Linear Algebra and its Applications 431 (2009) 1587–1602

Contents lists available at ScienceDirect

Linear Algebra and its Applications

journal homepage: www.elsevier.com/locate/laa

Biderivations of triangular algebras

Dominik Benkovič
University of Maribor, FNM, Koroška 160, 2000 Maribor, Slovenia

A R T I C L E I N F O A B S T R A C T

Article history: Let A be a triangular algebra. A bilinear map ϕ : A×A→A is called

Received 24 March 2009 a biderivation if it is a derivation with respect to both arguments.
Accepted 21 May 2009 In this paper we define the concept of an extremal biderivation,
Available online 18 June 2009 and prove that under certain conditions a biderivation of a trian-
Submitted by C.K. Li
gular algebra A is a sum of an extremal and an inner biderivation.
The main result is then applied to (block) upper triangular matrix
AMS classification: algebras and nest algebras. We also consider the question when a
16W25 derivation of a triangular algebra is an inner derivation.
15A78 © 2009 Elsevier Inc. All rights reserved.
47L35

Keywords:
Triangular algebra
Nest algebra
Triangular matrix algebra
Biderivation
Inner biderivation
Derivation
Inner derivation

1. Introduction

Let C be a commutative ring with identity. A triangular algebra is every algebra of the form
 
A M
,
B

where A and B are unital algebras over C and M is an (A, B)-bimodule which is faithful as a left A-
module as well as a right B-module. Basic examples of triangular algebras are upper triangular matrix
algebras and nest algebras. Cheung [9,10] obtained interesting results on abstract triangular algebras;
he described commuting maps and Lie derivations of these algebras. In a joint paper with Eremita

E-mail address: dominik.benkovic@uni-mb.si

0024-3795/$ - see front matter © 2009 Elsevier Inc. All rights reserved.
doi:10.1016/j.laa.2009.05.029
1588 D. Benkovič / Linear Algebra and its Applications 431 (2009) 1587–1602

[1], the author of this paper studied commuting traces of biadditive maps and Lie isomorphisms of
triangular algebras. Wong [14] treated Jordan isomorphisms of triangular algebras, while Zhang and
Yua [16] described Jordan derivations.
Let A be an algebra on a commutative ring C and let Z (A) be its centre. Further let [x, y] = xy − yx
be the commutator (the Lie product) of the elements x, y ∈ A. A linear map d : A → A is a derivation if
d(xy) = d(x)y + xd(y) for all x, y ∈ A.
If a derivation d is of the form d(x) = [x, a], where a ∈ A, then d is said to be an inner derivation. A
bilinear map ϕ : A × A → A is a biderivation if it is a derivation with respect to both components,
meaning that
ϕ (xy, z) = xϕ (y, z) + ϕ (x, z) y and ϕ (x, yz) = yϕ (x, z) + ϕ (x, y) z
for all x, y, z ∈ A. If A is a noncommutative algebra then the map
ϕ (x, y) = λ [x, y] for all x, y ∈ A, (1.1)
where λ ∈ Z (A), is a basic example of biderivation. Biderivation of the form (1.1) are therefore called
inner biderivations.
Biderivations are a subject of research in various areas. Brešar et al. [7] proved that all biderivations
on noncommutative prime rings are inner, i.e. of the form (1.1). This theorem has proved to be useful
in the study of commutating maps. More details about commuting maps, biderivations and their
generalizations can be found in [2–6]. The reader is especially encouraged to read the survey paper [5,
Section 3] where applications of biderivations to other fields are also described.
In this paper we will investigate biderivations of triangular algebras. An indirect motivation for this
comes from two recent papers. Zhang et al. [15, Theorem 2.1] showed that biderivations of nest algebras
are usually inner. There are, however, special cases of nest algebras with non-inner biderivations. Zhao
et al. [17] used the results in [15] to prove that every biderivation of an upper triangular matrix algebra
is a sum of an inner biderivation and a special biderivation, which they call an extremal biderivation.
The paper is organized in the following way. The second section summarizes the main properties
and describes triangular algebras we will work with. As biderivations are closely related to derivations,
in the third section we shall consider the question whether every derivation of a triangular algebra is
an inner derivation.
The referee pointed out to us that the problem of innerness of derivations was already considered
in a somewhat different context by Forrest and Marcoux [13] and as well as by Cheung in his disser-
tation [11]. The original version of our paper contained detailed proofs of all results, but receiving the
information by the referee we have shortened the section. The important result of the third section
is Theorem 3.5 which describes sufficient conditions for every derivation of a triangular algebra to be
inner. We also provide examples of triangular algebras with non-inner derivations. In the final section
we first consider the concept of an extremal biderivation. It turns out that this kind of derivations
appear naturally on triangular algebras. We say that ψ : A × A → A is an extremal biderivation if it is
of the form
ψ (x, y) = [x, [y, a]] for all x, y ∈ A, (1.2)
where a ∈ A such that [a, [A, A]] = 0. The main result of the paper is Theorem 4.11. A corollary to
this theorem states that under certain conditions a biderivation of a triangular algebra is a sum of
an inner biderivation (1.1) and an extremal biderivation (1.2). Finally we apply the main result to a
(block) triangular matrix algebra and to a nest algebra (Corollaries 4.13 and 4.14, respectively) and
provide examples of triangular algebras with biderivations which are not sums of inner and extremal
biderivations.

2. Triangular algebras

Let A and B be unital algebras over a commutative ring C, and let M be a unital (A, B)-bimodule,
which is faithful as a left A-module and also as a right B-module. Recall that a left A-module M is
faithful if aM = 0 implies that a = 0. The C-algebra
 D. Benkovič / Linear Algebra and its Applications 431 (2009) 1587–1602 1589

  
a m
A = Tri (A, M, B) = b
; a ∈ A, m ∈ M, b ∈ B

under the usual matrix operations is called a triangular algebra. Let us define two natural projections
πA : A → A and πB : A → B by
   
a m a m
πA : b
→ a and πB : b
→ b.

By [9, Proposition 3] we know that the center Z (A) of A coincides with

  
a 0
b
; am = mb for all m ∈ M .

