SOME ASPECTS OF RING THEORY

A. W. GOLDIE Ring theory arises from two main sources; (i) associative algebras, especially group-algebras, and matrix theory, (ii) polynomial rings, algebraic number theory and algebraic geometry. The first of these sources led to the development of artinian rings which is characterised by its non-commutative nature and develops by the methods of the theory of ideals and modules. The second led to commutative noetherian rings which has been studied by more basic and natural methods permitted by the commutative law of multiplication, together with these of ideal andfieldtheory. At first the two developments were wholly separate but have drawn together in the last decade, because of recognition of the universal value of the theory of modules and the gathering strength of the non-commutative theory. Since an artinian ring is noetherian, it is natural enough to examine possible properties of noetherian rings which generalise in an illuminating way suitable parts of both the artinian theory and the commutative theory. The search for results of this type has already brought to light new classes of rings and revitalised some of the older results, especially in artinian rings.
1. Domains, prime and semi-prime rings

The technique of forming a ring of quotients is extensively employed in commutative ring theory, but is not used in artinian rings because nothing is gained, they are already quotient rings. A non zero-divisor is already a unit of the ring. In Van der Waerden [116], the first edition of 1930, the author asked whether a non-commutative integral domain can be embedded in a division ring, but the edition of 1937 carries a reference to the famous counter-example of Malcev [85]. In the meantime Wedderburn [117] had shown that the quotient method worked for euclidean domains and in Ore [92] there appeared the condition which is given below. Some special cases of polynomial domains were discussed in Littlewood[81]. Interest waned until Tamari [112] showed in 1953 that a Birkhoff-Witt algebra (this is a noetherian domain) has a quotient division ring. The general problem was settled in Goldie [37]. These papers study the quotient problem only, the general question of embeddability being a different matter; the reader is referred to B. H. Neumann [91] and P. M. Cohn [18]. Let R be a ring. An element of R is regular if it is neither a left nor right zero divisor. A set # c R of regular elements is a right Ore set if any pair aeR,
Received 14 February, 1969.
[BULL. LONDON MATH. SOC, 1 (1969), 129-154]
BULL. 2

130

A. W. GOLDIE

gives rise to another pair ateR, q e # such that

ac{ = cav A ring Q is a right quotient ring of R with respect to # provided that

0) 2 ^ ;
(ii) the elements of # are units in Q; (iii) the elements of Q have the form ac~1-, aeR, ce#.

Q exists if and only if ^ satisfies the right Ore condition, this result is due to Ore for domains and to Asano generally; see Jacobson [59]. The necessity of the Ore condition is clear, since c~la must be expressible in the form at c^1. Its sufficiency is usually established by an elaborate form of the commutative procedure using classes of ordered pairs (a, c), but there is a different approach in Lambek [74 Chapter 2], with the help of partial endomorphisms. The existence of a quotient ring depends on the rank of the ring. A ring R has finite (right) rank when it has no infinite direct sums of non-zero right ideals. A right ideal U of R is uniform if U ^ 0 and Ui n U2 i1 0 for any pair of non-zero right ideals Uu U2 which lie in U. Each non-zero right ideal / contains a uniform right ideal. Suppose not, then R has a sequence of non-zero right ideals {/,/'} with the properties:

Now 11 I2' ... is an infinite direct sum, the existence of which contradicts the assumption that R has finite rank. Let 17! ... Um and V1 ... Vn be direct sums of uniform right ideals of R, each having maximal length. Each sum is an essential right ideal, it has non-zero intersection with every non-zero right ideal of R. In more usual terms, R is an essential extension of each sum, considered as a right /^-module. Now m = n, which was proved in Goldie [37] by a procedure following that in the Steinitz exchange theorem. The invariant integer m is the (right) rank of R. The term infinite right rank covers the case where R has infinite direct sums of right ideals.
THEOREM (1.1). The (right) rank of an integral domain R takes only the values 1 and oo. R has a right quotient ring if and only if the rank is 1.

For the satisfaction of the Ore condition is equivalent to R itself being a uniform right ideal, in other words, R has rank 1. When R has finite rank it contains a uniform right ideal, and because R is a domain, one readily verifies that R itself is a uniform right ideal. The requirement of finite rank is a weak sort of maximum condition on ideals, it is vacuous for commutative domains. An artinian domain is already a division ring.f Certainly we can say that a right noetherian domain has rank one. This t By an artinian ring we mean either a right or left (or both) artinian ring. The precise gloss is unnecessary in the general text.

