You are on page 1of 29

Some Notes On Clean Rings

Nicholas A. Immormino

Clean Rings

A ring is said to be clean if every element in the ring can be written as the sum of a unit and
an idempotent of the ring. More generally, an element in a ring is called clean if it can be written
as the sum of a unit and an idempotent of the ring. These rings were introduced by Nicholson [34]
in his study of lifting idempotents and exchange rings. In these notes we review some well-known
properties and examples of clean rings, and then we classify the rings that belong to an interesting
subclass of clean rings. For a more general introduction to clean rings, see the survey of clean rings
given by Nicholson and Zhou [37]. For a different perspective on this class of rings, see the history
of commutative clean rings given by McGovern [29].

1 Some Properties of Clean Rings


In this section we review some useful properties of clean rings. We begin
\prod by showing that every
homomorphic image of a clean ring is clean, that a direct product of rings Ri is clean if and only
if each of the rings Ri is clean, and that a full matrix ring Mn (R) is clean if the underlying ring R
is clean. Then we consider the notion of lifting idempotents modulo an ideal, and we prove that a
ring R is clean if and only if for any ideal I of R such that I \subseteq J(R) the quotient ring R/I is clean
and idempotents lift modulo I. These results are well-known, but proofs are included for the sake
of completeness.

Basic Properties
The next two results were proved by Anderson and Camillo [1].
(1.1) Proposition. Every homomorphic image of a clean ring is clean.
Proof. Since multiplication is preserved by every ring homomorphism, the homomorphic image of
a unit (resp. idempotent) is a unit (resp. idempotent) of its ring. Since addition is also preserved
by every ring homomorphism, the result follows.
(1.2) Proposition. A direct product of rings

\prod

Ri is clean if and only if each ring Ri is clean.

Proof. Since multiplication in a direct product of rings is defined componentwise, an element in


a direct product of rings is a unit (resp. idempotent) of that ring if and only if the entry in each of
its components is a unit (resp. idempotent) of its ring. Since addition in a direct product of rings
is also defined componentwise, the result follows from a simple computation.
The next result is due to Han and Nicholson [18].
1

2
(1.3) Theorem. A full matrix ring Mn (R) is clean if the underlying ring R is clean.
Proof. Han and Nicholson [18, Theorem] showed that if the identity 1 of a ring R can be written
as a finite sum 1 = e1 + e2 + \cdot \cdot \cdot + en of mutually orthogonal idempotents ei such that each corner
ring ei Rei is clean, then the ring R is clean---the interested reader is encouraged to see their proof.
Since the set of matrix units \{ Eii \} ni=1 is a complete set of orthogonal idempotents for Mn (R) with
each corner ring Eii Mn (R)Eii isomorphic to R, the result follows.

Lifting Idempotents
For an ideal I of a ring R, we say that idempotents lift modulo \bfitI if for each element x \in R
such that x - x2 \in I there is some idempotent e of R with e - x \in I. Meanwhile, a ring is said to
be exchange if it satisfies the exchange property of Crawley and J\'
onsson [13] when regarded as a
left module over itself. Exchange rings were introduced by Warfield [46], and he proved that their
definition is left-right symmetric. The interested reader is encouraged to see either [13] or [46] for
the formal definition of the exchange property. In his study of lifting idempotents and exchange
rings, Nicholson [34] called a ring suitable if idempotents lift modulo every left ideal of the ring,
and he proved that this definition is left-right symmetric. Moreover, he showed that these suitable
rings coincide with the exchange rings of Warfield. In other words, a ring is exchange if and only if
idempotents lift modulo every left (equivalently, right) ideal of the ring. Nicholson [34] introduced
clean rings as an example of a class of exchange rings when he proved the following.
(1.4) Proposition. (1) Every clean ring is exchange. (2) An abelian ring is clean if and only if
it is exchange.
Proof. (1) Let R be a clean ring, let x be any element in R, and let I be any left ideal of R such
that x - x2 \in I. As we mentioned above, a ring is exchange if and only if idempotents lift modulo
every left ideal of the ring according to Nicholson [34]. Thus it suffices to show that there is some
idempotent e of R with e - x \in I. Since R is a clean ring, we can write x = u + f for some unit u
and idempotent f of R. Let v denote the inverse of u, and check that e = v(1 - f )u is idempotent
with e - x = v(x - x2 ) \in I as desired. Therefore every clean ring is exchange.
(2) This proof is omitted---see Nicholson [34, Proposition 1.8].
(1.5) Corollary. Idempotents lift modulo every left ideal of a clean ring.
Proof. This result follows from (1.4) since idempotents lift modulo every left (equivalently, right)
ideal of an exchange ring according to Nicholson [34].
For an ideal I of a ring R, we say that units lift modulo \bfitI if for each element x \in R such that
x + I is a unit of the quotient ring R/I there is some unit u of R with u - x \in I. It can be shown
that for any ring R and any ideal I \subseteq J(R), an element x + I is a unit of R/I if and only if x is a
unit of R. In particular, this tells us that units lift modulo any ideal of a ring R that is contained
in its Jacobson radical. We use this fact to prove the following characterization of clean rings due
to Han and Nicholson [18].
(1.6) Theorem. Let I be any ideal of a ring R such that I \subseteq J(R). Then R is clean if and only
if the quotient ring R/I is clean and idempotents lift modulo I.

3
Proof. Suppose that R is a clean ring. Then R/I is clean as a result of (1.1), while idempotents
lift modulo I by (1.5). This proves the necessity. Conversely, suppose that the quotient ring R/I is
clean and that idempotents lift modulo I, and let r be any element in R. Since R/I is a clean ring,
we can write r + I = x + e + I for some unit x + I and idempotent e + I of R/I. Moreover, since
idempotents lift modulo I, we can assume that e is an idempotent of R. Since I \subseteq J(R), we also
know that units lift modulo I. In particular, since r - e + I = x + I is a unit of R/I, this tells us
that r - e is a unit of R. It follows that r can be written as the sum of a unit and an idempotent
of R by writing r = (r - e) + e. This proves the sufficiency. Therefore R is clean if and only if the
quotient ring R/I is clean and idempotents lift modulo I.
It is well known that idempotents lift modulo every nil ideal of a ring---for an elementary proof
of this important classical result, the interested reader should see Koh [24]. Since every nil ideal of
a ring R is contained in its Jacobson radical, the following result is a corollary to (1.6).
(1.7) Corollary. Let N be any nil ideal of a ring R. Then R is clean if and only if the quotient
ring R/N is clean.
Proof. As noted above, this result follows from (1.6) since every nil ideal of a ring R is contained
in its Jacobson radical, and idempotents lift modulo every nil ideal.

Two More Properties


A ring R is said to be semipotent if each left ideal of R that is not contained in its Jacobson
radical contains a nonzero idempotent---this notion is left-right symmetric. It is not difficult to see
that a ring R is semipotent if and only if for each a \in R that is not contained in J(R) there exists
some nonzero x \in R such that xax = x. A semipotent ring R is said to be potent if idempotents
lift modulo its Jacobson radical---these rings are sometimes called \bfitI -rings, while semipotent rings
are sometimes known as \bfitI \bfzero -rings. Nicholson [34, Proposition 1.9] proved that every exchange ring
is potent. Since every clean ring is exchange by (1.4), we have the following.
clean

=\Rightarrow

exchange

=\Rightarrow

potent

=\Rightarrow

semipotent

These implications are known to be irreversible. In particular, Camillo and Yu [7, p. 4746] proved
that the ring in a well-known example due to Bergman [19, Example 1] is exchange but not clean.
Meanwhile, Nicholson [34, p. 272] showed that a potent ring need not be exchange, and Nicholson
and Zhou [38, Example 25] showed that a semipotent ring need not be potent. Following the lead
of Han and Nicholson [18], we give a direct proof that every clean ring is semipotent.
(1.8) Proposition. Every clean ring is semipotent.
Proof. Let R be any clean ring, and let I be any left ideal of R such that I \nsubseteq J(R). We need to
show that I contains a nonzero idempotent. It is well known that every quasiregular left ideal of a
ring R is contained in its Jacobson radical. Since I \nsubseteq J(R), this tells us that there is some element
a \in I that is not quasiregular. Since R is a clean ring, we can write a = u + e for some unit u and
idempotent e of R. Let v denote the inverse of u, and check that v(1 - e)u is idempotent and that
v(1 - e)u = v(1 - e)(a - e) = v(1 - e)a is an element of I. Meanwhile, notice that e - a = - u is a
unit of R. Since a is not quasiregular, this tells us that e \not = 1. Since u and v are both units of R, it

4
follows that v(1 - e)u \not = 0, and hence v(1 - e)u is a nonzero idempotent of I. Therefore every clean
ring is semipotent.
(1.9) Corollary. Every clean ring is potent.
Proof.

This result follows immediately from (1.8) and (1.5).

A nonzero idempotent e of a ring R is called a primitive idempotent if the corner ring eRe
contains no nontrivial idempotents---in other words, if the ring eRe contains no idempotents other
than 0 and e. It can be shown that a nonzero idempotent e of a ring R is a primitive idempotent
if and only if for any nonzero idempotent f of R the inclusion f \in Re implies that Rf = Re. This
fact was pointed out by Nicholson [33, p. 346]. Meanwhile, an idempotent e of a ring R is called a
local idempotent if the corner ring eRe is a local ring. Since a local ring contains no nontrivial
idempotents, every local idempotent is primitive. The following result due to Nicholson [33] shows
that the converse is true for any semipotent ring.
(1.10) Proposition. Every primitive idempotent in a semipotent ring is local.
Proof. Let R be a semipotent ring, and let e be any primitive idempotent of R. We need to show
that e is a local idempotent of R. In particular, we need to show that the corner ring eRe is a local
ring. Let a be any element in the ring eRe that is not contained in J(eRe). It is well known that
J(eRe) = J(R) \cap eRe for any idempotent e of a ring R---see Lam [25, Theorem 21.10]. It follows
that a is not contained in J(R), and hence Ra \nsubseteq J(R). Since R is a semipotent ring, this tells us
that the left ideal Ra contains some nonzero idempotent f . Notice that Rf \subseteq Ra \subseteq Re. Since e is
a primitive idempotent of R, this implies that Rf = Re, which forces Ra = Re. In particular, this
tells us that e is contained in Ra, and hence there is some element x \in R such that xa = e. Since
a is an element of eRe, it follows that (exe)a = e, and hence a has a left inverse in eRe. Since this
is the case for every element of eRe that is not contained in J(eRe), it follows that the corner ring
eRe is a local ring. Therefore every primitive idempotent of R is local.
(1.11) Corollary. Every primitive idempotent in a clean ring is local.
Proof.