Moreover, πA (Z (A)) ⊆ Z (A) and πB (Z (A)) ⊆ Z (B), and there exists a unique algebra isomorphism
τ : πA (Z (A)) → πB (Z (A)) such that am = mτ (a) for all m ∈ M. The most important examples of
triangular algebras are upper triangular matrix algebras, block upper triangular matrix algebras and
nest algebras.
Upper triangular matrix algebras. Let Mk×m (C ) denotes the set of all k × m matrices and let Tn (C )
denotes the algebra of all n × n upper triangular matrices over C. For n  2 and each 1  l  n − 1 the
algebra Tn (C ) can be represented as a triangular algebra of the form
 
Tl (C ) Ml×(n−l) (C )
Tn (R) = .
Tn−l (C )
Block upper triangular matrix algebras. Let N be the set of all positive integers and let n ∈ N. For every
positive integer m, m  n, we denote by k̄ = (k1 , k2 , . . ., km ) ∈ Nm an ordered m-vector of positive inte-
gers such that n = k1 + k2 + · · · + km . The block upper triangular matrix algebra Bnk̄ (C ) is a subalgebra
of Mn (C ) of the form
⎛ ⎞
Mk1 (C ) Mk1 ×k2 (C ) ··· Mk1 ×km (C )
⎜ 0
⎜
Mk2 (C ) ··· Mk2 ×km (C )⎟
⎟
Bnk̄ (C ) = ⎜ .. .. .. .. ⎟.
⎝ . . . . ⎠
0 0 ··· Mkm (C )
Note that Mn (C ) and Tn (C ) are two special cases of block upper triangular matrix algebras. If we have
n  2 and Bnk̄ (C ) =
/ Mn (C ) then Bnk̄ (C ) is a triangular algebra and can be represented as
⎛ ⎞
k̄
⎝Bl (C ) Ml×(n−l) (C )⎠
1
Bnk̄ (C ) = k̄
,
Bn2−l (C )

where 1  l < m and k̄1 ∈ Nl , k̄2 ∈ Nm−l .

Nest algebras. A nest is a chain N of closed subspaces of a complex Hilbert space H containing {0}
and H which is closed under arbitrary intersections and closed linear span. The nest algebra associated
to N is the algebra
T (N ) = {T ∈ B (H ) |T (N ) ⊆ N for all N ∈ N } .
A nest N is called trivial if N = {0, H }. A nontrivial nest algebra is a triangular algebra. Namely,
if N ∈ N \{0, H } and E is the orthonormal projection onto N, then N1 = E (N ) and N2 = (1 − E )(N )
are nests of N and N ⊥ , respectively. Moreover, T (N1 ) = E T (N )E, T (N2 ) = (1 − E )T (N )(1 − E ) are
nest algebras and
 
T (N1 ) E T (N ) (1 − E )
T (N ) = .
T (N2 )
Let us mention that finite dimensional nest algebras are isomorphic to complex block upper triangular
matrix algebras.
Let 1A and 1B be identities of the algebras A and B, respectively, and let 1 be the identity of the
triangular algebra A. Throughout this paper we shall use following notation
1590 D. Benkovič / Linear Algebra and its Applications 431 (2009) 1587–1602

   
1A 0 0 0
e = 0
, f =1−e= 1B
.
We immediately notice that e and f are orthogonal idempotents of A and so A may be represented
as

A = 1A1 = (e + f ) A (e + f ) = eAe + eAf + f Af .

Here eAe is a subalgebra of A isomorphic to A, f Af is a subalgebra of A isomorphic to B and eAf is a
(eAe, f Af )-bimodule isomorphic to the bimodule M. To simplify notation we will use the following
convention: a = eae ∈ A = eAe, b = fbf ∈ B = f Af and m = emf ∈ M = eAf . Then each element
x = exe + exf + fxf ∈ A can be represented in the form
= eae + emf + fbf = a + m + b,
x
where a ∈ A, b ∈ B, m ∈ M.

3. Derivations of triangular algebras

This section is dedicated to the treatment of derivations of triangular algebras. The main theorem
(Theorem 4.11) answering the question on innerness of biderivations on triangular algebras relies on a
condition involving derivations. The central question of this section is when derivations of triangular
algebras are inner. As we have already mentioned in introduction the problem of innerness of deri-
vations was already considered in [13,11]. Therefore, the proofs of some results in this section will be
omitted.
Recall that d : A → A is a derivation if d(xy) = d(x)y + xd(y) for all x, y ∈ A. A derivation is inner
if it is of the form d(x) = [x, x0 ] for some x0 ∈ A. We begin with the following results (see [11, Theorem
2.2.1]):

Proposition 3.1. Let d : A → A be a derivation. Then d = d1 + d2 , where d1 : A → A is an inner der-

ivation and d2 : A → A a derivation satisfying d2 (e) = 0.

Proposition 3.2. Let d : A → A be a derivation with the property d(e) = 0. Then d is of the form
   
a m d1 (a) f (m)
d
b
= d2 (b)
, (3.1)

where d1 : A → A is a derivation, d2 : B → B is a derivation and f : M → M is a linear map satisfying

f (am) = d1 (a)m + af (m) and f (mb) = f (m)b + md2 (b) (3.2)
for all a ∈ A, b ∈ B, m ∈ M .

Using [11, Proposition 2.2.3], we obtain:

Proposition 3.3. A derivation d : A → A is an inner derivation if and only if for each map f : M → M,
defined in (3.1), there exists a0 ∈ A and b0 ∈ B such that f (m) = a0 m + mb0 for all m ∈ M .

Let M be a unital (A, B)-bimodule. The map f : M → M satisfying f (am) = af (m) and f (mb) = f (m)b
for all a ∈ A, b ∈ B, m ∈ M is called a bimodule homomorphism. A bimodule homomorphism f : M → M
is of the standard form if there exist a0 ∈ Z (A), b0 ∈ Z (B) such that f (m) = a0 m + mb0 for all m ∈ M.

Corollary 3.4. Suppose that every derivation of the triangular algebra Tri(A, M, B) is inner. Then every
bimodule homomorphism f : M → M is of the standard form.

Proof. Let f : M → M be a bimodule homomorphism. It is clear from Proposition 3.2 that

   
a m 0 f (m)
d
b
= 0
 D. Benkovič / Linear Algebra and its Applications 431 (2009) 1587–1602 1591

is a derivation of the algebra Tri(A, M, B). Since we assumed that every derivation is inner we conclude
(using Proposition 3.3) that there exist elements a0 ∈ A and b0 ∈ B such that f (m) = a0 m + mb0 for
all m ∈ M. Note that a and b are in fact central elements. 

