INVARIANCE OF GENERALISED REYNOLDS IDEALS

UNDER DERIVED EQUIVALENCES

ALEXANDER ZIMMERMANN

Abstract. For any algebraically closed field k of positive characteristic p and any non negative
integer n Külshammer defined ideals Tn A⊥ of the centre of a symmetric k-algebra A. We show
that for derived equivalent algebras A and B there is an isomorphism of the centres of A and
B mapping Tn A⊥ to Tn B ⊥ for all n. Recently Héthelyi, Horváth, Külshammer and Murray
showed that this holds for Morita equivalent algebras.

Introduction
Let k be an algebraically closed field of characteristic p > 0 and let A be a finite dimensional
symmetric k-algebra with non degenerate symmetrising bilinear form ( , ) on A. Külshammer
defined in [6] ideals Tn A⊥ of the centre of A by the following construction. Let KA be the
n
k-subspace of A generated by ab − ba for all a, b ∈ A and set Tn A := {x ∈ A | xp ∈ KA}. Let
Tn A⊥ is the subspace orthogonal to Tn A with respect to the form ( , ) on A. Note that Tn A⊥
is then an ideal of ZA as ZA = KA⊥ and that Tn A is a ZA submodule of A
ccsd-00009060, version 2 - 29 Mar 2006

n n
In [7] Külshammer shows that the equation (ζn (z), x)p = (z, xp ) for any x, z in the centre
of A defines a mapping ζn from the centre of A to the centre of A. Moreover, ζn (A) = Tn A⊥ .
Many properties of group algebras can be shown using the ideals Tn A⊥ . Concerning the ideals
Tn A⊥ , Héthelyi et al. show in [2] that Z0 A ⊆ (T1 A⊥ )2 ⊆ HA, where HA is the Higman
ideal of A, and where Z0 A is the sum of the centres of those blocks of A which are simple
algebras. They show that for odd p the left inclusion is an equality, whereas for p = 2 one gets
Z0 A = (T1 A⊥ )3 = (T1 A⊥ ) · (T2 A⊥ ). Finally, the authors show that e · (Tn A⊥ ) · e = Tn (eAe)⊥ for
any idempotent e of A. Now, the authors use the fact that within all algebras Morita equivalent
to A there is an, up to isomorphism unique, smallest algebra B = eAe Morita equivalent to A,
the basic algebra. If A is symmetric, B is symmetric as well, as follows in a more general context
by [11]. Multiplication by this idempotent induces an isomorphism between the centers of an
algebra A and its basic algebra B. Hence, the corresponding ideals Tn A⊥ and Tn B ⊥ are sent
to each other by this isomorphism. Composing two of them gives a corresponding statement for
Morita equivalent algebras.
In [2, question 5.4] Héthelyi et al. ask whether for two symmetric algebras A and B, the
condition that the derived categories of A and of B are equivalent imply the existence of an
isomorphism ϕ of their centres so that ϕ induces an isomorphism between the ideals Tn A⊥ and
Tn B ⊥ for all n ∈ N. The main objective of this paper is to give a positive answer to this question.
This way we provide new invariants for an equivalences between the triangulated categories
D (A) and Db (B) for algebras A and B. A number of invariants are known. Suppose Db (A) ≃
b

D b (B) as triangulated categories, then we get an isomorphism of the Hochschild homology

HH∗ (A) ≃ HH∗ (B) and the Hochschild cohomology HH ∗ (A) ≃ HH ∗ (B) (cf Rickard [9]), the
cyclic homology HC∗ (A) ≃ HC∗ (B), the cyclic cohomology HC ∗ (A) ≃ HC ∗ (B) of the algebra
A (Keller [4]), or by a result of Thomason and Trobaugh the K-theory K∗ (A) ≃ K∗ (B). Some
of them are quite useful in specializing the degree. So is HH 0 (A) ≃ Z(A) the centre of the
algebra A, or rankZ (K0 (A)) equals the number of isomorphism classes of simple A-modules.
Nevertheless, if these few computable invariants coincide, it is in general very difficult to decide
Date: February 17, 2005; revised: August 25, 2005.
2000 Mathematics Subject Classification. 20C05, 16L60, 18E30.