This result follows immediately from (1.8) and (1.10).

Some Notes
(1) Let R be any ring, and let e be any idempotent of R with complement f = 1 - e. Han and
Nicholson [18, Lemma] showed that if the corner rings eRe and f Rf are both clean, then R is also
\v
clean. The converse is not true in general according to Ster
[40, Example 3.4]. In other words, the
corner rings of a clean ring need not be clean. However, the corner rings of any exchange ring are
exchange by [34, Proposition 1.10], and the corner rings of any potent ring (resp. semipotent ring)
are potent (resp. semipotent) by [33, Corollary 1.7].
(2) An idempotent e of a ring R is said to be a full idempotent if ReR = R. It is not known
if the corner rings of a clean ring that arise from a full idempotent of the ring are clean. It is also
not known if the converse of (1.3) is true---in other words, it is not known if the underlying ring R
must be clean in order for a full matrix ring Mn (R) to be clean. Moreover, it is not known if being
clean is Morita invariant.

5
(3) A ring R is called quasinormal if eRf Re = 0 for every noncentral idempotent e of R with
complement f = 1 - e. It is easy to check that every abelian ring is quasinormal, while the upper
triangular matrix ring T2 (\BbbZ /2\BbbZ ) is quasinormal but not abelian. Therefore this notion is a proper
generalization of that of an abelian ring. We know from (1.4) that an abelian ring is clean if and
only if it is exchange. More generally, Wei and Li [47, Proposition 4.1] showed that a quasinormal
ring is clean if and only if it is exchange.
(4) A ring is called quasi-duo if every one-sided maximal ideal of the ring is in fact two-sided.
Yu [49, Theorem 4.2] showed that a quasi-duo ring is clean if and only if it is exchange. As it turns
out, this result generalizes the above result of Wei and Li since every quasinormal exchange ring is
quasi-duo according to [47, Theorem 3.12].
(5) A ring R is said to have artinian primitive factors if the quotient ring R/U is artinian
for every primitive ideal U of R---the interested reader should see Lam [25, Definition 11.3] for the
definition of a primitive ideal. Chen [8, Theorem 1] showed that a ring with artinian primitive
factors is clean if and only if it is exchange.
(6) A ring is said to be root clean if each element in the ring can be written as the sum of a
unit and a square root of 1. Hiremath and Hegde [20, Proposition 2.12] showed that a ring is root
clean if and only if it is a clean ring in which 2 is invertible. The sufficiency was proved earlier by
Camillo and Yu [7]. More specifically, they showed that a ring in which 2 is invertible is clean if
and only if it is root clean---see [7, Proposition 10].
(7) As we mentioned above, Camillo and Yu [7, p. 4746] proved that the ring in a well-known
example due to Bergman [19, Example 1] is exchange but not clean. More specifically, the ring in
Bergman's example is an exchange ring in which 2 is invertible that is not root clean. Since a ring
in which 2 is invertible is clean if and only if it is root clean according to [7, Proposition 10], the
ring in Bergman's example is indeed exchange but not clean. The interested reader is encouraged
to see [7], [19], [27], and [40] for more on this example. No other examples of exchange rings that
are not clean are known to the author.

2 Some Examples of Clean Rings


In this section we look at several important classes of rings whose rings are known to be clean.
These include semiperfect rings, unit regular rings, and strongly \pi -regular rings. We review some
useful properties of these rings and of some closely related rings. In particular, we are interested
in the relationships between these rings and clean rings. All rings of interest are defined below.

Division, Boolean \& Local Rings


The following lemma due to Anderson and Camillo [1] shows that the units, idempotents, and
quasiregular elements of any ring are clean. The immediate result of this lemma is that every ring
that consists entirely of these types of elements is a clean ring. We study this interesting subclass
of clean rings in \S 3.

6
(2.1) Lemma. The units, idempotents, and quasiregular elements of any ring are clean.
Proof. Since 0 is idempotent, any unit u can be written as the sum of a unit and an idempotent
by writing u = u + 0. Meanwhile, notice that 2e - 1 is a square root of 1 (hence a unit) when e is
idempotent. Since 1 - e is idempotent when e is an idempotent, any idempotent e can be written
as the sum of a unit and an idempotent by writing e = (2e - 1) + (1 - e). Finally, since 1 - x is a
unit by definition when x is quasiregular, any quasiregular element x can be written as the sum of
a unit and an idempotent by writing x = - (1 - x) + 1. This completes the proof.
(2.2) Corollary. Every division ring, boolean ring, and local ring is clean.
Proof. This follows immediately from (2.1) since every division ring, boolean ring, and local ring
consists entirely of units, idempotents, and quasiregular elements.
The following theorem shows the precise relationship between local rings and clean rings (resp.
exchange rings, potent rings, semipotent rings). In particular, it tells us that a ring is local if and
only if it is clean (resp. exchange, potent, semipotent) and has no nontrivial idempotents. The first
direct proof that (1) \leftrightarrow (2) below was given by Anderson and Camillo [1]. The fact that (1) \leftrightarrow (3)
was proved earlier by Warfield [45], and the equivalences (1) \leftrightarrow (4) and (1) \leftrightarrow (5) were essentially
proved by Nicholson [33].
(2.3) Theorem. The following are equivalent for any ring R.
(1)
(2)
(3)
(4)
(5)

R
R
R
R
R

is
is
is
is
is

local;
clean and has no nontrivial idempotents;
exchange and has no nontrivial idempotents;
potent and has no nontrivial idempotents;
semipotent and has no nontrivial idempotents.

Proof. It is not difficult to see that a local ring has no nontrivial idempotents. Since every local
ring is clean by (2.2), it follows that (1) \Rightarrow (2). Meanwhile, recall that every clean ring is exchange
by (1.4), that every exchange ring is potent according to Nicholson [34, Proposition 1.9], and that
every potent ring is semipotent by definition. This tells us that (2) \Rightarrow (3) \Rightarrow (4) \Rightarrow (5). It remains
only for us to show that every semipotent ring with no nontrivial idempotents is local. Recall that
every primitive idempotent in a semipotent ring is local according to (1.10). Since the identity 1 is
a primitive idempotent in any ring with no nontrivial idempotents, it follows that (5) \Rightarrow (1). This
completes the proof.
(2.4) Corollary. An integral domain is clean if and only if it is local.
Proof. This result follows immediately from (2.3) since an integral domain contains no nontrivial
idempotents.

Perfect \& Semiperfect Rings


A ring R is said to be left artinian if every descending chain of left ideals in R has a minimal
element---in other words, if for every descending chain I1 \supseteq I2 \supseteq I3 \supseteq \cdot \cdot \cdot of left ideals in the ring
there is some positive integer n such that In = In+1 = In+2 = \cdot \cdot \cdot . To be right artinian requires
instead that this property holds for every descending chain of right ideals in the ring. We will say

7
that a ring is artinian if it is both left and right artinian---this notion is not left-right symmetric.
It is well known that the Jacobson radical of every left (resp. right) artinian ring is nilpotent, and
hence every nil one-sided ideal in a left (resp. right) artinian ring is nilpotent.
A left artinian ring is said to be semisimple if its Jacobson radical is equal to 0. As it turns
out, this notion is left-right symmetric. In particular, it follows that any left or right artinian ring
whose Jacobson radical is equal to 0 is (left and right) artinian. It is not difficult to see that every
homomorphic image of a left (resp. right) artinian ring is left (resp. right) artinian, and hence the
quotient ring R/J(R) is semisimple for any left (resp. right) artinian ring R. It is well known that
a commutative ring is semisimple if and only if it is isomorphic to a finite direct product of fields.
(2.5) Proposition. Every semisimple ring is clean.
Proof. It is well known that every semisimple ring is isomorphic to a finite direct product of full
matrix rings over division rings---this is the Wedderburn-Artin Theorem. Since every division ring
is clean by (2.2), every full matrix ring with entries from a division ring is clean as a result of (1.3).
Since every direct product of clean rings is clean by (1.2), the result now follows. Therefore every
semisimple ring is clean.
(2.6) Proposition. Every left (right) artinian ring is clean.
Proof. Let R be any left (right) artinian ring. Then its Jacobson radical is nilpotent (hence nil),
and it follows from (1.7) that the ring R is clean if and only if R/J(R) is clean. Since the quotient
ring R/J(R) is semisimple, as mentioned above, for any left (right) artinian ring R, the result now
follows from (2.5). Therefore every left (right) artinian ring is clean.
(2.7) Corollary. Every finite ring is clean.
Proof.

This result follows immediately from (2.6) since every finite ring is artinian.