Two basic examples of triangular algebras are the algebra of the upper triangular matrices Tn (C )
and the nest algebra T (N ) where N is a nontrivial nest. It is proven in [8] that every derivation of
Tn (C ) is inner. A similar claim holds for nest algebras (see [12, Theorem 19.7]). We will prove that every
derivation of the block upper triangular matrix algebra is also inner (Proposition 3.6). On the basis of
Proposition 3.3 we see that the question of innerness of derivations of triangular algebras is linked to
the characterization of the special bimodule maps (3.2). The following theorem follows from Cheung’s
result [11, Corollary 2.3.8]. However, for the sake of completeness we give its proof. It should also be
mentioned that Forrest and Marcoux [13, Theorem 2.8] proved similar result for Banach triangular
algebras.

Theorem 3.5. Let A =Tri(A, M, B). If

(i) every derivation of A is inner,

(ii) every derivation of B is inner,
(iii) each bimodule homomorphism f : M → M is of the standard form,
then all the derivations of the algebra A are inner.

Proof. Let d be a derivation of the triangular algebra A. By Proposition 3.1 we may assume that d(e) =
0. Since each derivation of algebras A and B is inner we have that derivations d1 : A → A, d2 : B → B
have the following form
 
d1 (a) = a, a in d2 (b) = b, b for all a ∈ A, b ∈ B,

where a ∈ A and b ∈ B are fixed elements. Let d : A → A be the inner derivation d(x) = [x, x0 ],
where x0 = ea e + fb f . Then d = d − d is a derivation of the algebra A that taking the form
        
 a m d1 (a) f (m) a, a mb − a m 0 g (m)
d = − = ,
b d2 (b) b, b 0
where the map g : M → M satisfies g (am) = ag (m) and g (mb) = g (m)b for all a ∈ A, b ∈ B, m ∈ M.
The assumption (iii) implies that there exist elements a0 and b0 such that g (m) = a0 m + mb0 for all

m ∈ M. We now use Proposition 3.3 to see that d is an inner derivation. Therefore d is a sum of inner

derivations d and d . This completes the proof. 

As an application of Theorem 3.5 we prove the following result:

Proposition 3.6. Let C be a commutative ring with identity. Every derivation of the block upper triangular
matrix algebra Bnk̄ (C ) is inner.

Proof. We prove this proposition by induction on n. In case n = 1, we have B1k̄ (C ) = C and every
derivation of the C-algebra C is clearly inner. Namely, the algebra C does not have nonzero derivations.
Let n  2. If Bnk̄ (C ) = Mn (C ), then every derivation of Mn (C ) is inner. Now we assume Bnk̄ (C ) =
/ Mn (C ).
Then Bnk̄ (C )is a triangular algebra and can be represented as
⎛ ⎞
k̄1
B (C ) M l×(n−l) ( C )
Bnk̄ (C ) = ⎝ l k̄
⎠,
Bn2−l (C )
k̄ k̄
where l, n − l < n. Both algebras A = Bl 1 (C ) and B = Bn2−l (C ) are block upper triangular matrix
algebras, which have dimension less than n. Hence by induction hypothesis we may assume that
1592 D. Benkovič / Linear Algebra and its Applications 431 (2009) 1587–1602

all derivations of algebras A and B are inner. Therefore, the first two conditions of Theorem 3.5 are
satisfied. Let us prove that Bnk̄ (C ) also satisfies the third condition of Theorem 3.5. Let f : Ml×(n−l) (C ) →
Ml×(n−l) (C ) be an (A, B)-bimodule homomorphism. Since both algebras A and B are central algebras
(this means that Z (A) = C1A and Z (B) = C1B ) it suffices to prove that bimodule homomorphism f is
of the form
f (m) = λm for all m ∈ Ml×(n−l) (C ) ,
where λ ∈ C. Let {eij , 1  i, j  n} be the system of matrix units of Mn (C ). Note that every unital (A, B)-
bimodule Ml×(n−l) (C ) is generated by matrix units {eij , 1  i  l, l + 1  j  n}. Since f is a bilinear map,
it is enough to see that
 
f eij = λeij for all 1  i  l, l + 1  j  n.
Since ell ∈ A, el+1,l+1 ∈ B and f is a bimodule homomorphism, we obtain
     
f el,l+1 = f ell el,l+1 el+1,l+1 = ell f el,l+1 el+1,l+1 = λel,l+1 ,
where λ ∈ C. On the basis of this identity for all 1  i  l, l + 1  j  n we get
     
f eij = f eil el,l+1 el+1,j = eil f el,l+1 el+1,j
 
= eil λel,l+1 el+1,j = λeij ,
since eil ∈ A, el+1,j ∈ B. Hence we see that every bimodule homomorphism is of the standard form,
and so the assumption (iii) of Theorem 3.5 is fulfilled. Therefore every derivation of Bnk̄ (C ) is inner.


We conclude this section with three examples of triangular algebras with non-inner derivations.
Example 3.7 is constructed on the basis of the assumption (iii) from the previous theorem (the cor-
responding bimodule homomorphism is not of the standard form). In Example 3.8 one of algebras
forming the triangular algebra has a non-inner derivation. In these two examples the algebras A and B
are both commutative. In Example 3.9 we use noncommutative algebras.

Example 3.7. Let A = Tri(Z, Q, Z) and let the map d : A → A be defined as

   
1
a m 0 m
d
b
= 2 .
0

Then d is a derivation of the algebra A that is not inner.

Example 3.8. Let C be a commutative ring with identity and let

     
a11 a12 b 0
A= ; a11 , a12 ∈ C , B = ; b∈C
a11 b

be commutative unital subalgebras of the algebra T2 (C ). Then the algebra M = T2 (C ) becomes an

(A, B)-bimodule, that is a faithful left A-module and a faithful right B-module. Let us denote A =
Tri(A, M, B). The map δ : A → A, δ(a(e11 + e22 ) + te12 ) = te12 is a derivation that is not inner. It
is easy to see that the map f : M → M, f (a11 e11 + a12 e12 + a22 e22 ) = −a22 e22 satisfies the condi-
tions
f (am) = δ (a) m + af (m) and f (mb) = f (m) b
for all a ∈ A, b ∈ B, m ∈ M. Therefore
   
a m δ (a) f (m)
d
b
= 0
is a non-inner derivation of the triangular algebra.
 D. Benkovič / Linear Algebra and its Applications 431 (2009) 1587–1602 1593

Example 3.9. Let A be an algebra with identity that has a non-inner derivation δ : A → A. Let A =
Tri(A, A, A). Then the map d : A → A defined by
   
a m δ (a) δ (m)
d
b
= δ (b)
is a non-inner derivation.