1
2 ALEXANDER ZIMMERMANN

whether two algebras have equivalent derived categories or not. So, invariants which are more
easy to determine in examples will be very welcome. Our result provides some of them.
The main result Theorem 1 will be proven in Section 3. Since there is no analogue of a
basic algebra for derived equivalences, we need to proceed differently from Héthelyi’s et al.’s
proof for Morita equivalence. Section 1 recalls some of the relevant notation and results from
homological algebra, for the convenience of the reader. We use the characterisation [7, (46)]; or
[2, Lemma 2.1] of Tn A⊥ as the image of the mapping ζnA and define in Section 2 the mapping ζnA
in a functorial manner by means of a composition of mappings between A ⊗k Aop -modules. We
apply the derived equivalence to each of the factors and using results in [11], and some delicate
commutativity considerations we are able to show that the mapping induced by a standard
derived equivalence on the morphism sets are indeed as asked. For notations concerning derived
categories and equivalences we follow [5]. Other references covering the needed background are
for example Gelfand-Manin [1] or Weibel [10]. The notation may differ slightly there.
Acknowledgement: I want to thank Burkhard Külshammer for having had a look into a first
draft of the manuscript, for pointing out a mistake there and for most helpful suggestions how to
fix this gap. I want to thank John Murray for explaining to me during the AMS/DMV-meeting
in May 2005 some of the historical background around the origins of the ideals Tn A⊥ .

1. A crash course on the relevant homological algebra

For the reader’s convenience and to fix notation we shall recall some basic facts in homological
algebra as it is needed in the sequel. Basic source is the book [5], and for some more general
aspects Gelfand-Manin [1], Weibel [10], or Rickard [9] as well as [11].
For a commutative noetherian ring k and a finitely generated k-algebra A we denote the
category of finitely generated left A-modules by A − mod and the category of all A-modules
by A − M od. Let K(A − mod) be the category of complexes in A − mod modulo homotopy.
Recall that the derived category D b (A) of bounded complexes of finitely generated A-modules is
formed by bounded complexes in K(A − mod) and formally inverting morphisms which induce
isomorphisms on homology. Recall furthermore that A − mod is a full subcategory of D b (A) by
mapping a module M to a complex with homogeneous components 0 in all degrees except in
degree 0 where the homogeneous component is M (cf e.g. Gelfand-Manin [1, III §5 Proposition
2]). Hence, any two objects M and N of A-mod may be considered as object in D b (A), and
then HomDb (A) (M, N ) = HomA (M, N ). This fact will be used at various places.
Recall that in case X is a complex in Db (A) whose homogeneous components are all projective,
then for any complex Y in D b (Aop ) one has Y ⊗A X = Y ⊗LA X. In case A and B are two algebras
over a field k, then for any X in Db (A ⊗k B op ) there is an X̃ in Db (A ⊗k B op ) so that X ≃ X̃
and so that all homogeneous components of X̃ are projective as A-modules and as B op -modules
(cf [5, Lemma 6.3.12]).
Let B be a k-algebra which is projective as k-module. By a result due to Keller (cf e.g. [3])
or [5, Chapter 8]) D b (A) is equivalent to Db (B) as triangulated categories if and only if there
is a complex X in Db (B ⊗k A) so that X ⊗LA − : D b (A) −→ Db (B) is an equivalence. Such
equivalences are called standard and X is called (two-sided) tilting complex. If B is symmetric,
then A is symmetric as well (cf [11]) and then the inverse equivalence to X ⊗LA − is given by
Homk (X, k) ⊗LB −. Moreover, Rickard has shown that this X ⊗LA − ⊗LA Homk (X, k) actually
defines an equivalence D b (A ⊗k Aop ) −→ Db (B ⊗k B op ) where the (A ⊗k Aop )-module A is
mapped to B (cf Rickard [9]; or [5, Proposition 6.2.6]). Hence X induces an isomorphism
Z(A) = EndA⊗k Aop (A) = EndDb (A⊗k Aop ) (A) ≃
EndDb (B⊗k B op ) (B) = EndB⊗k B op (B) = Z(B).
This isomorphism is explicitly exhibited in [5, Proposition 6.2.6]. In [11] it is shown that under
the equivalence induced by tensoring with X the (A ⊗k Aop )-module Homk (A, k) is mapped to
Homk (B, k).
 DERIVED EQUIVALENCES AND REYNOLDS IDEALS 3