A ring R is said to be semiprimary if R/J(R) is semisimple and J(R) is nilpotent. Since the
Jacobson radical of every left (resp. right) artinian ring is nilpotent, it follows that every left (resp.
right) artinian ring is semiprimary.
(2.8) Proposition. Every semiprimary ring is clean.
Proof. Let R be any semiprimary ring. Then its Jacobson radical is nilpotent (hence nil), and it
follows from (1.7) that the ring R is clean if and only if R/J(R) is clean. Since the quotient ring
R/J(R) is semisimple by definition for any semiprimary ring R, the result now follows from (2.5).
Therefore every semiprimary ring is clean.
A subset of a ring R is called left \bfitT -nilpotent if for any sequence \{ a1 , a2 , a3 , . . . \} of elements
in the subset there is some positive integer n such that a1 a2 \cdot \cdot \cdot an = 0. To be right \bfitT -nilpotent
requires instead that an \cdot \cdot \cdot a2 a1 = 0. We will call a subset \bfitT -nilpotent if it is both left and right
T -nilpotent---this notion is not left-right symmetric. It is not difficult to see that the implications
below hold for any one-sided ideal of a ring.
nilpotent

=\Rightarrow

T -nilpotent

=\Rightarrow

left T -nilpotent

=\Rightarrow

nil

A ring R is said to be left perfect if the quotient ring R/J(R) is semisimple and J(R) is left
T -nilpotent. To be right perfect requires instead that J(R) is right T -nilpotent. We will say that

8
a ring is perfect if it is both left and right perfect---this notion is not left-right symmetric. Since
every nilpotent ideal of a ring is both left and right T -nilpotent, it follows that every semiprimary
ring is perfect. It is well known that a commutative ring is perfect if and only if it is isomorphic to
a finite direct product of (commutative) local rings each of which has a T -nilpotent maximal ideal.
In general, Bass [4, Theorem P] showed that a ring is left (resp. right) perfect if and only if every
descending chain of principal right (resp. left) ideals in the ring has a minimal element---notice the
unusual switch from left to right.
(2.9) Proposition. Every left (right) perfect ring is clean.
Proof. Let R be any left (resp. right) perfect ring. Then its Jacobson radical is left (resp. right)
T -nilpotent (hence nil), and it follows from (1.7) that the ring R is clean if and only if R/J(R) is
clean. Since R/J(R) is semisimple by definition for any left (right) perfect ring R, the result now
follows from (2.5). Therefore every left (right) perfect ring is clean.
A ring R is said to be semiperfect if the quotient ring R/J(R) is semisimple and idempotents
lift modulo J(R). Since every semisimple ring is clean according to (2.5), it follows from (1.6) that
every semiperfect ring is clean. It is well known that a commutative ring is semiperfect if and only
if it is isomorphic to a finite direct product of (commutative) local rings. In general, it was shown
by Mueller [31, Theorem 1] that a ring is semiperfect if and only if its identity 1 can be written as
a finite sum of orthogonal local idempotents. Mueller also showed that a ring is semiperfect if and
only if every primitive idempotent is local and there is no infinite set of orthogonal idempotents in
the ring. A ring is sometimes said to be orthogonally finite if it has no infinite set of orthogonal
idempotents. The following characterization of semiperfect rings is due to Camillo and Yu [7]. It
tells us that a ring is semiperfect if and only if it is clean (resp. exchange, potent, semipotent) and
orthogonally finite.
(2.10)
(1)
(2)
(3)
(4)
(5)

Theorem. The following are equivalent for any ring R.


R is semiperfect;
R is clean and orthogonally finite;
R is exchange and orthogonally finite;
R is potent and orthogonally finite;
R is semipotent and orthogonally finite.

Proof. It is well known that every semiperfect ring is orthogonally finite. Since every semiperfect
ring is clean, as noted above, this tells us that (1) \Rightarrow (2). Meanwhile, recall that every clean ring is
exchange by (1.4), that every exchange ring is potent according to Nicholson [34, Proposition 1.9],
and that every potent ring is semipotent by definition. It follows that (2) \Rightarrow (3) \Rightarrow (4) \Rightarrow (5). We
only need to show that every orthogonally finite semipotent ring is semiperfect. Recall that every
primitive idempotent in a semipotent ring is local according to (1.10). Since a ring is semiperfect
if and only if it is orthogonally finite and every primitive idempotent in the ring is local according
to Mueller [31, Theorem 1], it follows that (5) \Rightarrow (1). This completes the proof.
A ring R is called semilocal if the quotient ring R/J(R) is semisimple. It is known that every
semilocal ring is orthogonally finite, and it follows immediately from (2.10) that a semilocal ring is
clean if and only if it is semiperfect. In particular, this means that a semilocal ring is clean if and
only if idempotents lift modulo its Jacobson radical. It is well known that a commutative ring is
semilocal if and only if it has only a finite number of maximal ideals. More generally, it is known
that any ring R that has only finitely many maximal ideals is semilocal, and that the converse is

9
true when the quotient ring R/J(R) is commutative. It is not difficult to see that a semilocal ring
need not be clean. For example, the ring \BbbZ (6) of all rational numbers with denominators relatively
prime to 6 (when written in lowest terms) is semilocal since it is a commutative ring with only two
maximal ideals, but it is not clean according to (2.3) since it is a non-local ring with no nontrivial
idempotents.
(2.11) Proposition. A semilocal ring is clean if and only if it is semiperfect.
Proof. As we mentioned above, it is known that every semilocal ring is orthogonally finite. Since
every semiperfect ring is semilocal by definition, the result follows from (2.10).
Bass [4] introduced the notions of perfect and semiperfect rings as homological generalizations
of semiprimary rings. Semiperfect rings in particular will play a central role in our study of clean
group rings in the next chapter. The implications below should now be clear.
left/right artinian

=\Rightarrow

semiprimary

=\Rightarrow

left perfect

=\Rightarrow

semiperfect

=\Rightarrow

clean

These implications are known to be irreversible. For an example of a noncommutative ring that is
semiprimary but neither left nor right artinian, the interested reader should see Lam [26, Ex. 20.5].
Meanwhile, since every semiprimary ring is (left and right) perfect, any left perfect ring that is not
right perfect is not semiprimary---the interested reader should see Bass [4, p. 476] for an example
of such a ring. Finally, it is not difficult to see that the ring \BbbZ (2) of all rational numbers with odd
denominators (when written in lowest terms) is semiperfect but not perfect, and that the infinite
direct product \BbbZ (2) \times \BbbZ (2) \times \cdot \cdot \cdot is clean but not semiperfect.
In addition to the relationships above, we should note that every division ring is artinian, and
that every local ring is semiperfect. Of course these implications are also known to be irreversible.
For example, the ring \BbbZ /4\BbbZ of integers modulo 4 is artinian but not a division ring, and the direct
product \BbbZ /4\BbbZ \times \BbbZ /4\BbbZ is semiperfect but not local.

Regular \& \bfitpi -Regular Rings


A ring R is said to be regular---in the sense of von Neumann [42]---if for every element r \in R
there exists an element x \in R such that rxr = r. More generally, an element r in a ring R is called
regular if there exists an element x \in R such that rxr = r. It is not difficult to see that a ring R
is regular if and only if every principal left (equivalently, right) ideal of the ring is generated by an
idempotent---in other words, if for every element a \in R there exists some idempotent e of R such
that Re = Ra. Moreover, since the Jacobson radical of any ring contains no nonzero idempotents,
it follows that the Jacobson radical of a regular ring must be equal to 0. Warfield [46, p. 34] proved
that every regular ring is exchange, but a regular ring need not be clean according to Camillo and
Yu [7, p. 4746]. The following result due to Lee, Yi, and Zhou [27] gives a necessary and sufficient
condition for a regular ring to be clean.
(2.12) Theorem. A regular ring is clean if and only if its indecomposable images are clean.
Proof.

This proof is omitted---see Lee, Yi, and Zhou [27, Corollary 7].

A ring R is called unit regular if for every element r \in R there exists a unit u of R such that
rur = r. More generally, an element r in a ring R is called unit regular if there exists some unit
u \in R such that rur = r. It is not difficult to see that a ring R is unit regular if and only if every

10
element r \in R can be written as the product r = ue of a unit u and an idempotent e of R. Camillo
and Khurana [6, Theorem 1] proved that every unit regular ring is clean, but a single element in a
ring can be unit regular without being clean according to Khurana and Lam [23, Example 4.5].
(2.13) Theorem. Every unit regular ring is clean.
Proof.

This proof is omitted---see Camillo and Khurana [6, Theorem 1].

A ring R is called strongly regular if for every element r \in R there exists an element x \in R
such that r2 x = r. This definition is left-right symmetric. In fact, Azumaya [3, Lemma 1] showed
that a ring R is strongly regular if and only if for each element r \in R there exists a unique element
z \in R such that rzr = r and zrz = z with rz = zr. It can be shown that a ring is strongly regular
if and only if it is both regular and reduced, and that a ring is strongly regular if and only if it is
an abelian regular ring. It is also well known that every strongly regular ring is unit regular, and
therefore every strongly regular ring is clean by (2.13). We give a direct proof of this result.
(2.14) Proposition. Every strongly regular ring is unit regular and clean.
Proof. Let r be any element in a strongly regular ring R. Then there is some element z \in R such
that rzr = r with rz = zr. It is not difficult to check that e = rz is idempotent, while the element
u = r - (1 - e) is a unit of R with the inverse v = ze - (1 - e). Thus we can write r as the product
of a unit and an idempotent of R by writing r = ue, and as the sum of a unit and an idempotent
of R by writing r = u + (1 - e). Therefore every strongly regular ring is unit regular and clean.
It is well known that the notions of regular, unit regular, and strongly regular are equivalent in
the commutative case. Examples of commutative regular rings include all boolean rings and fields.
Meanwhile, Jacobson [21, Theorem 11] showed that if for every element r in a ring R there exists a
positive integer n such that rn = r, then every element in R has finite additive order and the ring
R is commutative. This notion is a proper generalization of that of a boolean ring. Moreover, it is
easy to see that every such ring is strongly regular, and therefore is a commutative regular ring.
A ring R is called \bfitpi -regular if for every element r \in R there exists an element x \in R such that
rn xrn = rn for some positive integer n. Notice that the equation rn xrn = rn implies that xrn is an
idempotent, and that xrn is equal to 0 if and only if rn = 0. In particular, this tells us that every
non-nil left ideal of a \pi -regular ring contains a nonzero idempotent---this result was pointed out by
Kaplansky [22, p. 63]. It follows that every \pi -regular ring is semipotent and its Jacobson radical is
a nil ideal. A semipotent ring with a nil Jacobson radical is sometimes called a Zorn ring. Since
the Jacobson radical of every \pi -regular ring is a nil ideal, it now follows from (1.7) that a \pi -regular
ring R is clean if and only if the quotient ring R/J(R) is clean.
A ring R is called strongly \bfitpi -regular if for every element r \in R there exists an element x \in R
such that rn+1 x = rn for some positive integer n. These rings were studied for nearly thirty years
before Dischinger [14] proved that their definition is left-right symmetric. Azumaya [3, Theorem 3]
showed that a ring R is strongly \pi -regular if and only if for every element r \in R there exists some
element z \in R such that rn zrn = rn with rz = zr for some positive integer n. The following result
was first proved by Burgess and Menal [5, Proposition 2.6].
(2.15) Theorem. Every strongly \pi -regular ring is clean.
Proof. This proof is omitted---see Nicholson [35, Theorem 1] for an elementary proof of this result
whose original proof involves sheaf-theoretic representations.