Note that the triangular algebra A = Tri(A, A, A) from Example 3.9 satisfies the assumption (iii)
from Theorem 3.5. According to Example 3.9 and Theorem 3.5 we can conclude:

Remark 3.10 [13, Lemma 3.1]. Let A be an algebra with identity and let A = Tri(A, A, A) be a triangular
algebra. Every derivation of A is inner if and only if every derivation of A is inner.

4. Biderivations of triangular algebras

In this section we will use extremal biderivations to prove the main Theorem 4.11, which sets out
the conditions under which a biderivation of a triangular algebra is a sum of an inner and an extremal
biderivation. We shall conclude the section with some examples and with an idea on how to construct
examples of non-inner biderivations.
Recall that a map ϕ : A × A → A is a biderivation if it is a derivation in both components. This
means that
ϕ (xy, z) = xϕ (y, z) + ϕ (x, z) y and ϕ (x, yz) = yϕ (x, z) + ϕ (x, y) z
for all x, y, z ∈ A. An inner biderivation ϕ is of the form ϕ(x, y) = λ[x, y] where λ ∈ Z (A). We begin
with a few lemmas.

Lemma 4.1 [4, Corollary 2.4]. Let ϕ : A × A → A be a biderivation. Then

ϕ (x, y) [u, v] = [x, y] ϕ (u, v)
for all x, y, u, v ∈ A.
Proof. Consider ϕ(xu, yv) for arbitrary x, y, u, v ∈ A. Since ϕ is a derivation in the first argument, we
have
ϕ(xu, yv) = ϕ(x, yv)u + xϕ(u, yv),
and since it is also a derivation in the second argument it follows that
ϕ(xu, yv) = ϕ(x, y)vu + yϕ(x, v)u + xϕ(u, y)v + xyϕ(u, v).
On the other hand, first using the derivation law in the second and after that in the first argument
we get
ϕ(xu, yv) = ϕ(x, y)uv + xϕ(u, y)v + yϕ(x, v)u + yxϕ(u, v).
Comparing both relations we obtain ϕ(x, y)[u, v] = [x, y]ϕ(u, v) for all x, y, u, v ∈ A. 
Lemma 4.2. Let ϕ : A × A → A be a biderivation. Then

(i) ϕ(x, 1) = 0 = ϕ(1, x) for all x ∈ A;

(ii) ϕ(x, 0) = 0 = ϕ(0, x) for all x ∈ A;
(iii) ϕ(e, e) = −ϕ(e, f ) = −ϕ(f , e) = ϕ(f , f ).

Proof. The identity ϕ(x, 1) = 0 follows from

ϕ (x, 1) = ϕ (x, 1 · 1) = 1ϕ (x, 1) + ϕ (x, 1) 1.
Similarly ϕ(1, x) = 0, while ϕ(x, 0) = ϕ(0, x) = 0 is obvious.
1594 D. Benkovič / Linear Algebra and its Applications 431 (2009) 1587–1602

To prove (iii) we use (i) and the equality e + f = 1. Then we may write
ϕ (e, f ) = ϕ (e, f ) − ϕ (e, 1) = −ϕ (e, 1 − f ) = −ϕ (e, e) .
Similarly ϕ(f , e) = −ϕ(e, e). It is also true that
ϕ (f , f ) = ϕ (1 − e, 1 − e) = ϕ (1, 1) − ϕ (1, e) − ϕ (e, 1) + ϕ (e, e)
= ϕ (e, e) .
Thus we conclude that ϕ(e, e) = −ϕ(e, f ) = −ϕ(f , e) = ϕ(f , f ). 

Lemma 4.3. Let A = Tri(A, M, B) be a triangular algebra and let ϕ : A × A → A be a biderivation. If

x, y ∈ A are such that [x, y] = 0, then ϕ(x, y) = eϕ(x, y)f .

Proof. Let x, y ∈ A and write

ϕ (x, y) = eϕ (x, y) e + eϕ (x, y) f + f ϕ (x, y) f .
The assumption [x, y] = 0 and Lemma 4.1 imply that
ϕ (x, y) [e, emf ] = [x, y] ϕ (e, emf ) = 0 and
[e, emf ] ϕ (x, y) = ϕ (e, emf ) [x, y] = 0

for all m ∈ M. Therefore eϕ(x, y)eM = 0 and Mf ϕ(x, y)f = 0. Since M is a faithful left A-module and
a faithful right B-module we see that eϕ(x, y)e = 0 = f ϕ(x, y)f . Therefore ϕ(x, y) = eϕ(x, y)f . 

Remark 4.4. Let x0 ∈ A, x0 ∈

/ Z (A) and suppose that [[x, y], x0 ] = 0 for all x, y ∈ A. Then the map
ψ : A × A → A defined by
ψ (x, y) = [x, [y, x0 ]] for all x, y ∈ A (4.1)
is a biderivation.

Proof. Using Jacobi identity it follows from [[x, y], x0 ] = 0 that

[x, [y, x0 ]] = [y, [x, x0 ]] + [[x, y] , x0 ] = [y, [x, x0 ]]
for all x, y ∈ A. Hence ψ(x, y) = [x, [y, x0 ]] is a symmetric bilinear map. So it is enough to prove that
ψ is a derivation in the first component. This is true since
ψ (xy, z) = [xy, [z, x0 ]] = x [y, [z, x0 ]] + [x, [z, x0 ]] y
= xψ (y, z) + ψ (x, z) y
for all x, y, z ∈ A. 

From here onward we shall call biderivations of the form (4.1) extremal biderivations. Such bideri-
vations occur naturally on triangular algebras.

Remark 4.5. A triangular algebra A = Tri(A, M, B) has a nonzero extremal biderivation if and only if
there exists 0 =
/ m0 ∈ M such that [A, A]m0 = 0 = m0 [B, B].

Proof. Let ψ(x, y) = [x, [y, x0 ]] be a nonzero extremal biderivation of A where

x0 = ea0 e + em0 f + fb0 f , a0 ∈ A, b0 ∈ B, m0 ∈ M .
Assume that [[x, y], x0 ] = 0 for all x, y ∈ A. If we substitute x = e and y = emf , where m ∈ M, into
this equality we see that a0 m = mb0 for all m ∈ M. This means that ea0 e + fb0 f ∈ Z (A). We further
substitute x = eae and y = ea e, where a, a ∈ A. Then we obtain that [a, a ]m0 = 0 for all a, a ∈ A.
 D. Benkovič / Linear Algebra and its Applications 431 (2009) 1587–1602 1595

This implies that [A, A]m0 = 0. Similarly by the substitution x = fbf and y = fb f , where b, b ∈ B, we
obtain m0 [B, B] = 0. Since ea0 e + fb0 f ∈ Z (A) it follows that
ψ (x, y) = [x, [y, x0 ]] = [x, [y, m0 ]]
for all x, y ∈ A. Hence we have found an element 0 = / m0 ∈ M satisfying [A, A]m0 = 0 = m0 [B, B].
Now the reverse implication. We suppose that there exists 0 = / m0 ∈ M such that [A, A]m0 = 0 =
m0 [B, B]. Let x0 = em0 f ∈ A. Then x0 ∈
/ Z (A) and
          
a m a m 0 m0 0 a, a m0
[x, y] x0 = , = = 0,
b b 0 0
         
0 m0 a m a m 0 m0 b, b
x0 [x, y] = ,  = =0
0 b b 0

for all a, a ∈ A, b, b ∈ B. Thus [[x, y], x0 ] = 0 for all x, y ∈ A and A has a nonzero extremal biderivation.