SOME ASPECTS OF RING THEORY

131

leads to the important case of Birkhoff-Witt algebras or universal enveloping algebras. Let L be a finite-dimensional Lie algebra over a field K and U(L) be its universal enveloping algebra, see Jacobson [61; Chapter 5]. Choose a basis uu ...,un of L over K and suppose that in L
fc = l

U(L) is isomorphic to the ring of polynomials over K in indeterminates x l 5 ...,xn subject to the equations
n
Y \" ^ " Y V *_/ - * i *,- -A-j \ ' cti y * 7 7 .^ 1 /_, / ijk A'k, I, J 1, %>t ...,n.

U(L) is a right and left noetherian domain and thus has right and left quotient rings; these are isomorphic division rings. U(L) exhibits the better sort of behaviour, but the following instance is different. Example. Let F be a field with an isomorphism a -> a which is not an automorphism and F be the subfield of images. Let F[x, ] be the ring of polynomials of the form a0 + xax +... + xk ak, k ^ 0, at e F. Multiplication follows the rule ax = xa. The right ideals of this domain are principal; it has a right quotient ring, but not a left quotient ring. The left Ore condition breaks down because F ^ F. This is worth noting as a principal right ideal ring which is not left noetherian. Simple noetherian rings have never been determined and results obtained for simple rings usually extend to a wider context. The following is an example of a simple domain of the easiest one-sided ideal form but not a division ring. Example. Let F() be the ring of rational functions in . over a field F of zero characteristic and take F [x, ] to be the ring of differential polynomials in t over F(^). Its elements are subject to the rule

at = t(x+du/d; aeF(^).
This is an integral domain, its right (and left) ideals are all principal. It is a simple ring (no proper two-sided ideals). The ideas developed so far for domains are now applied to a wider class of rings. A ring R is semi-prime if it has no nilpotent ideals. A ring R is a prime ring if the condition aRb = 0, where a,beR, implies that either a = 0 or b = 0. An ideal A of a ring I? is a semi-prime (prime) ideal when the factor ring R/A is a semi-prime (prime) ring. A prime ring is a semi-prime ring. A commutative prime ring is an integral domain and the semi-prime case has little interest. An artinian semi-prime ring is semi-simple and if prime is a simple ring, which is a full matrix ring over a division ring.

132

A. W. GOLDIE

The following theorem is due to Goldie [38]. It had been proved in [37] for prime rings under two-sided conditions and in Lesieur-Croisot [76] for prime rings under one-sided conditions. Later proofs appear in Gabriel [33], Lambek [74], Procesi-Small [99].
THEOREM (1.2). A ring R has a right quotient ring Q which is a semi-simple artinian ring if and only if

(1) R is a semi-prime ring, (2) R has finite right rank, (3) JR has maximum condition on right annihilators. We take the present opportunity of inflicting another proof on the reader; it has at any rate the merit of brevity. The third condition needs explanation. Let S be a subset of R. Denote

r(S) = (xeR; Sx = 0).

Then r(S) is a right ideal, the right annihilator of S. Left annihilators are similarly defined, a typical one is An interesting ideal in any ring R is the set Z(R) = (zeR; r(z) is an essential right ideal). Z(R) is a two-sided ideal of R. In the prescence of the maximum condition on right annihilators Z(R) becomes a nil ideal. The concept is due to R. E. Johnson [63].
LEMMA (1.3). Let Rbe a ring satisfying conditions (1) and (3). Then a nil right (left) ideal of R is zero.

Proof. As aR is nil if and only if Ra is nil, we consider Ra ^ 0. The set of right annihilators r(za), z ranging over R with za # 0, has maximal elements. Let r(b) be maximal, b = ta. Suppose that (yb)k = 0, (yb)11'1 ^ 0, where yeR, and k > 1. Then r(b) = r((j;Z>)'c~1)) by the maximality and hence byb = 0. Thus bRb = 0 and (Rb)2 = 0, b = 0, since R has no nilpotent left ideals. This contradiction shows that Ra = 0 and the lemma follows.
LEMMA (1.4). Let R be any ring which satisfies conditions (2) and (3). For each aeR there exists n > 0 such that a"R+r(a") is an essential right ideal.