11
A commutative ring R is called zero-dimensional---in regard to Krull dimension---if its prime
ideals are all maximal. It is well known that a commutative ring is zero-dimensional if and only if
it is strongly \pi -regular, and therefore every zero-dimensional commutative ring is clean as a result
of (2.15). Meanwhile, it is also known that a commutative ring R is zero-dimensional if and only if
the quotient ring R/J(R) is regular and J(R) is a nil ideal. This tells us that a commutative ring
is regular if and only if it is zero-dimensional and reduced. More generally, it can be shown that a
ring is strongly regular if and only if it is strongly \pi -regular and reduced.
(2.16) Corollary. Every zero-dimensional commutative ring is clean.
Proof. As mentioned above, every zero-dimensional commutative ring is strongly \pi -regular, and
therefore is clean by (2.15).
Von Neumann [42] introduced regular rings as a proper generalization of semisimple rings. The
interested reader is encouraged to see Goodearl [17] and Tuganbaev [41] for more on regular rings
and rings close to regular. The implications below are well known.
commutative regular

strongly \pi -regular

=\Rightarrow

clean

=\Rightarrow

regular

=\Rightarrow

exchange

=\Rightarrow

=\Rightarrow

unit regular

=\Rightarrow

zero-dimensional comm.

=\Rightarrow

=\Rightarrow

strongly regular

=\Rightarrow

=\Rightarrow

These implications are known to be irreversible. In particular, since any noncommutative division
ring (e.g. Hamilton's quaternions) is strongly regular but not commutative, the pair of horizontal
implications on the left is irreversible. Meanwhile, Chen [11, Remark 4.3] constructed an example
of a ring that is unit regular but not strongly \pi -regular, which tells us that the pair of horizontal
implications in the middle is irreversible, and Camillo and Yu [7, p. 4746] proved that the ring in
a well-known example due to Bergman [19, Example 1] is regular but not clean, which shows that
the pair of horizontal implications on the right is irreversible. Finally, notice that the ring \BbbZ /4\BbbZ of
integers modulo 4 is a zero-dimensional commutative ring but is not regular, and therefore all four
vertical implications above are irreversible---more generally, the ring \BbbZ /n\BbbZ of integers modulo n is
zero-dimensional for any positive integer n, but is regular if and only if n is square-free according
to Ehrlich [16, Theorem 5].
In addition to the relationships above, we should mention that every left (right) perfect ring is
strongly \pi -regular. It is an easy exercise to check that a ring R is strongly \pi -regular if and only if
for any element a \in R the descending chain Ra \supseteq Ra2 \supseteq Ra3 \supseteq \cdot \cdot \cdot of principal left ideals in R has
a minimal element. Since a ring R is right perfect if and only if every descending chain of principal
left ideals in R has a minimal element according to Bass [4, Theorem P], it follows that every right
perfect ring is strongly \pi -regular. In fact, since the notion of a strongly \pi -regular ring is left-right
symmetric, this tells us that every left or right perfect ring is strongly \pi -regular. This implication
is known to be irreversible. For example, any infinite direct product of fields is strongly \pi -regular,
but not perfect.

Some Notes
(1) A ring is said to be left noetherian if every ascending chain of left ideals in the ring has
a maximal element---in other words, if for every ascending chain I1 \subseteq I2 \subseteq I3 \subseteq \cdot \cdot \cdot of left ideals in
the ring there is a positive integer n such that In = In+1 = In+2 = \cdot \cdot \cdot . To be right noetherian

12
requires instead that this property holds for every ascending chain of right ideals. We will say that
a ring is noetherian if it is both left and right noetherian---this notion is not left-right symmetric.
It is well known that every left (resp. right) artinian ring is left (resp. right) noetherian, and that
the converse is true for all semiprimary rings---this is the Hopkins-Levitzki Theorem.
(2) It is not difficult to see that a noetherian ring need not be clean. For example, the ring of
integers \BbbZ is noetherian but not clean, and the ring \BbbZ (6) of all rational numbers with denominators
relatively prime to 6 (when written in lowest terms) is both semilocal and noetherian but not clean.
It is well known that every left (right) noetherian ring is orthogonally finite. Since an orthogonally
finite ring is clean (resp. exchange, potent, semipotent) if and only if it is semiperfect by (2.10), it
follows that a left (right) noetherian ring is clean (resp. exchange, potent, semipotent) if and only
if it is semiperfect. Moreover, since a ring is semiperfect if and only if it is orthogonally finite and
every primitive idempotent in the ring is local according to Mueller [31, Theorem 1], it follows that
a left (right) noetherian ring is semiperfect (resp. clean, exchange, potent, semipotent) if and only
if every primitive idempotent in the ring is local.
(3) A ring R is called semiregular if the quotient ring R/J(R) is regular and idempotents lift
modulo J(R). This class of rings includes all regular rings, semiperfect rings, and zero-dimensional
commutative rings, and it is well known that every semiregular ring is exchange. The relationship
between semiperfect rings, semiregular rings, and exchange rings is shown below.
semiperfect

semiregular

=\Rightarrow

exchange

=\Rightarrow

clean

=\Rightarrow

=\Rightarrow

These implications are known to be irreversible. In particular, we gave an example above of a ring
that is clean but not semiperfect, while Monk [30, Theorem 2] showed that an exchange ring need
not be semiregular---more specifically, he showed that a commutative exchange (hence clean) ring
whose Jacobson radical is equal to 0 need not be regular. Meanwhile, we also mentioned above that
the ring in a well-known example due to Bergman [19, Example 1] is regular but not clean. Since
every regular ring is semiregular, it follows that the ring in Bergman's example is semiregular but
not clean, and therefore both of the vertical implications above are irreversible. No example of an
exchange ring that is neither regular nor clean is known to the author.
(4) A ring R is said to have stable range one if for any elements a, b \in R with Ra + Rb = R,
there is some element y \in R such that a + yb is a unit of R---this notion is left-right symmetric. It
is well known that every semiperfect ring, unit regular ring, and strongly \pi -regular ring has stable
range one. Moreover, Chen and Li [10, Corollary 3] proved that every quasi-duo exchange ring has
stable range one, and Yu [48, Theorem 1] proved that every exchange ring with artinian primitive
factors has stable range one. However, it follows immediately from [36, Corollary] that a clean ring
need not have stable range one.
(5) A semiregular ring is called unit semiregular if it has stable range one. It is well known
that a regular ring has stable range one if and only if it is unit regular. More generally, Aydo\u gdu,
\"
Lee, and Ozcan
[2, Theorem 2.1] showed that a semiregular ring R has stable range one if and only
if R/J(R) is unit regular and idempotents lift modulo J(R). Since every unit regular ring is clean
by (2.13), it follows from (1.6) that every semiregular ring with stable range one is clean---in other
words, every unit semiregular ring is clean. It is not known to the author whether every exchange
ring with stable range one is clean.

13
(6) A ring R is said to have idempotent stable range one if for any elements a, b \in R with
Ra + Rb = R, there exists some idempotent e of R such that a + eb is a unit of R---this notion is
left-right symmetric. Chen [9, Theorem 12] proved that every ring with idempotent stable range
one is clean---this follows easily from the fact that Ra + R( - 1) = R for any element a in a ring R.
He also proved that the converse is true for abelian rings, and hence an abelian ring is clean if and
only if it has idempotent stable range one. Meanwhile, Wang et al. [44] proved that all semiperfect
rings and unit regular rings have idempotent stable range one.

3 An Interesting Subclass of Clean Rings


In this section, we study the rings that consist entirely of units, idempotents, and quasiregular
elements. We know from (2.1) that the units, idempotents, and quasiregular elements of any ring
are clean, and thus every ring that consists entirely of these types of elements is a clean ring. But
exactly which rings belong to this interesting subclass of clean rings? We give a complete solution
to this problem below. In particular, we prove that any abelian ring that consists entirely of units,
idempotents, and quasiregular elements is either a boolean ring, a local ring, or isomorphic to the
direct product of two division rings, while any nonabelian ring that consists entirely of these types
of elements is isomorphic to either the full matrix ring M2 (D) for some division ring D, or to the
ring of a Morita context with zero pairings where both of the underlying rings are division rings.