Remark 4.6. Let A = Tri(A, M, B) be a triangular algebra. Let I be the left ideal of A generated by all
the commutators [A, A] and let J be the right ideal of B generated by all commutators [B, B]. If I = A or
J = B then A does not have extremal biderivations.

Proof. Without loss we may suppose that the left ideal of the algebra A generated by all the commu-
tators of A is equal to A. Let m0 ∈ M such that [A, A]m0 = 0. Then
⎛ ⎞
 n   
n   

m0 = 1A m0 = ⎝ ai ai , ai ⎠ m0 = ai ai , ai m0 = 0.
i =1 i =1

From Remark 4.5 it now follows that the triangular algebra does not have extremal biderivations.


The following examples show that the upper triangular matrix algebra Tn (C ), the block upper
triangular matrix algebra Bnk̄ (C ) and the nest algebra T (N ) all admit extremal biderivations. These
examples are based on [15, Example 2.1, Theorem 2.2]:

Example 4.7. Let A = Tn (C ) be an upper triangular matrix algebra and let us denote by eij , 1  i  j  n,
the standard matrix units. Then
ψ (x, y) = [x, [y, e1n ]] for all x, y ∈ A
is an extremal biderivation. Namely, if we write A = Tri(C, M1×(n−1) , Tn−1 (C )), then clearly we have
[C, C ]e1n = 0 = e1n [Tn−1 (C ), Tn−1 (C )].

Example 4.8. Let N be a nest on a Hilbert space H. Suppose that there exists N1 , N2 ∈ N \{0, H } such
that dim N1 = 1 and dim N2⊥ = 1. Then the nest algebra T (N ) can be represented in the following
form
⎛    ⎞
C B H  ,C
  C 
T (N ) = ⎝ T N B C, H  ⎠ ,
C
for a suitable subspace H  of H and some nest N  on H  . For each element
⎛ ⎞
0 0 λ
x0 = ⎝ 0 0⎠ , λ ∈ C
0
we have [[T (N ), T (N )], x0 ] = 0. Therefore ψ(x, y) = [x, [y, x0 ]] is an extremal biderivation of T (N ).
1596 D. Benkovič / Linear Algebra and its Applications 431 (2009) 1587–1602

Remark 4.9. A block upper triangular matrix algebra Bnk̄ (C ) has a nonzero extremal biderivation if and
only if k̄ = (1, k2 , . . ., km−1 , 1).

Proof. Let m be a positive integer and let A = Bnk̄ (C ), where k̄ = (1, k2 , . . ., km−1 , 1) ∈ Nm . Note that
matrix unit e1n satisfies [[A, A], e1n ] = 0 and so ψ(x, y) = [x, [y, e1n ]] is an extremal biderivation.
Conversely, we may assume that the corresponding vector k̄ of A = Bnk̄ (C ) is of the form k̄ =
(k1 , k2 , . . ., km ), where k1  2. Now we can write
 
Mk1 (C ) Mk1 ×(n−k1 ) (C )
A = k̄ .
Bn1−k1 (C )

Since the left ideal of Mk1 (C ) which is generated by all commutators [Mk1 (C ), Mk1 (C )] is equal to
Mk1 (C ), Remark 4.6 implies that the algebra A does not have a nonzero extremal biderivation. 

Proposition 4.10. Let A be a triangular algebra and let ϕ : A × A → A be a biderivation. If ϕ(e, e) = / 0,

then ϕ = ψ + θ , where ψ(x, y) = [x, [y, ϕ(e, e)]] is an extremal biderivation and θ is a biderivation that
satisfy θ(e, e) = 0.

Proof. Let ϕ be a biderivation and ϕ(e, e) = / 0. Since [e, e] = 0 we see from Lemma 4.3 that ϕ(e, e) =
eϕ(e, e)f . So ϕ(e, e) ∈
/ Z (A) and it satisfies (use Lemma 4.1)
ϕ (e, e) [x, y] = [e, e] ϕ (x, y) = 0 and
[x, y] ϕ (e, e) = ϕ (x, y) [e, e] = 0
for all x, y ∈ A. So [[x, y], ϕ(e, e)] = 0 for all x, y ∈ A and the map ψ(x, y) = [x, [y, ϕ(e, e)]] is an
extremal biderivation of the triangular algebra A. We note that
ψ (e, e) = [e, [e, ϕ (e, e)]] = [e, [e, eϕ (e, e) f ]]
= eϕ (e, e) f = ϕ (e, e) .
Let θ = ϕ − ψ . Then θ is a biderivation of A satisfying θ(e, e) = ϕ(e, e) − ψ(e, e) = 0. 
From Proposition 4.10 it follows that we may always subtract an extremal biderivation ψ(x, y) =
[x, [y, ϕ(e, e)]] from biderivations ϕ of the triangular algebra A. Henceforth we shall limit ourselves to
those biderivations of triangular algebras for which ϕ(e, e) = 0. The key question now becomes when
are such biderivations inner. In other words what conditions must a triangular algebra satisfy so that
the aforementioned biderivations are inner. The sufficient conditions are supplied by the following
theorem:

Theorem 4.11. Let A = Tri(A, M, B) be a triangular algebra. If the following conditions hold:

(i) πA (Z (A)) = Z (A) and πB (Z (A)) = Z (B),

(ii) at least one of the algebras A and B is noncommutative,
(iii) if α a = 0, α ∈ Z (A), 0 =/ a ∈ A, then α = 0,
(iv) each derivation of A is inner,
then every biderivation ϕ : A × A → A that satisfies ϕ(e, e) = 0 is of the form ϕ(x, y) = λ[x, y], where
λ ∈ Z (A).