We finish with some notation. Let C be a category and let X, Y and Z be any three objects
in C. We denote for any morphism ϕ ∈ HomC (X, Y ) the induced mapping HomC (Z, ϕ) :
HomC (Z, X) −→ HomC (Z, Y ) which is defined by (HomC (Z, ϕ)) (ψ) := ϕ ◦ ψ for any ψ ∈
HomC (Z, X). If Z is clear from the context, we write HomC (Z, ϕ) =: ϕ∗ for short.

2. Interpreting ζ
Recall from Section 1 that HomDb (A⊗k Aop ) (A, A) = EndA⊗k Aop (A) ≃ Z(A) .
Furthermore, by the adjointness formulas (cf e.g. Mac Lane [8, VI (8.7)]), we get
Homk (A ⊗A⊗k Aop A, k) ≃ HomA⊗k Aop (A, Homk (A, k))
f →
7 (a 7→ (b 7→ f (a ⊗ b)))
and since
P canonically by the very definition of a tensor product A ⊗A⊗k Aop A ≃ A/KA where
KA = a,b∈A k · (ab − ba) is the k-vector space generated by commutators, we have a functorial
isomorphism
Homk (A/KA, k) ≃ HomA⊗k Aop (A, Homk (A, k))
f →
7 (a 7→ (b 7→ f (ab)))
The mapping A/KA ∋ a 7→ ap ∈ A/KA was first defined by Richard Brauer who called it the
Frobenius mapping and proved that it is well defined (cf Külshammer [6, II]) and semilinear.
Denote by k(n) the n times Frobenius twisted copy of k.
The Frobenius mapping induces a well defined mapping
Homk (A/KA, k) −→ Homk (A/KA, k(1) )
f 7→ (a 7→ f (ap ))
The mapping
F r k ∗ : Homk (A/KA, k) −→ Homk (A/KA, k(1) )
f 7→ (a 7→ f (a)p )
induces a mapping
HomA⊗k Aop (A, Homk (A, k)) −→ HomA⊗k Aop (A, Homk (A, k(1) ))
and since for any algebra B one has a fully faithful embedding of B − mod into Db (B) by
considering a B-module as a complex with differential 0 and modules concentrated in degree 0
only, this in turn gives a mapping
HomDb (A⊗k Aop ) (A, Homk (A, k)) −→ HomDb (A⊗k Aop ) (A, Homk (A, k(1) )).
Put A∗ := Homk (A, k). Recall that a k-algebra A is symmetric if and only if there is an
isomorphism of A⊗k Aop -bimodules A ≃ A∗ , or equivalently there is a non degenerate symmetric
bilinear form ( , ) : A × A −→ k satisfying (a, cb) = (ac, b) for any a, b, c ∈ A (cf e.g. [5, Chapter
n n
9]). Then the mapping ζnA is defined by the equation (ζn (z), x)p = (z, xp ) which can be written
as composition of the mappings in the following diagram (‡):
HomDb (A⊗k Aop ) (A, A) −→ HomDb (A⊗k Aop ) (A, A∗ )
↓ ((F rk )∗ )n
(‡) A
↑ ζn HomDb (A⊗k Aop ) (A, Homk (A, k(n) )
↑ ((F rA )∗ )n
HomDb (A⊗k Aop ) (A, A) −→ HomDb (A⊗k Aop ) (A, A∗ )
where the horizontal arrows are induced by the isomorphism
A −→ A∗
a 7→ (b 7→ (a, b))
which is coming from the symmetrising bilinear form ( , ) : A ⊗k A −→ k. of A.
4 ALEXANDER ZIMMERMANN