Proof. Because of (3) there is an n ^ 1 such that r(o") = r(an+1). a"R n r(a") = 0. Let / be a right ideal and suppose that

Then

I n (cfR+ria")) = 0 .
The sum I+a I + a I+... is direct and, because R has finite rank, we conclude that 7 = 0. Lemma (1.3) is due to Utumi [114] as regards proof and Lemma (1.4) uses an idea of Lesieur-Croisot [76].
n 2n

SOME ASPECTS OF RING THEORY

133

Proof of Theorem (1.2). Under the stated conditions we prove that an essential right ideal E contains a regular element. E is not a nil ideal; it has an element In the latter av = 0 with r(ax) = rfc,2). Either r(aj) r\E = 0 or r(a^) nE^O. case, choose a2 e r{a^) n E with a2=0 and r(a2) = r(a22). If r ^ ) n r(a2) n E # 0, then the process continues. At the general stage we have a direct sum atR ... ak R (KaJ n ... n r(afc) n ), where 0 ^ ^ ) n ... n r(ak-i) n E and afc ^ 0, r(ak) = r(afc2). The process has to stop, because R has finite rank; let this happen at the k-th stage. Then r(fli) n ... n r(ak) n E = 0 = r(ax) n ... n r(aft) and hence r(fl12 + ...+Jk2) = (r(fll) n ... n r() = 0. Let Z be the singular ideal of R and zeZ. Then z"JR r(z") is an essential right ideal for some n > 0 and r(z") is also essential. Hence z"R = 0, Z is a nil ideal and hence is zero. Set c = at2 + ... + ak2eE; as r(c) = 0 we deduce that cR is essential by Lemma (1.4). Hence l(c) c Z, so that l(c) = 0 and c is a regular element. This establishes the existence of regular elements in R. Suppose that a,deR with d regular and set E = (xejR; axedR). Then dR is essential and hence so is E, so that E contains a regular element dt. The right Ore condition is satisfied and R has a right quotient ring Q. Next suppose that F is an essential right ideal in Q, then F n R is essential in R jR. Now F n J has a regular element, which is a unit in Q, and hence F = Q. Let J be a right ideal and K a right ideal of Q such that J n X = 0 and J K is essential (use Zorn's lemma); then J K = Q. Thus the module QQ is semi-simple and Q is a semi-simple ring. Conversely, let R have a semi-simple right quotient ring Q. Then a right ideal E of .R is essential if and only if EQ = Q. To see this, suppose / is a non-zero right ideal of Q, then / n R # 0 and / nRnE # 0, taking to be essential in .R. Hence / n Q # 0, which means that EQ is essential in Q; then Q = Q as QQ is a direct sum of simple modules. On the other hand, when EQ = Q is given and / is a nonzero right ideal of R then IQ n Q # 0, trivially and hence / n E # 0, so that is an essential right ideal of R. These conditions are equivalent to saying that E has a regular element, because yeEQ and 1 = ec" 1 , eei?, ceR and regular. On the other hand, when E has a regular element, then EQ = Q and E is essential in R. We now conclude that J is a semi-prime ring, because if A is a nilpotent ideal R T of R, l(N) is essential as a right ideal and has a regular element. Thus N = 0. Let S = (E/ a ; aeA) be a direct sum of non-zero right ideals of R which is essential as a right ideal. S has a regular element c, expressible as a finite sum

134

A. W. GOLDIE

Now cR is essential and lies in J ai + ...+J a n ; it follows that A has only the indices a l s ...,a n , and R has finite right rank. Finally, the maximum condition holds for right annihilators, because

for any non-empty subset S of R, the subscripts denoting the ring in which the annihilator is taken. The following corollary is evident.
COROLLARY (1.5). A prime ring R has a simple artinian ring Q as its right quotient ring if and only if conditions (2) and (3) hold.