Proof. Let x = a + m + b and y = a + m + b be arbitrary elements of A, where a, a ∈ A, b, b ∈ B

and m, m ∈ M. Since ϕ is bilinear it follows that
 
ϕ (x, y) = ϕ a + m + b, a + m + b
         
= ϕ a, a + ϕ a, m + ϕ a, b + ϕ m, a + ϕ m, m
 D. Benkovič / Linear Algebra and its Applications 431 (2009) 1587–1602 1597

       
+ ϕ m, b + ϕ b, a + ϕ b, m + ϕ b, b (4.2)

for all x, y ∈ A. Our goal is to describe all nine parts of the biderivation (4.2). We shall do this in four
steps.
Step 1. We first prove that
ϕ (a, b) = 0 = ϕ (b, a) for all a ∈ A, b ∈ B. (4.3)
According to the assumption ϕ(e, e) = 0 and so using Lemma 4.2 part (iii) it is true that ϕ(e, f ) =
ϕ(f , e) = 0. Since [a, b] = 0 = [b, a], Lemma 4.3 implies that ϕ(a, b) = eϕ(a, b)f and ϕ(b, a) = eϕ(b, a)f
for all a ∈ A, b ∈ B. Now we can conclude that

ϕ (a, b) = eϕ (ae, fb) f = aϕ (e, fb) f + eϕ (a, fb) ef

= af ϕ (e, b) f + aϕ (e, f ) b = 0
and

ϕ (b, a) = eϕ (fb, ae) f = af ϕ (b, ae) f + eϕ (f , ae) b

= aϕ (f , e) b + eϕ (f , a) eb = 0
for all a ∈ A, b ∈ B.
Step 2. We prove that there exists α ∈ πA (Z (A)) such that
ϕ (a, m) = −ϕ (m, a) = α am and (4.4)
ϕ (m, b) = −ϕ (b, m) = α mb
for all a ∈ A, b ∈ B. Let the map f : M → M be defined by f (m) = ϕ(e, m) for all m ∈ M. Then f is a
bimodule homomorphism. Namely for all a ∈ A, b ∈ B, m ∈ M we have that

f (am) = ϕ (e, am) = aϕ (e, m) + ϕ (e, a) m = af (m) and

f (mb) = ϕ (e, mb) = mϕ (e, b) + ϕ (e, m) b = f (m) b,

where we used that ϕ(e, a) = ϕ(e, b) = 0. The assumption (iv) implies that each derivation of the
triangular algebra A is inner, so Corollary 3.4 implies that the bimodule homomorphism f is of the
standard form
= a0 m + mb0 , a0 ∈ Z (A) , b0 ∈ Z (B) .
f (m)
Now we use the assumption (i) to see that a0 ∈ πA (Z (A)) and b0 ∈ πB (Z (A)). We may write
 
ϕ (e, m) = f (m) = a0 + τ −1 (b0 ) m = α m for all m ∈ M,
where α = a0 + τ −1 (b0 ) ∈ πA (Z (A)).
Similarly the map g : M → M defined by g (m) = ϕ(m, e) for all m ∈ M is a bimodule homomor-
phism. So there exists β ∈ πA (Z (A)) so that ϕ(m, e) = β m for all m ∈ M. We prove that
ϕ (e, m) = α m = −ϕ (m, e) for all m ∈ M . (4.5)
It is sufficient to prove that α + β = 0. According to condition (ii) we may assume that A is a noncom-
mutative algebra. Choose a, a ∈ A such that [a, a ] =
/ 0. Since ϕ(e, m) = α m in ϕ(m, e) = β m we see
using Lemma 4.1 that
   
ϕ a, a [e, m] = a, a ϕ (e, m) = α a, a m and
   
ϕ a, a [m, e] = a, a ϕ (m, e) = β a, a m

for all m ∈ M. Adding the above equalities yields (α + β)[a, a ]M = 0. The faithfulness of the left
A-module M now implies (α + β)[a, a ] = 0. Since [a, a ] =
/ 0 we conclude using the condition (iii)
that α + β = 0.
1598 D. Benkovič / Linear Algebra and its Applications 431 (2009) 1587–1602

Further using (4.5) and ϕ(e, m) + ϕ(f , m) = 0 and ϕ(m, e) + ϕ(m, f ) = 0 we see that
ϕ (m, f ) = α m = −ϕ (f , m) for all m ∈ M . (4.6)
Let a ∈ A and m ∈ M be arbitrary elements. Then we may write
ϕ (a, m) = ϕ (ae, m) = aϕ (e, m) + ϕ (a, m) e = aϕ (e, m) ,
since the equality ϕ(a, f ) = 0 implies
ϕ (a, m) e = ϕ (a, mf ) e = mϕ (a, f ) e + ϕ (a, m) fe = 0.
Now use (4.5) to get ϕ(a, m) = aϕ(e, m) = α am. This proves the first equality (4.4). The other three
can be proven with a similar calculation using (4.5) and (4.6).
Step 3. Let us prove that for all a, a ∈ A and b, b ∈ B we have
     
ϕ a, a = α a, a and ϕ b, b = τ (α) b, b . (4.7)

Using that ϕ(a, e) = 0 = ϕ(e, a) for all a ∈ A we see

       
ϕ a, a = ϕ e (ae) , a = eϕ ae, a + ϕ e, a ae
     
= eaϕ e, a + eϕ a, a e = eϕ a, a e.

Hence ϕ(a, a ) ∈ eAe = A. Use Lemma 4.1 and (4.5) to get

   
ϕ a, a [e, m] = a, a ϕ (e, m) = α a, a m
for all m ∈ M. This means (ϕ(a, a ) − α[a, a ])M = 0 and from the faithfulness of the left A-module
M we obtain ϕ(a, a ) = α[a, a ] for all a, a ∈ A.
Similarly ϕ(b, b ) ∈ f Af = B and
   
[e, m] ϕ b, b = ϕ (e, m) b, b = α m b, b

for all m ∈ M. Hence M (ϕ(b, b ) − τ (α)[b, b ]) = 0. The faithfulness of the right B-module M now
implies the desired result.
Step 4. The only remaining equality we need to prove is
ϕ (m, n) = 0 for all m, n ∈ M . (4.8)
Since [m, n] = 0 it follows that ϕ(m, n) = eϕ(m, n)f . Fix n ∈ M. Then the map h : M → M defined by
h(m) = ϕ(m, n) for all m ∈ M is a bimodule homomorphism. Using (4.4) we see that

h (am) = ϕ (am, n) = aϕ (m, n) + ϕ (a, n) m

= aϕ (m, n) + α anm = ah (m)
and

h (mb) = ϕ (mb, n) = mϕ (b, n) + ϕ (m, n) b

= −mnτ (α) b + ϕ (m, n) b = h (m) b
for all a ∈ A, b ∈ B, m ∈ M. According to (iv), Corollary 3.4 and (i) we may conclude that there exists
αn ∈ Z (A) such that
ϕ (m, n) = h (m) = αn m for all m ∈ M . (4.9)
Without loss of generality we again assume that A is a noncommutative algebra and let a, a ∈ A be
fixed elements with nonzero commutator. Using Lemma 4.1 and (4.9) we may write
   