Abelian Rings
We begin our study of rings consisting entirely of units, idempotents, and quasiregular elements
by classifying those rings that consist entirely of any two of these types of elements. In particular,
we prove that any such ring must be either a division ring, a boolean ring, or a local ring. Let us
define the following subsets of a ring R.
\scrU (R) = \{ r \in R | r is a unit of R\}
\scrI \scrD (R) = \{ r \in R | r is an idempotent of R\}
\scrQ \scrR (R) = \{ r \in R | r is a quasiregular element of R\}
The first part of the following theorem was proved independently by Chen and Cui [12] and by the
author (unpublished). This was proved earlier for commutative rings by Anderson and Camillo [1].
We give a new proof of this result.
(3.1) Theorem. Let R be any ring.
(1) R = \scrU (R) \cup \scrI \scrD (R) if and only if R is a division ring or a boolean ring.
(2) R = \scrU (R) \cup \scrQ \scrR (R) if and only if R is a local ring.
(3) R = \scrI \scrD (R) \cup \scrQ \scrR (R) if and only if R is a division ring or a boolean ring.
Proof. (1) Suppose that the ring R consists entirely of units and idempotents. It is not difficult
to see that any such ring is reduced, and therefore is abelian. In particular, the ring R either has
no nontrivial idempotents, and hence is a division ring, or it has a nontrivial central idempotent,
and hence is decomposable. Suppose that R is decomposable, and let A \times B be any direct product

14
decomposition of R. If either of the rings A or B contained a non-idempotent x, then there would
be an element (x, 0) or (0, x) in A \times B that was neither a unit nor an idempotent of A \times B. Since
the ring R consists entirely of units and idempotents, this tells us that A and B are both boolean
rings. Since the direct product of any two boolean rings is itself a boolean ring, it follows that the
ring R is boolean in this case. Therefore every ring that consists entirely of units and idempotents
is either a division ring or a boolean ring. The converse is easy to check.
(2) This result is well known, and its proof is omitted.
(3) Suppose that the ring R consists entirely of idempotents and quasiregular elements. Then
for any element r \in R, the element 1 - r is either idempotent or quasiregular. It is not difficult to
check that r is idempotent if (and only if) 1 - r is idempotent, and that r is a unit if (and only if)
1 - r is quasiregular. This tells us that the ring R consists entirely of units and idempotents, and
hence it must be either a division ring or a boolean ring by (1). Therefore every ring that consists
entirely of idempotents and quasiregular elements is either a division ring or a boolean ring. The
converse is easy to check.
It is not difficult to see that division rings, boolean rings, and local rings are not the only rings
that consist entirely of units, idempotents, and quasiregular elements. For example, the interested
reader is encouraged to check that the direct product \BbbZ /2\BbbZ \times \BbbZ /3\BbbZ consists entirely of these types
of elements even though it is not a division ring, a boolean ring, or a local ring. More generally, we
show that the direct product of any two division rings consists entirely of units, idempotents, and
quasiregular elements.
(3.2) Proposition. Let R be the direct product of two division rings. Then R consists entirely
of units, idempotents, and quasiregular elements.
Proof. Since multiplication in a direct product of rings is defined componentwise, an element in
any such ring is a unit (resp. idempotent, quasiregular element) of that ring if and only if the entry
in each of its components is a unit (resp. idempotent, quasiregular element) of its ring. Since the
ring R is a direct product of two division rings, this tells us that an element in R is a unit (resp.
quasiregular element) of the ring R if and only if the entry in each of its components is a nonzero
(resp. nonidentity) element of its ring. It is not difficult now to see that each element r \in R must
satisfy at least one of the following conditions, where each of the entries denoted by \ast can be any
nonzero, nonidentity element of its ring.
r = (1, 1), (1, \ast ), (\ast , 1), or (\ast , \ast ) \in \scrU (R)
r = (0, 0), (0, 1), (1, 0), or (1, 1) \in \scrI \scrD (R)
r = (0, 0), (0, \ast ), (\ast , 0), or (\ast , \ast ) \in \scrQ \scrR (R)
Therefore the direct product of any two division rings consists entirely of units, idempotents, and
quasiregular elements.
The next result shows that any decomposable ring that consists entirely of units, idempotents,
and quasiregular elements must be either a boolean ring or isomorphic to a direct product of two
division rings. Incidentally, this result tells us that every decomposable ring that consists entirely
of these types of elements is necessarily abelian.
(3.3) Theorem. Let R be a decomposable ring that consists entirely of units, idempotents, and
quasiregular elements. Then R is a boolean ring or a direct product of two division rings.

15
Proof. Let A \times B be any direct product decomposition of the ring R. Since multiplication in a
direct product of rings is defined componentwise, an element in A \times B is a unit (resp. idempotent,
quasiregular element) of the ring A \times B if and only if the entry in each of its components is a unit
(resp. idempotent, quasiregular element) of its ring. If either of the rings A or B contained some
element x that was neither a unit nor an idempotent of its ring, then there would be an element
(x, 1) or (1, x) in A \times B that was not a unit, an idempotent, or a quasiregular element of A \times B.
Since the ring R consists entirely of units, idempotents, and quasiregular elements, it follows that
the rings A and B must both consist entirely of units and idempotents. We know from (3.1) that
any such ring is either a division ring or boolean ring. If either of the rings A or B contained some
unit u other than 1, while the other ring contained some nontrivial idempotent e, then there would
be an element (u, e) or (e, u) in A \times B that is not a unit, an idempotent, or a quasiregular element
of A \times B. Since the ring R consists entirely of units, idempotents, and quasiregular elements, this
tells us that A and B are either both boolean rings or both division rings. Since the direct product
of any two boolean rings is itself a boolean ring, it follows that the ring R is either a boolean ring
or a direct product of two division rings. Therefore every decomposable ring that consists entirely
of units, idempotents, and quasiregular elements is either a boolean ring or a direct product of two
division rings.
We are now ready to classify the abelian rings that consist entirely of units, idempotents, and
quasiregular elements. In particular, we prove that any abelian ring that consists entirely of these
types of elements must be either a boolean ring, a local ring, or isomorphic to a direct product of
two division rings.
(3.4) Theorem. Let R be an abelian ring that consists entirely of units, idempotents, and quasiregular elements. Then R is a boolean ring, a local ring, or a direct product of two division rings.
Proof. Since the ring R is abelian, it either has no nontrivial idempotents or a nontrivial central
idempotent. If R has no nontrivial idempotents, then it consists entirely of units and quasiregular
elements, and hence is a local ring by (3.1). Meanwhile, if R has a nontrivial central idempotent,
then it is decomposable, and hence is either a boolean ring or a direct product of two division rings
by (3.3). Therefore every abelian ring that consists entirely of units, idempotents, and quasiregular
elements is either a boolean ring, a local ring, or a direct product of two division rings.
(3.5) Corollary. Let R be a commutative ring that consists entirely of units, idempotents, and
quasiregular elements. Then R is a boolean ring, a commutative local ring, or a direct product of
two fields.
Proof.

This follows immediately from (3.4).

Nonabelian Rings
It still remains to classify the nonabelian rings that consist entirely of units, idempotents, and
quasiregular elements. We give a complete solution to this problem below. In particular, we prove
that any nonabelian ring that consists entirely of units, idempotents, and quasiregular elements is
isomorphic to either the full matrix ring M2 (D) for some division ring D, or to the ring of a Morita
context with zero pairings where both of the underlying rings are division rings. We begin with the
following lemma.

16
(3.6) Lemma. Let R be any ring that consists entirely of units, idempotents, and quasiregular
elements. Then eRe is a division ring for every noncentral idempotent e of R.
Proof. Let e be a nontrivial idempotent of R with complement f = 1 - e, and let us identify the
ring R with its Peirce decomposition with respect to e.
\biggl(
\biggr)
eRe eRf
\sim
R =
f Re f Rf
It is not difficult to check that the inverse of a diagonal matrix (if one exists) is a diagonal matrix.
Since the ring R consists entirely of units, idempotents, and quasiregular elements, it follows that
the subring of all diagonal matrices in this decomposition of R also consists entirely of these types
of elements. Moreover, this subring is isomorphic to the direct product eRe \times f Rf , and since e is
a nontrivial idempotent of R, we know that eRe and f Rf are both (nonzero) rings. It now follows
from the proof of (3.3) that the corner ring eRe is either a division ring or a boolean ring. We need
to show that eRe must be a division ring when e is noncentral. We prove the contrapositive.
Suppose that the corner ring eRe is not a division ring. Since it must be either a division ring
or a boolean ring, this tells us that eRe must contain a nontrivial idempotent---in other words, it
must contain some idempotent other than 0 or e. Let a be any nontrivial idempotent of eRe with
complement b = e - a, and consider the following matrices, where each entry denoted by x can be
any element in eRf , while each entry denoted by y can be any element in f Re.
\biggr)
\biggr)
\biggl(
\biggr)
\biggl(
\biggr)
\biggl(
\biggl(
b 0
a 0
b x
a x
, X4 =
, X3 =
, X2 =
X1 =
y 0
y 0
0 0
0 0
Since X1 and X2 both have a row in which each entry is 0, while X3 and X4 both have a column in
which each entry is 0, none of these matrices is a unit of R. Now consider the following matrix.
\biggr)
\biggr)
\biggl(
\biggr) \biggl(
\biggl(
b - x
a x
e 0
=
-
I - X1 =
0 f
0 0
0 f
In particular, notice that b would need to have a left inverse in eRe in order for this matrix I - X1
to have a left inverse in R. However, this element b cannot have a left inverse in eRe since it is a
nontrivial idempotent of eRe. This tells us that I - X1 is not invertible in R, and hence X1 is not
quasiregular in R. Similar arguments show that the matrices X2 , X3 , and X4 are not quasiregular
in R. Since the ring R consists entirely of units, idempotents, and quasiregular elements, it follows
that each of these matrices must be idempotent. A simple computation shows that the matrix X1
is idempotent if and only if ax = x, while X2 is idempotent if and only if bx = x. Moreover, since
x is an element of eRf , we also know that ex = x. These last three equalities tell us that x = 0, as
follows, while a similar argument shows that y = 0.
x = bx = (e - a)x = ex - ax = x - x = 0
Since this is the case for any elements x \in eRf and y \in f Re, it follows that eRf = 0 = f Re, and
hence e is a central idempotent of R. More specifically, this tells us that e is a central idempotent
of R when the ring eRe is a not division ring. Therefore eRe is a division ring for every noncentral
idempotent e of R by contrapositive.
The next result classifies the full matrix rings Mn (R) that consist entirely of units, idempotents,
and quasiregular elements for some integer n > 1. In particular, it shows that the ring of all n \times n
matrices with entries from some arbitrary ring consists entirely of these types of elements for some

