q ·); M )
∼
= HomB (H 1 ((P ·)∗ ); M ):
∼

Theorem 4.4. De;ne a pair of full subcategories of Mod B

U(P ·) = Ker(H0 (P ·) ⊗B −); V(P ·) = Ker(HomB (H1 ((P ·)∗ ); −)):
Then, the following hold:
(1) (U(P ·); V(P ·)) is a torsion theory for Mod B.
(2) We have a pair of functors
HomA (H0 (P ·); −) : X(P ·) → V(P ·); H0 (P ·) ⊗B − : V(P ·) → X(P ·)
which de;ne an equivalence.
(3) We have a pair of functors
H1 ((P ·)∗ ) ⊗ − : Y(P ·) → U(P ·);
A HomB (H1 ((P ·)∗ ); −) : U(P ·) → Y(P ·)
which de;ne an equivalence.

Proof. (1) According to Lemmas 3.2 and 4.2, we can apply Corollary 3.6 for a tilting
complex Q · to conclude that (U(P ·); V(P ·)) is a torsion theory for Mod B.
(2) For any X ∈ X(P ·), by Lemmas 2:13; 4:1(1) and 4:3(3) we have
Hom (H1 ((P ·)∗ ); Hom (H0 (P ·); X )) ∼
B A = H−1 (F(G(X )))
∼
= H−1 (X )
= 0:
32 M. Hoshino et al. / Journal of Pure and Applied Algebra 167 (2002) 15–35

Also, since by Lemma 3.2(1) and Corollary 3.6 H1 ((P ·)∗ )⊗A H0 (P ·)=0, H1 ((P ·)∗ )⊗A
H0 (P ·) ⊗B M = 0 for all M ∈ V(P ·). The last assertion follows by Lemmas 2:13; 4:1
and 4:3.
(3) For any Y ∈ Y(P ·), by Lemmas 2:13; 3:1; 4:1(1) and 4:3(2) we have
H0 (P ·) ⊗B H1 ((P ·)∗ ) ⊗A Y ∼
= H0 (F(G(Y [1])))
∼
= H0 (Y [1])

= 0:
Also, since H1 ((P ·)∗ ) ⊗A H0 (P ·) = 0, for any N ∈ U(P ·) we have
HomA (H0 (P ·); HomB (H1 ((P ·)∗ ); N )) ∼
= HomB (H1 ((P ·)∗ ) ⊗A H0 (P ·); N )

= 0:
The last assertion follows by Lemmas 2:13; 4:1 and 4:3.

Denition 4.5. Let (U; V) be a torsion theory for an abelian category A. Then (U; V)
is called splitting if Ext 1A (V; U) = 0.

For a left A-module M , we denote by proj dimA M (resp., inj dimA M ) the projective
(resp., the injective) dimension of M .

Proposition 4.6. The torsion theory (U(P ·); V(P ·)) for Mod B is splitting if and
only if Ext2A (X(P ·); Y(P ·)) = 0. In particular; (U(P ·); V(P ·)) is splitting if either
proj dim X 6 1 for all X ∈ X(P ·) or inj dim Y 6 1 for all Y ∈ Y(P ·).

Proof. For any X ∈ X(P ·) and Y ∈ Y(P ·), we have

Ext 1B (HomA (H0 (P ·); X ); H1 ((P ·)∗ ) ⊗A Y ) ∼
= HomD(Mod B) (G(X ); G(Y [1])[1])
∼
= HomD(Mod A) (X; Y [2])
∼
= Ext2A (X; Y ):

5. Torsion theories for artin algebras

In this section, we deal with the case where R is a commutative artin ring, I is an
injective envelope of an R-module R=rad(R) and A is a 5nitely generated R-module. We
denote by mod A the full abelian subcategory of Mod A consisting of 5nitely generated
d−1
modules. Throughout this section, P · is also a complex P −1 → P
P 0 with the P i being
5nitely generated projective A-modules. Note that H (P ); H (*(P ·)) ∈ mod A for all
n · n

n ∈ Z. We set
Xc (P ·) = X(P ·) ∩ mod A and Yc (P ·) = Y(P ·) ∩ mod A:
 M. Hoshino et al. / Journal of Pure and Applied Algebra 167 (2002) 15–35 33