0 = ϕ a, a [m, n] = a, a ϕ (m, n) = αn a, a m
 D. Benkovič / Linear Algebra and its Applications 431 (2009) 1587–1602 1599

for all m, n ∈ M. The faithfulness of the left A-module M implies αn [a, a ] = 0 and from the assumption
(iii) it follows that αn = 0 for all n ∈ M. Therefore ϕ(m, n) = αn m = 0 for all m, n ∈ M.
At the end let λ = α + τ (α) ∈ Z (A). Substituting (4.3), (4.4), (4.7) and (4.8) into the formula (4.2)
leads to
 
ϕ (x, y) = α a, a + α am − α a m + α mb − α m b + τ (α) b, b
      
α 0 a m a m
= ,
τ (α) b b
= λ [x, y]
for all x, y ∈ A. Hence every biderivation ϕ such that ϕ(e, e) = 0 is an inner biderivation. This com-
pletes the proof. 

If a triangular algebra A satisfies the conditions of the main Theorem 4.11, then Proposition 4.10
implies that every biderivation of the algebra A is a sum of an extremal and an inner biderivation.

Remark 4.12. In Theorem 4.11, the condition (iv) may be replaced by a weaker condition that each
bimodule homomorphism f : M → M is of the form
f (m) = am + mb, a ∈ Z (A) , b ∈ Z (B)
for all m ∈ M.

Two corollaries describing biderivations of the (block) upper triangular matrix algebra and the
nest algebra follow directly from the main theorem. As mentioned in the introduction the relevant
examples were given in [15,17].

Corollary 4.13. Let C be a commutative domain with identity. If n  3, then each biderivation of Bnk̄ (C ) is
the sum of an extremal biderivation and an inner biderivation. In particular, every biderivation of Tn (C ) is
the sum of an extremal biderivation and an inner biderivation.

Proof. Let be n  3. If Bnk̄ (C ) = Mn (C ) then every biderivation of the matrix algebra Mn (C ) is inner
biderivation (namely Mn (C ) is a noncommutative prime algebra and so this result follows from [5,
Theorem 3.5]). Now, let Bnk̄ (C ) =
/ Mn (C ). Let us prove that the algebra Bnk̄ (C ) satisfies all the conditions
of Theorem 4.11. The block upper triangular matrix algebra can be seen as a triangular algebra of the
form
⎛ ⎞
  k̄1
A M B (C ) M l×(n−l) (C )
B
=⎝ l k̄
⎠.
Bn1−l (C )

Since Z (Bnk̄ (C )) = C1, Z (A) = C1A and Z (B) = C1B , the first condition is satisfied. Since we have
n  3, the second  condition
 is also satisfied (namely at least one of the algebras A and B is noncom-
mutative). As Z Bnk̄ (C ) = C1 and C is a commutative domain, the algebra Bnk̄ (C ) also satisfies the
condition (iii). It is also true that every derivation of the algebra Bnk̄ (C ) is inner (see Proposition 3.6).
Hence, using Theorem 4.11 and Proposition 4.10 we conclude that every biderivation of Bnk̄ (C ) is a sum
of an extremal biderivation and an inner biderivation. Clearly upper triangular matrix algebra Tn (C ) is
a special case of a block upper triangular matrix algebra. 

Corollary 4.14. Let N be a nest of a Hilbert space H, where dim H  3. Then each biderivation ϕ of nest
algebra T (N ) is the sum of an extremal biderivation and an inner biderivation.
1600 D. Benkovič / Linear Algebra and its Applications 431 (2009) 1587–1602

Proof. Let N be a nest on a Hilbert space H. If N = {0, H } then T (N ) = B(H ). Since B(H ) is a non-
commutative centrally closed prime algebra it follows by [5, Theorem 3.5] that every biderivation of
B(H ) is inner. In this case the algebra B(H ) does not have extremal biderivations.
Let N be a nontrivial nest. Then there exists 0 = / N ∈ N and the nest algebra can be seen as a
triangular algebra
 
T (N1 ) M
T (N ) = ,
T (N2 )

where A = T (N1 ) and B = T (N2 ) are the corresponding nest algebras. Since dim H  3 at least one
of the algebras A or B is noncommutative. Since nest algebras are central algebras we have that
πA (Z (T (N ))) = Z (A) = C1A and πB (Z (T (N ))) = Z (B) = C1B . Since Z (T (N )) = C1 and C is a field,
a nest algebra satisfies the condition (iii) of Theorem 4.11. The condition (iv) is also satisfied because
every derivation of a nest algebra is inner [12, Theorem 19.7]. Hence using Theorem 4.11 and Proposition
4.10 we conclude that every biderivation of a nest algebra is a sum of an extremal biderivation and an
inner biderivation. 

Remark 4.15. Let A be a commutative algebra with identity and let A = Tri(A, A, A) be a triangular
algebra. A map ϕ : A × A → A, defined by
     
a m a m 0 mm
ϕ b  =
b 0

is a biderivation, which is not the sum of an extremal biderivation and and inner biderivation.

Proof. The map ϕ is bilinear and symmetric. It can be verified by a direct calculation that ϕ is a
biderivation. We have ϕ(e12 , e12 ) = e12 . Let θ(x, y) = [x, [y, x0 ]] + λ[x, y] be a sum of an extremal
and inner biderivation, where x0 ∈ A and λ ∈ Z (A). Then
θ (e12 , e12 ) = [e12 , [e12 , x0 ]] + λ [e12 , e12 ] = 0.
Therefore the biderivation ϕ cannot be written as a sum of an extremal and inner biderivation. 