17
integer n > 1 if and only if it is equal to the full matrix ring M2 (D) for some division ring D. The
sufficiency is essentially due to Li [28]. Recall that two matrices X and Y are said to be similar if
there is an invertible matrix P such that P - 1 XP = Y .
(3.7) Proposition. Let R be the ring of all n \times n matrices with entries from some arbitrary ring.
If n > 1, then R consists entirely of units, idempotents, and quasiregular elements if and only if it
is equal to M2 (D) for some division ring D.
Proof. Suppose that the ring R consists entirely of units, idempotents, and quasiregular elements
for some integer n > 1, and let us write R = Mn (S). Since n > 1, the matrix unit E = E11 and its
complement F = I - E11 are both noncentral idempotents of R. Since the ring R consists entirely
of units, idempotents, and quasiregular elements, it follows from (3.6) that the ring F RF must be
a division ring. Moreover, notice that we have the following.
F RF \sim
= S \times \cdot \cdot \cdot \times S
\underbrace{}
\underbrace{}
n - 1 times

Since the direct product on the right-hand side of the isomorphism above must be a division ring,
it follows that n = 2 and that S must be a division ring. Therefore R is equal to M2 (D) for some
division ring D. This proves the necessity.
Conversely, suppose that the ring R is equal to M2 (D) for some division ring D. Then for any
matrix X \in R, it follows immediately from a technical lemma due to Li [28, Lemma 2.4] that if X
is neither a unit nor a quasiregular element of R, then it must be similar to the following matrix.
\biggr)
\biggl(
1 1
Y =
0 0
Notice that this matrix Y is idempotent. It is not difficult to check that any matrix similar to an
idempotent is idempotent. In particular, this tells us that if X is neither a unit nor a quasiregular
element of the ring R, then it must be an idempotent of R. Therefore R consists entirely of units,
idempotents, and quasiregular elements. This proves the sufficiency.
A ring is called semiprime if it has no nonzero nilpotent ideals. Du [15] proved that if eRe is
a division ring for every noncentral idempotent e of a nonabelian semiprime ring R, then that ring
R must be isomorphic to the full matrix ring M2 (D) for some division ring D. We prove that every
nonabelian semiprime ring that consists entirely of units, idempotents, and quasiregular elements
is isomorphic to M2 (D) for some division ring D.
(3.8) Theorem. Let R be any nonabelian ring that consists entirely of units, idempotents, and
quasiregular elements. If R is semiprime, then it is isomorphic to M2 (D) for some division ring D.
Proof. Suppose that R is semiprime. Since the ring R consists entirely of units, idempotents, and
quasiregular elements, it follows immediately from (3.6) that the ring eRe must be a division ring
for every noncentral idempotent e of R. As we noted above, Du [15, Lemma 21] proved that every
nonabelian semiprime ring that has this property is isomorphic to the full matrix ring M2 (D) for
some division ring D. Therefore R is isomorphic to M2 (D) for some division ring D.
Let A, B be some pair of rings, let M be any (A, B)-bimodule, let N be any (B, A)-bimodule,
and let \psi : M \times N \rightarrow A and \varphi : N \times M \rightarrow B be some binary functions, written multiplicatively as
\psi (x, y) = xy and \varphi (y, x) = yx. The 6-tuple (A, B, M, N, \psi , \varphi ) is called a Morita context if these

18
binary functions \psi and \varphi preserve addition in each variable and satisfy the following associativity
conditions for every element a \in A, b \in B, x \in M , and y \in N .
a(xy) = (ax)y, (xb)y = x(by), (xy)a = x(ya),
b(yx) = (by)x, (ya)x = y(ax), (yx)b = y(xb).
In this case, the binary functions \psi and \varphi are called pairings, and if \psi and \varphi are both equal to 0,
then the Morita context (A, B, M, N, \psi , \varphi ) is said to be a Morita context with zero pairings.
Let (A, B, M, N, \psi , \varphi ) be any Morita context, and notice that the following set of matrices is a
ring with respect to the usual operations for addition and multiplication of matrices when \psi and \varphi
are used to define multiplication between elements of the bimodules M and N .
\Biggl\{ \biggl(
\Biggr\}
\biggl(
\biggr)
\biggr) \bigm|
A M
a x \bigm| \bigm|
=
a \in A, b \in B, x \in M, y \in N
N B
y b \bigm|
Any such ring is known as the the ring of a Morita context. The next result shows that if either
of the bimodules M or N is nonzero, then the ring of any Morita context (A, B, M, N, \psi , \varphi ) with
zero
pairings consists entirely of units, idempotents, and quasiregular elements if and only if the rings
A and B are both division rings.
(3.9) Proposition. Let R be the ring of a Morita context (A, B, M, N, \psi , \varphi ) with zero pairings.
If either of the bimodules M or N is nonzero, then R consists entirely of units, idempotents, and
quasiregular elements if and only if A and B are both division rings.
Proof. Suppose that the ring R consists entirely of units, idempotents, and quasiregular elements
for some bimodules M and N , at least one of which is nonzero. Since at least one of the bimodules
M or N is nonzero, the matrix units E = E11 and F = E22 are both noncentral idempotents of R.
Since the ring R consists entirely of units, idempotents, and quasiregular elements, it now follows
from (3.6) that ERE and F RF must both be division rings. Moreover, notice that ERE \sim
= A and
that F RF \sim
= B. Therefore the rings A and B must both be division rings when either of M or N
is nonzero. This proves the necessity.
Conversely, suppose that A and B are both division rings. Then any nonzero element in either
of these rings is a unit of its ring. Let a be any nonzero element in A with inverse u, and let b be
any nonzero element in B with inverse v. Then for any elements x \in M and y \in N , since R is the
ring of a Morita context with zero pairings, we have the following.
\biggl(

a x
y b

\biggr) - 1

\biggl(
=

u
- uxv
- vyu
v

\biggr)

This shows that a matrix in R is a unit (resp. quasiregular element) of the ring R when each entry
on its main diagonal is a nonzero (resp. nonidentity) element of its ring. It is not difficult now to
see that each matrix X \in R must satisfy at least one of the following conditions, where each of the
entries denoted by \ast can be any nonzero, nonidentity element of its ring, while each entry denoted

19
by x can be any element in M , and
\biggl(
\biggr)
1 x
X =
,
y 1
\biggl(
\biggr)
0 0
X =
,
0 0
\biggl(
\biggr)
0 x
X =
,
y 0

each entry denoted by y can be any element in N .


\biggl(
\biggr) \biggl(
\biggr)
\biggl(
\biggr)
1 x
\ast x
\ast x
,
, or
\in \scrU (R)
y \ast
y 1
y \ast
\biggl(
\biggr) \biggl(
\biggr)
\biggl(
\biggr)
0 x
1 x
1 0
,
, or
\in \scrI \scrD (R)
y 1
y 0
0 1
\biggl(
\biggr) \biggl(
\biggr)
\biggl(
\biggr)
0 x
\ast x
\ast x
,
, or
\in \scrQ \scrR (R)
y \ast
y 0
y \ast

Therefore R consists entirely of units, idempotents, and quasiregular elements when A and B are
both division rings. This proves the sufficiency.
Nicholson [32] called a ring an NJ-ring if the ring consists entirely of regular and quasiregular
elements. Examples of NJ-rings include all regular rings and local rings. Moreover, since the units
and idempotents of any ring are regular, every ring that consists entirely of units, idempotents, and
quasiregular elements is an NJ-ring. The converse is not true in general, since a regular ring need
not be clean. Nicholson [32] showed that an NJ-ring must be either a regular ring, a local ring, or
isomorphic to the ring of a Morita context with zero pairings where the underlying rings are both
division rings. We prove that every nonabelian, nonsemiprime ring that consists entirely of units,
idempotents, and quasiregular elements is isomorphic to the ring of some Morita context with zero
pairings where the underlying rings are both division rings.
(3.10) Theorem. Let R be any nonabelian ring that consists entirely of units, idempotents, and
quasiregular elements. If R is not semiprime, then it is isomorphic to the ring of a Morita context
with zero pairings where the underlying rings are both division rings.
Proof. Suppose that R is not semiprime. Then R is a nonabelian, nonsemiprime ring that consists
entirely of units, idempotents, and quasiregular elements. Since every ring that consists entirely of
these types of elements is an NJ-ring, it follows that R is a nonabelian, nonsemiprime NJ-ring. As
we noted above, Nicholson [32, Theorem 2] showed that an NJ-ring must be either a regular ring,
a local ring, or isomorphic to the ring of a Morita context with zero pairings where the underlying
rings are both division rings. In particular, since every regular ring is semiprime, while every local
ring is abelian, it follows that every nonabelian, nonsemiprime NJ-ring must be isomorphic to the
ring of some Morita context with zero pairings where the underlying rings are both division rings.
Therefore R is isomorphic to the ring of a Morita context with zero pairings where the underlying
rings are both division rings.

A Special Case
It is well known that every nilpotent element is quasiregular. In particular, if xn = 0 for some
integer n > 1, then an easy computation shows that 1 + x + x2 + \cdot \cdot \cdot + xn - 1 is an inverse of 1 - x.
We classified the rings that consist entirely of units idempotents, and quasiregular elements above.
But when do these rings consist entirely of units, idempotents, and nilpotent elements? We give a
complete solution to this problem below. Specifically, we prove that any abelian ring that consists
entirely of units, idempotents, and nilpotent elements is either a boolean ring or a local ring with
a nil Jacobson radical, while any nonabelian ring that consists entirely of these types of elements