Proposition 5.1. The following are equivalent:

(1) Xc (P ·) ∩ Yc (P ·) = {0} and H0 (P ·) ∈ Xc (P ·).
(2) Xc (P ·) ∩ Yc (P ·) = {0} and $(X ) ∈ Xc (P ·); &(X ) ∈ Yc (P ·) for all X ∈ mod A.
(3) (Xc (P ·); Yc (P ·)) is a torsion theory for mod A.
(4) Xc (P ·) consists of the modules generated by H0 (P ·) and Yc (P ·) consists of the
modules cogenerated by H−1 (*(P ·)).

Proof. By the same arguments as in the proof of Theorem 3.5.

Lemma 5.2. The following are equivalent:

(1) {P ·[i] | i ∈ Z} is a generating set for D(mod A).
(2) Xc (P ·) ∩ Yc (P ·) = {0}.

Proof. By the same arguments as in the proof of Lemma 2.6.

Lemma 5.3. The following hold:

(1) If DA ∈ Xc (P ·); then H−1 (P ·) = 0; i.e. P · ∼
= H0 (P ·) in D(mod A).
(2) H (*(P )) ∈ Xc (P ) if and only if H (P ) ∈ Yc (P ·).
0 · · −1 ·

Proof. For any P ∈ proj A, we have functorial isomorphisms

*(P) ∼
= DA ⊗A P and P∼
= HomA (DA; *(P)):

Thus,

H0 (*(P ·)) ∼
= DA ⊗A H0 (P ·) and H−1 (P ·) ∼
= HomA (DA; H−1 (*(P ·)))

and hence,

HomA (H0 (*(P ·)); H−1 (*(P ·))) ∼

= HomA (DA ⊗A H0 (P ·); H−1 (*(P ·)))

= HomA (H0 (P ·); HomA (DA; H−1 (*(P ·))))

∼

= HomA (H0 (P ·); H−1 (P ·)):

∼

Lemma 5.4. Assume Xc (P ·)∩Yc (P ·)={0} and H0 (P ·) ∈ Xc (P ·). Then the following
are equivalent:
(1) H0 (*(P ·)) ∈ Xc (P ·).
(2) Xc (P ·) is stable under DA ⊗A −.
(3) H−1 (P ·) ∈ Yc (P ·).
(4) Yc (P ·) is stable under HomA (DA; −).

Proof. (1) ⇒ (2). Let X ∈ Xc (P ·). Then by Proposition 5.1 X is generated by H0 (P ·)

and hence DA ⊗A X is generated by DA ⊗A H0 (P ·) ∼= H0 (*(P ·)) ∈ Xc (P ·).
34 M. Hoshino et al. / Journal of Pure and Applied Algebra 167 (2002) 15–35

(2) ⇒ (3). Since H0 (*(P ·)) ∼

= DA ⊗A H0 (P ·) ∈ Xc (P ·), by Lemma 5.3(2) we have
−1 · ·
H (P ) ∈ Yc (P ).
(3) ⇒ (4) ⇒ (1). By the dual arguments.

Proposition 5.5. The following are equivalent:

(1) P · is a tilting complex.
(2) Xc (P ·) ∩ Yc (P ·) = {0}; H0 (P ·) ∈ Xc (P ·) and H−1 (P ·) ∈ Yc (P ·).
(3) (Xc (P ·); Yc (P ·)) is a torsion theory for mod A and H−1 (P ·) ∈ Yc (P ·).
(4) (Xc (P ·); Yc (P ·)) is a torsion theory for mod A and Xc (P ·) is stable under DA⊗A
−.
(5) (Xc (P ·); Yc (P ·)) is a torsion theory for mod A and Yc (P ·) is stable under
HomA (DA; −).

Proof. By Proposition 5.1, Lemmas 2:7; 5:2 and 5:4.

Denition 5.6. Let A be an abelian category and C a full subcategory of A closed

under extensions. Then an object X ∈ C is called Ext-projective (resp., Ext-injective)
if Ext 1A (X; C) = 0 (resp., Ext 1A (C; X ) = 0).