Especially note that if in the above example the algebra A is a commutative ring C with an identity,
then the upper triangular matrix algebra T2 (C ) supports a biderivation that is not a sum of an extremal
and an inner biderivation. This means that Corollary 4.13 does not hold for n = 2. Similarly, Corollary
4.14 does not hold in the case dim H = 2.
How does a biderivation of T2 (C ) looks like? Can any biderivation of T2 (C ) be written as the sum of
an inner derivation, an extremal biderivation and a biderivation from Remark 4.15? Let {e11 , e12 , e22 }
be the system of matrix units of T2 (C ). We denote by x = a11 e11 + a12 e12 + a22 e22 and y = b11 e11 +
b12 e12 + b22 e22 arbitrary elements of T2 (C ). First we gather together basic examples of biderivations
of the algebra T2 (C ). Let
ϕ1 (x, y) = α(a11 − a22 )(b11 − b22 )e12 , α ∈ C (4.10)
be an extremal biderivation of T2 (C ), which is generated by an element α e12 , this means ϕ1 (x, y) =
α[x, [y, e12 ]]. Let
ϕ2 (x, y) = β a12 b12 e12 , β ∈ C (4.11)
be a biderivation of T2 (C ) that is defined in Remark 4.15. By a direct calculation one can easily prove
that there are another two basic types of biderivations of algebra T2 (C ):

ϕ3 (x, y) = γ (a11 − a22 )b12 e12 , γ ∈ C, (4.12)

ϕ4 (x, y) = δ(b11 − b22 )a12 e12 , δ ∈ C . (4.13)

Moreover, ϕ3 and ϕ4 are not sums of an inner derivation, an extremal biderivation and a biderivation of
form (4.11). In a special case if γ + δ = 0, then ϕ3 + ϕ4 is an inner biderivation. The next proposition
 D. Benkovič / Linear Algebra and its Applications 431 (2009) 1587–1602 1601

tells us that a biderivation of the algebra T2 (C ) can be written as a sum of the aforementioned four
basic types of biderivations. Since the proof of this proposition is elementary and long we omit it here.

Proposition 4.16. Let ϕ be a biderivation of T2 (C ). Then there exist biderivations ϕ1 , ϕ2 , ϕ3 , ϕ4 of forms

(4.10), (4.11), (4.12), (4.13), respectively, such that ϕ = ϕ1 + ϕ2 + ϕ3 + ϕ4 .

Finally, we shall answer the question on how to construct a noninner biderivation of a (noncom-
mutative) triangular algebra. Recall that a map f : A →A is commuting if [f (x), x] = 0 for all x ∈ A.
In the study of commuting maps we are interested in the question whether they are of a standard form
f (x) = λx + μ (x) for all x ∈ A,
where λ ∈ Z (A) and μ : A →Z (A) are linear maps. Here we need to highlight the following fact:
If the equality [f (x), x] = 0 is linearized we obtain [f (x), y] = [x, f (y)] for all x, y ∈ A. With a direct
calculation we see that the map ϕ : A × A → A defined by
ϕ (x, y) = [f (x) , y] = [x, f (y)] for all x, y ∈ A, (4.14)
is a biderivation.

Remark 4.17. If each biderivation of A is inner, then every commuting map of A is of the standard
form.

Proof. Let f : A →A be a commuting map. The assumption implies that ϕ(x, y) = [f (x), y] is an inner
biderivation. Therefore [f (x), y] = λ[x, y] or, in other words, [f (x) − λx, A] = 0 for all x ∈ A. Hence
μ(x) = f (x) − λx ∈ Z (A) and f is of the standard form. 

The last remark implies that if on the triangular algebra A there exists a commuting map f that is
not of the standard form, then Eq. (4.14) yields a biderivation that is not inner. Commuting maps on
triangular algebras where treated in detail by Cheung [9]. We conclude with an example of a triangular
algebra that admits a biderivation that is not a sum of an extremal and an inner biderivation.

Example 4.18. Let C be a commutative ring with identity and let

  
a t
A= ; a, t ∈ C
a

be a subalgebra of T2 (C ). We consider a triangular algebra A = Tri(A, T2 (C ), A). If f : A → A is one

of commuting maps that are defined in [9, Example 1], then (4.14) can be used to get an example of a
biderivation ϕ that is not an inner biderivation. It can also be verified that the obtained biderivation
ϕ is not a sum of an extremal and an inner biderivation.

Aknowledgement

The author is thankful to the referee for careful reading of the paper and useful suggestions.

References

[1] D. Benkovič, D. Eremita, Commuting traces and commmutativity preserving maps on triangular algebras, J. Algebra, 280
(2004) 797–824.
[2] M. Brešar, Commuting mappings on von Neumann algebras, Proc. Amer. Math. Soc. 111 (1991) 501–510.
[3] M. Brešar, Centralizing mappings and derivations in prime rings, J. Algebra 156 (1993) 385–394.
[4] M. Brešar, On generalized biderivations and related maps, J. Algebra 172 (1995) 764–786.
[5] M. Brešar, Commuting maps: A survey, Taiwanese J. Math 8 (2004) 361–397.
[6] M. Brešar, M.A. Chebotar, W.S. Martindale 3rd, Functional Identities, Birkhaüser Verlag, 2007.
[7] M. Brešar, W.S. Martindale 3rd, C.R. Miers, Centralizing maps in prime rings with involution, J. Algebra 161 (1993) 342–357.
[8] S.P. Coelho, C.P. Milies, Derivations of upper triangular matrix rings, Linear Algebra Appl. 187 (1993) 263–267.
[9] W.-S. Cheung, Commuting maps of triangular algebras, J. London Math. Soc. 63 (2001) 117–127.
1602 D. Benkovič / Linear Algebra and its Applications 431 (2009) 1587–1602

[10] W.-S. Cheung, Lie derivations of triangular algebras, Linear Multilinear Algebra 51 (2003) 299–310.
[11] W.-S. Cheung, Mappings on Triangular Algebras, PhD Dissertation, University of Victoria, 2000.
[12] K.R. Davidson, Pitman Research Notes in Mathematics Series, Longman, Harlow, 1988.
[13] B.E. Forrest, L.W. Marcoux, Derivations of triangular Banach algebras, Indiana Univ. Math. J. 45 (1996) 441–462.
[14] T.-L. Wong, Jordan isomorphisms of triangular rings, Proc. Amer. Math. Soc. 133 (2005) 3381–3388.
[15] J.-H. Zhang, S. Feng, H.-X. Li, R.-H. Wu, Generalized biderivations of nest algebras, Linear Algebra Appl. 418 (2006) 225–233.
[16] J.-H. Zhang, W.-Y. Yua, Jordan derivations of triangular algebras, Linear Algebra Appl. 419 (2006) 251–255.
[17] Y. Zhao, D. Wang, R. Yao, Biderivations of upper triangular matrix algebras over commutative rings, in press.