20
is isomorphic to either the full matrix ring M2 (\BbbZ /2\BbbZ ), or to the ring of a Morita context with zero
pairings where the underlying rings are both \BbbZ /2\BbbZ .
We begin our study of rings consisting entirely of units, idempotents, and nilpotent elements by
classifying the abelian rings that consist entirely of these types of elements. It is easy to check that
all boolean rings and local rings with a nil Jacobson radical consist entirely of units, idempotents,
and nilpotent elements. The next result shows that any abelian ring that consists entirely of these
types of elements must be either a boolean ring or a local ring with a nil Jacobson radical.
(3.11) Proposition. Let R be any abelian ring that consists entirely of units, idempotents, and
nilpotent elements. Then R is a boolean ring or a local ring with a nil Jacobson radical.
Proof. Since the ring R is abelian, it either has no nontrivial idempotents or a nontrivial central
idempotent. Suppose that R has no nontrivial idempotents. Then it consists entirely of units and
nilpotent (hence quasiregular) elements, and thus it is a local ring by (3.1). Moreover, since every
nonunit of R is nilpotent, it is a local ring with a nil Jacobson radical. Suppose instead that R has
a nontrivial central idempotent. Then R is a decomposable ring. Let A \times B be any direct product
decomposition of R. Since the ring R consists entirely of units, idempotents, and nilpotent (hence
quasiregular) elements, it follows from (3.3) that the rings A and B are either both boolean rings
or both division rings. If either of A or B contained a unit u other than 1, then there would be an
element (u, 0) or (0, u) in A \times B that is not a unit, an idempotent, or a nilpotent element of A \times B.
Since the ring R consists entirely of units, idempotents, and nilpotent elements, this tells us that
A and B are both boolean rings. Since the direct product of two boolean rings is itself a boolean
ring, it follows that the ring R is boolean in this case. Therefore every abelian ring that consists
entirely of units, idempotents, and nilpotent elements is either a boolean ring or a local ring with
a nil Jacobson radical.
It still remains to classify the nonabelian rings that consist entirely of units, idempotents, and
nilpotent elements. We begin with the following corollary to (3.6).
(3.12) Corollary. Let R be any ring that consists entirely of units, idempotents, and nilpotent
elements. Then eRe is isomorphic to \BbbZ /2\BbbZ for every noncentral idempotent e of R.
Proof. Since the ring R consists entirely of units, idempotents, and nilpotent (hence quasiregular)
elements, we know from (3.6) that eRe must be a division ring for every noncentral idempotent e
of R. However, if the ring eRe contained some unit u \not = e, then the following matrix would not be
a unit, an idempotent, or a nilpotent element of the Peirce decomposition of R with respect to e.
\biggl(
\biggr)
u 0
X =
0 0
Therefore eRe must be isomorphic to \BbbZ /2\BbbZ for every noncentral idempotent e of R.
It is easy to see that the full matrix ring M2 (\BbbZ /2\BbbZ ) consists entirely of units, idempotents, and
nilpotent elements. The following result shows that every nonabelian semiprime ring that consists
entirely of these types of elements must be isomorphic to M2 (\BbbZ /2\BbbZ ).
(3.13) Proposition. Let R be any nonabelian ring that consists entirely of units, idempotents,
and nilpotent elements. If R is semiprime, then it is isomorphic to M2 (\BbbZ /2\BbbZ ).

21
Proof. Suppose that the ring R is semiprime. Since it consists entirely of units, idempotents, and
nilpotent (hence quasiregular) elements, we know from (3.8) that R must be isomorphic to M2 (D)
for some division ring D. Let us identify R with the full matrix ring M2 (D), and notice that the
matrix unit E = E11 is a noncentral idempotent of R with ERE \sim
= D. Moreover, since the ring R
consists entirely of units, idempotents, and nilpotent elements, it follows from (3.12) that the ring
ERE is isomorphic to \BbbZ /2\BbbZ . Therefore R is isomorphic to M2 (\BbbZ /2\BbbZ ).
It is easy to check that the ring of any Morita context with zero pairings where the underlying
rings are both isomorphic to \BbbZ /2\BbbZ consists entirely of units, idempotents, and nilpotent elements.
The following result shows that every nonabelian, nonsemiprime ring that consists entirely of these
types of elements must be isomorphic to the ring of a Morita context with zero pairings where the
underlying rings are both \BbbZ /2\BbbZ . In this case, it can be shown that the underlying bimodules must
both be elementary (abelian) 2-groups.
(3.14) Proposition. Let R be any nonabelian ring that consists entirely of units, idempotents,
and nilpotent elements. If R is not semiprime, then it is isomorphic to the ring of a Morita context
with zero pairings where the underlying rings are both \BbbZ /2\BbbZ .
Proof. Suppose that the ring R is not semiprime. Since it consists entirely of units, idempotents,
and nilpotent (hence quasiregular) elements, we know from (3.10) that R must be isomorphic to the
ring of a Morita context with zero pairings where the underlying rings and are both division rings.
Let us identify R with the ring of some Morita context (A, B, M, N, \psi , \varphi ) with zero pairings where
the rings A and B are both division rings, and notice that the matrix units E = E11 and F = E22
are noncentral idempotents of R with ERE \sim
= A and F RF \sim
= B. Since the ring R consists entirely
of units, idempotents, and nilpotent elements, it now follows from (3.12) that the rings ERE and
F RF must both be isomorphic to \BbbZ /2\BbbZ . Therefore R is isomorphic to the ring of a Morita context
with zero pairings where the underlying rings are both \BbbZ /2\BbbZ .

Some Notes
(1) A ring is said to be strongly clean if each element in the ring can be written as the sum
of a unit and an idempotent that commute. It follows easily from the proof of (2.1) that every ring
that consists entirely of units, idempotents, and quasiregular elements is strongly clean. It can be
shown that every ring that consists entirely of these types of elements is also optimally clean in
the sense of Shifflet [39]. We should note that optimally clean implies strongly clean by definition.
(2) As we mentioned in \S 1, it is known that the corner rings of a clean ring need not be clean.
However, the corner rings of any strongly clean ring must be strongly clean according to S\'anchez
Campos (unpublished), while the corner rings of any optimally clean ring must be optimally clean
according to Shifflet [39, Proposition 3.1.5]. Similarly, it can be shown that the corner rings of any
ring that consists entirely of units, idempotents, and quasiregular elements also consists entirely of
units, idempotents, and quasiregular elements.
(3) As we mentioned in \S 1, it is not known whether being clean is Morita invariant. However,
we know from (3.7) that consisting entirely of units, idempotents, and quasiregular elements is not
Morita invariant. Moreover, being strongly clean (or optimally clean) is also not Morita invariant
according to Wang and Chen [43, Example 1].

22
(4) A ring is called Dedekind finite (or sometimes, von Neumann finite) if ab = 1 implies
that ba = 1 for any pair of elements a, b in the ring. It can be shown that every ring that consists
entirely of units, idempotents, and quasiregular elements is Dedekind finite. It not known whether
a strongly clean ring or an optimally clean ring must be Dedekind finite, but a clean ring need not
be Dedekind finite according to Nicholson and Varadarajan [36, Corollary].
(5) Recall that a ring R is called quasinormal if eRf Re = 0 for every noncentral idempotent
e of R with complement f = 1 - e. It can be shown that every nonabelian, nonsemiprime ring that
consists entirely of units, idempotents, and quasiregular elements is quasinormal. More generally,
Nicholson [32] showed that every nonsemiprime NJ-ring is quasinormal. Unfortunately, this result
is hidden in the proof of [32, Theorem 2].
(6) We know from (3.6) that eRe must be a division ring for every noncentral idempotent e of
any ring R that consists entirely of units, idempotents, and quasiregular elements. It can be shown
that the converse is true for any nonabelian quasinormal ring R, but it is not known whether the
converse is true for every nonabelian ring R.
(7) Let R be any ring, and let M be any (R, R)-bimodule. Then the following set of matrices
is a ring with respect to the usual operations for addition and multiplication of matrices.
\Biggl\{ \biggl(
\Biggr\}
\biggr) \bigm|
r x \bigm| \bigm|
T (R, M ) =
r \in R, x \in M
0 r \bigm|
Any such ring is known as a trivial extension. It can be shown that a trivial extension T (R, M )
consists entirely of units, idempotents, and quasiregular elements if and only if the ring R consists
entirely of these types of elements and the equation ex = xf holds for every nontrivial idempotent
e of R and every element x \in M , where f = 1 - e. It follows easily from this result that a trivial
extension of the form T (R, R) consists entirely of units, idempotents, and quasiregular elements if
and only if R is local. In this case, the trivial extension T (R, R) is also local.
(8) Let A, B be a pair of rings, let X be any (A, B)-bimodule, let Y be any (B, A)-bimodule,
and consider the following sets of matrices.
\Biggl\{ \biggl(
\Biggr\}
\Biggl\{ \biggl(
\Biggr\}
\biggr) \bigm|
\biggr) \bigm|
a 0 \bigm| \bigm|
0 x \bigm| \bigm|
R =
a \in A, b \in B , M =
x \in X, y \in Y
0 b \bigm|
y 0 \bigm|
Notice that the set R is a ring with respect to the usual operations for addition and multiplication
of matrices, while the set M is an (R, R)-bimodule. Now consider the trivial extension T (R, M ).
\right\}
\left\{ \left(
\right) \bigm|
a 0 0 x \bigm| \bigm|
0 b y 0 \bigm| \bigm|
a \in A, b \in B, x \in X, y \in Y
T (R, M ) =
0 0 a 0 \bigm| \bigm|
0 0 0 b \bigm|
It is not difficult to see that the trivial extension T (R, M ) is isomorphic to the ring of the Morita
context (A, B, X, Y, \psi , \varphi ) where \psi and \varphi are zero pairings. It follows immediately from (3.9) that
T (R, M ) consists entirely of units, idempotents, and quasiregular elements for some M \not = 0 if and
only if A and B are both division rings. Meanwhile, it follows from (3.10) that every nonabelian,
nonsemiprime ring that consists entirely of units, idempotents, and quasiregular elements must be
isomorphic to T (R, M ) for some division rings A and B.

References

[1] D.D. Anderson and Victor P. Camillo.