Proposition 5.7. Assume P · is a tilting complex. Then the following hold:

(1) H0 (P ·) ∈ Xc (P ·) is Ext-projective and generates Xc (P ·).
(2) H−1 (*(P ·)) ∈ Yc (P ·) is Ext-injective and cogenerates Yc (P ·).

Proof. By Propositions 5.1, 5.5 and Lemmas 3.3 and 3.4.

Theorem 5.8. Let (X; Y) be a torsion theory for mod A such that X contains an
Ext-projective module X which generates X; Y contains an Ext-injective module Y
which cogenerates Y; and X is stable under DA⊗A −. Let MX· be a minimal projective
presentation of X and N · a minimal injective presentation of Y . Then
Y

P · = M · ⊕ Hom · (DA; N ·)[1]

X A Y

is a tilting complex such that X = Xc (P ·) and Y = Yc (P ·).

Proof. According to Proposition 5.5, we have only to show that X = Xc (P ·) and

Y = Yc (P ·). It follows by [8], Lemmas 2 and 3 that H0 (P ·) ∈ X and H−1 (*(P ·)) ∈
Y. Since X is a direct summand of H0 (P ·) and Y is a direct summand of H−1 (*(P ·)),
it follows that H0 (P ·) generates X and H−1 (*(P ·)) cogenerates Y. It now follows by
Remark 2.9, Lemmas 3.2 and 3.3(2) that X = Xc (P ·) and Y = Yc (P ·).

Remark 5.9. Let

S = {P · : P −1 → P 0 ∈ Kb (proj A) | P ·is a tilting complex for A}
on which we de5ne the equivalence relation P1· ∼ P2· provided add P1· = add P2· in
Kb (proj A). Let T be the collection of torsion theories (X; Y) for mod A such that X
 M. Hoshino et al. / Journal of Pure and Applied Algebra 167 (2002) 15–35 35

contains an Ext-projective module X which generates X, Y contains an Ext-injective

module Y which cogenerates Y, and X is stable under DA ⊗A −. Set
<(P ·) = (Xc (P ·); Yc (P ·)) for P · ∈ S;

=((X; Y)) = MX· ⊕ HomA· (DA; NY·)[1] for (X; Y) ∈ T:

Then, according to Corollary 3.7, Propositions 5.5, 5.7 and Theorem 5.8, < and =
induce a one-to-one correspondence between S= ∼ and T.

References

[1] M. Auslander, Coherent functors, in: Proceedings of the Conference on Categorical Algebra, La Jolla
1965, Springer, Berlin, 1966, pp. 189 –231.
[2] A.A. Beilinson, J. Bernstein, P. Deligne, Faisceaux pervers, AstNerisque 100 (1982).
[3] M. BOokstedt, A. Neeman, Homotopy limits in triangulated categories, Compositio Math. 86 (1993)
209–234.
[4] S.E. Dickson, A torsion theory for abelian categories, Trans. Amer. Math. Soc. 121 (1966) 233–235.
[5] D. Happel, C.M. Ringel, Tilted algebras, Trans. Amer. Math. Soc. 274 (1982) 399–443.
[6] T. Holm, Derived equivalence classi5cation of algebras of dihedral, semidihedral, and quaternion type,
J. Algebra 211 (1999) 159–205.
[7] M. Hoshino, Tilting modules and torsion theories, Bull. London Math. Soc. 14 (1982) 334–336.
[8] M. Hoshino, On splitting torsion theories induced by tilting modules, Commun. Algebra 11 (4) (1983)
427–439.
[9] M. Hoshino, Y. Kato, Tilting complexes de5ned by idempotents, preprint.
[10] A. Neeman, The Grothendieck duality theorem via Bous5eld’s techniques and Brown representability,
J. Amer. Math. Soc. 9 (1996) 205–236.
[11] R. Hartshorne, Residues and Duality, Lecture Notes in Mathematics, vol. 20, Springer, Berlin, 1966.
[12] J. Rickard, Morita theory for derived categories, J. London Math. Soc. 39 (2) (1989) 436–456.