3. Behaviour under derived equivalences

In this section we prove our main result.
Theorem 1. Let k be an algebraically closed field of characteristic p > 0 and let A and B be
finite dimensional k-algebras. If D b (A) ≃ D b (B) as triangulated categories, then there is an
isomorphism ϕ : ZA −→ ZB between the centres ZA of A and ZB of B so that ϕ(Tn A⊥ ) =
Tn B ⊥ for all positive integers n ∈ Z.
Remark 3.1. This answers to the positive question 5.4 posed by László Héthelyi, Erszébet
Horváth, Burkhard Külshammer and John Murray in [2].
Proof: Let F : Db (A) −→ Db (B) be a standard derived equivalence with two-sided tilting
complex X. Let X ′ be the inverse tilting complex. Then, in [11] it is shown that X ⊗A − ⊗A X ′
induces an equivalence G : D b (A ⊗k Aop ) −→ D b (B ⊗k B op ) mapping the A-A-bimodule A AA to
the B-B-bimodule B BB ,
G( A AA ) = B BB .
From [11, Lemma 1] we know that
G (Homk (A, k)) = Homk (B, k).
We shall show that  
G Homk (A, k(n) ) = Homk (B, k(n) )
for all n ∈ Z. Indeed, X ⊗A − ≃ HomA (X ′ , −) and − ⊗A X ′ ≃ HomA (X, −) by the adjointness
properties of Hom and ⊗-functors. Hence (cf [5, proof of Corollary 6.3.6]),
X ⊗A Homk (A, k(n) ) ⊗A X ′ ≃ HomA (X ′ , Homk (A, k(n) )) ⊗A X ′
≃ Homk (A ⊗A X ′ , k(n) ) ⊗A X ′
≃ Homk (X ′ , k(n) ) ⊗A X ′
(†) ≃ HomA (X, Homk (X ′ , k(n) ))
≃ Homk (X ⊗A X ′ , k(n) )
≃ Homk (B, k(n) )
We apply now G to the diagram (‡) of Section 2 and get a commutative diagram
HomDb (B⊗k B op ) (B, B) −→ HomDb (B⊗k B op ) (B, B ∗ )
G↑≃ G↑≃
HomDb (A⊗k Aop ) (A, A) −→ HomDb (A⊗k Aop ) (A, A∗ )
↓ ((F rk )∗ )n
A
↑ ζn HomDb (A⊗k Aop ) (A, Homk (A, k(n) ))
↑ ((F rA )∗ )n
HomDb (A⊗k Aop ) (A, A) −→ HomDb (A⊗k Aop ) (A, A∗ )
G↓≃ G↓≃
HomDb (B⊗k B op ) (B, B) −→ HomDb (B⊗k B op ) (B, B ∗ ) .
It is clear that the upper and the lower square are commutative, since they arise as squares
induced from applying an equivalence of categories.
Recall the notation we use as explained at the end of Section 1.
We obtain a commutative diagram
G
HomDb (A⊗k Aop ) (A, Homk (A, k)) −→ HomDb (B⊗k B op ) (B, Homk (B, k))
↓ Homk (A, (F rk )n ) ↓ϕ
G
HomDb (A⊗k Aop ) (A, Homk (A, k(n) )) −→ HomDb (B⊗k B op ) (B, Homk (B, k(n) ))
where ϕ = G ◦ Homk (A, (F r k )n ) ◦ G−1 .
We shall need to see that ϕ = Homk (B, (F r k )n ).
Claim 1. G ◦ Hom(A, F r k ) = Hom(B, F r k ) ◦ G.
 DERIVED EQUIVALENCES AND REYNOLDS IDEALS 5