Commutative rings whose elements are a sum of a unit and idempotent.
Communications in Algebra 30(7) (2002), 3327--3336.
MR 1914999 (2003e:13006)
\"
[2] Pinar Aydo\u
gdu, Yang Lee and A. Ci\u
\c gdem Ozcan.
Rings close to semiregular.
Journal of the Korean Mathematical Society 49(3) (2012), 605--622.
MR 2953039
[3] Gor\^
o Azumaya.
Strongly \pi -regular rings.
Journal of the Faculty of Science, Hokkaido University. Series 1, Mathematics 13 (1954), 34--39.
MR 67864 (16,788e)
[4] Hyman Bass.
Finitistic dimension and a homological generalization of semi-primary rings.
Transactions of the American Mathematical Society 95(3) (1960), 466--488.
MR 157984 (28 \#1212)
[5] W.D. Burgess and Pere Menal.
On strongly \pi -regular rings and homomorphisms into them.
Communications in Algebra 16(8) (1988), 1701--1725.
MR 949266 (89f:16015)
[6] Victor P. Camillo and Dinesh Khurana.
A characterization of unit regular rings.
Communications in Algebra 29(5) (2001), 2293--2295.
MR 1837978
[7] Victor P. Camillo and Hua-Ping Yu.
Exchange rings, units and idempotents.
Communications in Algebra 22(12) (1994), 4737--4749.
MR 1285703 (95d:16013)
[8] Huanyin Chen.
Exchange rings with artinian primitive factors.
Algebras and Representation Theory 2(2) (1999), 201--207.
MR 1702275 (2000d:16005)
[9] Huanyin Chen.
Rings with many idempotents.
International Journal of Mathematics and Mathematical Sciences 22(3) (1999), 547--558.
MR 1717176 (2000k:16040)
[10] Huanyin Chen and Fu'an Li.
Exchange rings having ideal-stable range one.
Science in China Series A: Mathematics 44(5) (2001), 579--586.
MR 1843753 (2002e:19001)

23

24
[11] Weixing Chen.
On semiabelian \pi -regular rings.
International Journal of Mathematics and Mathematical Sciences 2007, Article ID 63171, 10 pages.
MR 2320775 (2008d:16019)
[12] Weixing Chen and Shuying Cui.
On clean rings and clean elements.
Southeast Asian Bulletin of Mathematics 32(5) (2008), 855--861.
MR 2470416 (2009k:16058)
[13] Peter Crawley and Bjarni J\'
onsson
Refinements for infinite direct decompositions of algebraic systems.
Pacific Journal of Mathematics 14(3) (1964), 797--855.
MR 169806 (30 \#49)
[14] Friedrich Dischinger.
Sur les anneaux fortement \pi -r\'eguliers.
Les Comptes Rendus de l'Acad\'emie des Sciences, Paris. S\'eries A-B 283(8) (1976), A571--A573.
MR 422330 (54 \#10321)
[15] Xiankun Du.
The adjoint semigroup of a ring.
Communications in Algebra 30(9) (2002), 4507--4525.
MR 1936488 (2003i:20118)
[16] Gertrude Ehrlich.
Unit-regular rings.
Portugaliae Mathematica 27(4) (1968), 209--212.
MR 266962 (42 \#1864)
[17] K.R. Goodearl.
Von Neumann regular rings. Second edition.
Krieger Publishing Company, Malabar, FL, 1991.
MR 1150975 (93m:16006)
[18] Juncheol Han and W.K. Nicholson.
Extensions of clean rings.
Communications in Algebra 29(6) (2001), 2589--2595.
MR 1845131 (2002d:16043)
[19] David Handleman.
Perspectivity and cancellation in regular rings.
Journal of Algebra 48(1) (1977), 1--16.
MR 447329 (56 \#5642)
[20] V.A. Hiremath, Sharad Hegde.
On a subclass of semipotent rings.
Research Journal of Pure Algebra 2(2) (2012), 66--70.
Available online through http://www.rjpa.info.
[21] Nathan Jacobson.
Structure theory for algebraic algebras of bounded degree.
Annals of Mathematics, Second Series, 46(4) (1945), 695--707.
MR 14083 (7,238c)
[22] Irving Kaplansky.
Topological representation of algebras II.
Transactions of the American Mathematical Society 68(1) (1950), 62--75.
MR 32612 (11,317c)

25
[23] Dinesh Khurana and T.Y. Lam.
Clean matrices and unit-regular matrices.
Journal of Algebra 280(2) (2004), 683--698.
MR 2090058 (2005f:16014)
[24] Kwangil Koh.
On lifting idempotents.
Canadian Mathematical Bulletin 17(4) (1974), 607.
MR 407073 (53 \#10856)
[25] T.Y. Lam.
A first course in noncommutative rings. Second edition. Graduate Texts in Mathematics, 131.
Springer-Verlag, New York, 2001.
MR 1838439 (2002c:16001)
[26] T.Y. Lam.
Exercises in classical ring theory. Second edition. Problem Books in Mathematics.
Springer-Verlag, New York, 2003.
MR 2003255 (2004g:16001)
[27] Tsiu-Kwen Lee, Zhong Yi and Yiqiang Zhou.
An example of Bergman's and the extension problem for clean rings.
Communications in Algebra 36(4) (2008), 1413--1418.
MR 2406593 (2009b:16073)
[28] Bingjun Li.
Strongly clean matrix rings over noncommutative local rings.
Bulletin of the Korean Mathematical Society 46(1) (2009), 71--78.
MR 2488502 (2010b:16057)
[29] Warren Wm. McGovern.
Neat rings.
Journal of Pure and Applied Algebra 205(2) (2006), 243--265.
MR 2203615 (2006j:13025)
[30] G.S. Monk.
A characterization of exchange rings.
Proceedings of the American Mathematical Society 35(2) (1972), 349--353.
MR 302695 (46 \#1839)
[31] Bruno J. Mueller.
On semi-perfect rings.
Illinois Journal of Mathematics 14(3) (1970), 464--467.
MR 262299 (41 \#6909)
[32] W.K. Nicholson.
Rings whose elements are quasi-regular or regular.
Aequationes Mathematicae 9(1) (1973), 64--70.
MR 316497 (47 \#5044)
[33] W.K. Nicholson.
I-rings.
Transactions of the American Mathematical Society 207 (1975), 361--373.
MR 379576 (52 \#481)
[34] W.K. Nicholson.
Lifting idempotents and exchange rings.
Transactions of the American Mathematical Society 229 (1977), 269--278.
MR 439876 (55 \#12757)

26
[35] W.K. Nicholson.
Strongly clean rings and Fitting's lemma.
Communications in Algebra 27(8) (1999), 3583--3592.
MR 1699586 (2000d:16046)
[36] W.K. Nicholson and K. Varadarajan.
Countable linear transformations are clean.
Proceedings of the American Mathematical Society 126(1) (1998), 61--64.
MR 1452816 (98d:16043)
[37] W.K. Nicholson and Yiqiang Zhou.
Clean rings: a survey. Advances in ring theory, 181--198,
World Scientific Publishing, Hackensack, NJ, 2005.
MR 2181857
[38] W.K. Nicholson and Yiqiang Zhou.
Strong lifting.
Journal of Algebra 285(2) (2005), 795--818.
MR 2125465 (2006f:16006)
[39] Daniel R. Shifflet.
Optimally Clean Rings (Ph.D. dissertation).
Bowling Green State University (2011), 70 pages.
Available online through http://etd.OhioLINK.edu.
\v
[40] Janez Ster.
Corner rings of a clean ring need not be clean.
Communications in Algebra 40(5) (2012), 1595--1604.
MR 2924469
[41] Askar Tuganbaev.
Rings close to regular. Mathematics and its Applications, 545.
Kluwer Academic Publishers, Dordrecht, 2002.
MR 1958361 (2004d:16018)
[42] John von Neumann.
On regular rings.
Proceedings of the National Academy of Sciences of the United States of America 22(12) (1936), 707--713.
Available online at http://dx.doi.org/10.1073/pnas.22.12.707.
[43] Zhou Wang and Jianlong Chen.
On two open problems about strongly clean rings.
Bulletin of the Australian Mathematical Society 70(2) (2004), 279--282.
MR 2094295 (2005e:16050)
[44] Zhou Wang, Jianlong Chen, Dinesh Khurana, and T.Y. Lam.
Rings of idempotent stable range one.
Algebras and Representation Theory 15(1) (2012), 195--200.
MR 2872487
[45] R.B. Warfield, Jr.
A Krull-Schmidt theorem for infinite sums of modules.
Proceedings of the American Mathematical Society 22(2) (1969), 460--465.
MR 242886 (39 \#4213)
[46] R.B. Warfield, Jr.
Exchange rings and decompositions of modules.
Mathematische Annalen 199(1) (1972), 31--36.
MR 332893 (48 \#11218)

27
[47] Junchao Wei and Libin Li.
Quasi-normal rings.
Communications in Algebra 38(5) (2010), 1855--1868.
MR 2642031 (2011i:16065)
[48] Hua-Ping Yu.
Stable range one for exchange rings.
Journal of Pure and Applied Algebra 98(1) (1995), 105--109.
MR 1317002 (96g:16006)
[49] Hua-Ping Yu.
On quasi-duo rings.
Glasgow Mathematical Journal 37(1) (1995), 21--31.
MR 1316960 (96a:16001)

Index

\pi -regular ring, 10

primitive idempotent, 4

artinian primitive factors, 5


artinian ring, 7

quasi-duo ring, 5
quasinormal ring, 5, 22

clean element, 1
clean ring, 1

regular element, 9
regular ring, 9
right T -nilpotent, 7
right artinian ring, 6
right noetherian ring, 11
right perfect ring, 7
root clean ring, 5

Dedekind finite, 22
exchange property, 2
exchange ring, 2
full idempotent, 4

semilocal ring, 8
semiperfect ring, 8
semipotent ring, 3
semiprimary ring, 7
semiprime ring, 17
semiregular ring, 12
semisimple ring, 7
similar matrices, 17
stable range one, 12
strongly \pi -regular ring, 10
strongly clean ring, 21
strongly regular ring, 10
suitable ring, 2

I-ring, 3
I0 -ring, 3
idempotent stable range one, 13
idempotents lift modulo I, 2
left T -nilpotent, 7
left artinian ring, 6
left noetherian ring, 11
left perfect ring, 7
local idempotent, 4
Morita context, 17
--- with zero pairings, 18
ring of a ---, 18
Morita invariant, 4

T -nilpotent, 7
trivial extension, 22

NJ-ring, 19
noetherian ring, 12

unit regular element, 9


unit regular ring, 9
unit semiregular ring, 12
units lift modulo I, 2

optimally clean ring, 21


orthogonally finite, 8
pairings, 18
perfect ring, 8
potent ring, 3
primitive ideal, 5

von Neumann finite, 22


zero-dimensional ring, 11
Zorn ring, 10

28

You might also like