Proof: Observe that G = X ⊗A − ⊗A X ′ acts only on the contravariant variables. Going

through the isomorphisms (†), since F r k acts on the covariant variable only, this proves the
claim.
Therefore, the diagram
Hom(B,(F r k )n )
HomDb (B⊗k B op ) (B, B ∗ ) −→ HomDb (B⊗k B op ) (B, Homk (B, k(n) ))
≃↑ G ≃↑ G
Hom(A,(F r k )n )
HomDb (A⊗k Aop ) (A, A∗ ) −→ HomDb (A⊗k Aop ) (A, Homk (A, k(n) ))
is commutative and the vertical morphisms are isomorphisms since G is an equivalence, and
since the images of the various objects under G in their version A and B correspond to each
other.
Claim 2. G ◦ Hom(F r A , k) = Hom(F r B , k) ◦ G.
Before starting with the proof observe the following consequences. Once the claim is estab-
lished the diagram
Hom((F r B )n ,k)
HomDb (B⊗k B op ) (B, B ∗ ) −→ HomDb (B⊗k B op ) (B, Homk (B, k(n) ))
≃↑ G ≃↑ G
Hom((F r A )n ,k)
HomDb (A⊗k Aop ) (A, A∗ ) −→ HomDb (A⊗k Aop ) (A, Homk (A, k(n) ))
is commutative.
Observe that since Homk (F r A , k) is not A ⊗k Aop -linear, the functor G is not defined on
Homk (F r A , k). Hence, the only way to prove the commutativity of the above diagram is by
inspection of the values.
Proof of Claim 2: We need to make explicit the mappings
G : HomDb (A⊗k Aop ) (A, A∗ ) −→ HomDb (B⊗k B op ) (B, B ∗ )
and
G : HomDb (A⊗k Aop ) (A, Homk (A, k(1) )) −→ HomDb (B⊗k B op ) (B, Homk (B, k(1) )).
For this, it is useful, and possible, to replace B by X ⊗A X ′ and A by X ′ ⊗B X.
We first deal with the first identification. Then, again by the usual adjointness formula
between Hom and ⊗, one has to make explicit an isomorphism
G : Homk (A ⊗A⊗k Aop A, k) −→ Homk (B ⊗B⊗k B op B, k),
or, replacing B by X ⊗A X ′ and A by X ′ ⊗B X,
G : Homk ((X ′ ⊗B X) ⊗A⊗k Aop (X ′ ⊗B X), k) → Homk ((X ⊗A X ′ ) ⊗B⊗k B op (X ⊗A X ′ ), k).
The isomorphism
A ⊗A⊗k Aop A ≃ A/KA ≃ B/KB ≃ B ⊗B⊗k B op B
comes from a mapping
(x ⊗ y) ⊗ 1A 7→ (y ⊗ x) ⊗ 1B
where it is clear that this is well defined. The diagonal mapping
A −→ A ⊗A A ⊗A · · · ⊗A A
| {z }
p factors
a 7→ a ⊗ a ⊗ ··· ⊗ a
is exactly the p-power map A ∋ a 7→ ap ∈ A. Composing with the natural projection A −→
A/KA this defines the p-power mapping A ∋ a 7→ ap ∈ A/KA. If k is of characteristic p, then
this last mapping is additive and factors through A −→ A/KA.
6 ALEXANDER ZIMMERMANN

Now, we observe that A/KA is equally isomorphic to

A ⊗A A ⊗A · · · ⊗A A ⊗A⊗k Aop A.
| {z }
p−1 factors

Moreover, we have seen A ≃ X′ ⊗B X and B ≃ X ⊗A X ′ and one recovers an isomorphism

p−1 p−1
(X ′ ⊗B X)⊗A ⊗A⊗k Aop (X ′ ⊗B X) −→ (X ⊗A X ′ )⊗B ⊗B⊗k B op (X ⊗A X ′ )
(xi ⊗ yi )p−1
i=1 ⊗ (xp ⊗ yp ) 7→ ((yp ⊗ x1 ) ⊗ (yi ⊗ xi+1 )p−2
i=1 ) ⊗ (yp−1 ⊗ xp )
which is an incarnation of the isomorphism A/KA −→ B/KB. We need to show that this is
well-defined, but actually this is just a straight forward and detailed examination which ring
acts where in which way.
Therefore, the p-power map on A/KA is mapped to the p-power map on B/KB by a standard
derived equivalence.
We need to explain the second isomorphism
G : HomDb (A⊗k Aop ) (A, Homk (A, k(1) )) −→ HomDb (B⊗k B op ) (B, Homk (B, k(1) )).
Here, we observe that
HomDb (A⊗k Aop ) (A, Homk (A, k(1) )) ≃ Homk (A ⊗A⊗k Aop A, k(1) )
and the very same arguments and constructions as above hold. The only difference is that one
needs to consider semilinear mappings only at the end. The reorganization procedure is just the
same. In particular, the action of F r B consists in tensoring the whole term on the right p times
over B ⊗ B op . It is now immediate to see that this operation commutes with this reorganization
of factors as described by explaining ν. So,
G ◦ Hom(F r A , k) = Hom(F r B , k) ◦ G.

Claim 3. The images of G ◦ ζnA ◦ G−1 and of ζnB coincide.

Proof: Since ϕ : A −→ Homk (A, k) is an isomorphism of A ⊗k Aop -modules, and since
G is a functor, Gϕ is an isomorphism as well. As we know that choosing an isomorphism
B −→ Homk (B, k) is equivalent to choosing a symmetrising form making B into a symmetric
algebra, we may well work with this form instead of the original one. Actually, given two different
isomorphisms φ : B −→ Homk (B, k) and ψ : B −→ Homk (B, k), then for all x ∈ B one has
φ−1 ψ(x) = λx for an invertible central λ ∈ Z(B)∗ . So, the resulting ζnB differ by invertible
central elements. As a consequence, the images are identical.
We shall finish the proof of the theorem. By the previous claims the composition
Gφ (G(F r A )∗ G−1 )n
ZB −→ Homk (B/KB, k) −→ Homk (B/KB, k(n) )
((G(F r k )∗ )n )−1 Gφ−1
−→ Homk (B/KB, k) −→ ZB
is a mapping which differs from ζnB by some central unit of B and therefore the isomorphism
induced by G between the centres of A and B maps Tn A⊥ to Tn B ⊥ .
This finishes the proof of the theorem.

References
[1] Sergei Gelfand and Yuri Manin, Methods of homological algebra, Springer Verlag; Berlin-Heidelberg
1996.
[2] László Héthelyi, Erszébet Horváth, Burkhard Külshammer and John Murray, Central ideals and Cartan
invariants of symmetric algebras, to appear in J. Alg.
[3] Bernhard Keller, A remark on tilting theory and DG-algebras, manuscripta math. 79 (1993) 247-253.
[4] Bernhard Keller, Invariance and licalization for cyclic homology of DG-algebras, J. of pure and appl. Alg.
123 (1998) 223-273.
 DERIVED EQUIVALENCES AND REYNOLDS IDEALS 7

[5] Steffen König and Alexander Zimmermann, Derived Equivalences for Group Rings; Lecture Notes in
Mathematics 1685, Springer Verlag Berlin-Heidelberg 1998.
[6] Burkhard Külshammer, Bemerkungen über die Gruppenalgebra als symmetrische Algebra I, II, III and IV,
J. Alg. 72 (1981) 1-7; 75 (1982) 59-69; 88 (1984) 279-291; 93 (1985) 310-323
[7] Burkhard Külshammer, Group-theoretical descriptions of ring theoretical invariants of group algebras,
Progress in Mathematics 95 (1991) 425-441.
[8] Saunders Mac Lane, Homology, Springer Verlag Berlin-Heidelberg 1975.
[9] Jeremy Rickard, Derived equivalences as derived functors, J. London Math. Soc. 43 (1991), 37–48.
[10] Charles Weibel, Introduction to homological algebra, Cambridge university press 1994.
[11] Alexander Zimmermann, Tilted symmetric orders are symmetric orders, Arch. Math. 73 (1999) 15-17.

Université de Picardie,
Faculté de Mathématiques et LAMFA (UMR 6140 du CNRS),
33 rue St Leu,
F-80039 Amiens Cedex 1,
France
E-mail address: [email protected]