There also follows, since Q is a full ring of matrices over a division ring; Q = Dn, say. (1.6). A prime ring R with conditions (2) and (3) contains a prime ring R' which also has Q as its right quotient ring and R' is a full matrix ring Cn, where C has the division ring D as its right quotient ring.
COROLLARY

This corollary is due to C. Faith and Y. Utumi [29]. 2. Rings with nilpotent ideals The earliest result is due to Levitski in 1939 and states that in a right noetherian ring JR every nil right ideal is nilpotent. Nowadays this theorem is an easy consequence of Lemma (1.3). It readily follows that R has a maximal nilpotent ideal N which is um'que, the nilpotent radical and R/N is a semi-prime ring. N is the intersection of the prime ideals of R and indeed is the intersection of the minimal prime ideals, which are finite in number. For further details on radicals see Divinsky [23] and Jacobson [60]. It is impossible to generalise the procedures of Theorem (1.2) in order to obtain regular elements. Instead the concept has to be taken over factor rings. Let A be an ideal of R and set
< l(

6' (A) = (ceR; cxeA,xeR

implies that xe A)

(A) = (CER; xceA, xeR implies that xe A)

Cases

<$ (A) = # ' (A) n >( (A).

Under the conditions of Theorem (1.2) we have # (0) = W (0) c '#(0). occur where <#'(0) # '#(0), see [38] and [107].
THEOREM

(2.1). Let R be a right noetherian ring with nilpotent radical N and let P l 5 . . . , Pn be the minimal prime ideals of R. Then (1) '(0) < <f(N); = (2) (3)

(4) let aeR, ce#'() then ateR, c^^iN)

exist with act = cav

SOME ASPECTS OF RING THEORY

135

We know that #(JV) = <g'(N) c '<g(N) by applying Theorem (1.2) to the ring R/N. Part (1) of this theorem is due to Djabali [28] and (2), (3) are due to Goldie. (3) is a short way of saying that ce^'(0) and neN together imply that c + ne<tf'(O). This shows that no modification of the proof of Theorem (1.2) will give regular elements. Part (4) is implicit in Small [108]. It is tempting to call this the pseudo Ore condition. It can be used to prove (1). This approach leads to the valuable theorem due to Small [108]. (2.2). A right noetherian ring R has a right quotient ring Q, which is a right artinian ring, if and only if
THEOREM

That #(0) = #(N) is necessary had been proved earlier by Talintyre [111]. After this point the quotient problem for noetherian rings becomes very difficult and no decisive results have been obtained. There are three aspects to the problem: (i) when does a noetherian ring R have regular elements; (ii) when are there enough regular elements to satisfy the right Ore condition; (iii) what is the structure of a quotient ring ? For commutative noetherian rings the matter is settled by the maximal primes of zero, these are the maximal annihilator ideals and they are finite in number. An element is regular if and only if it does not lie in any of these primes, such elements exist if and only if the ring is faithful (Rx = 0 implies that JC = 0). The second question is trivial and, as for the third, a quotient ring is a semi-local ring in which the Jacobson radical has non-zero annihilator. Rings with zero singular ideal are of some interest. Such a ring, if commutative, is semi-prime but in general the structure is very complicated. They are not known in the artinian case except when indecomposable right ideals are uniserial. They are then determined as direct sums of blocked triangular matrix rings over division rings. See Goldie [40], Gordon [45] and Colby-Rutter [19]. The assumption that i? has zero singular ideal is useful technically, because c e # ' ( 0 ) implies that cR is an essential right ideal and hence l(c) = 0. Thus #(0) = #'(0). Djabali [28] settles some cases of the quotient problem for rings with zero singular ideal. In the general case when the existence of the quotient ring is assumed, related properties can be studied. (2.3). Let R be a right noetherian ring with a right quotient ring Q and N be the nilpotent radical of R. Then NQ is the nilpotent radical of Q. Let P be a prime ideal of R then either PQ = Q or #(0) c #(P) and PQ is a prime ideal of Q with PQnR = P. Let P' be a prime ideal of Q, then P' n R is a prime ideal of R and (P' nR)Q = P'.
THEOREM

This theorem is well known for commutative rings but its generalization is not immediate. It depends on Theorem (2.1) and so far is only known for noetherian rings. There is no published proof at present.

136

A. W. GOLDIE

Some necessary conditions for the existence of the quotient ring are obtained as follows. The transfer ideal of #(0) is the largest ideal T such that c + te #(0) for all c e #(0), t e T. There is also a transfer right ideal T", defined in the same way. Clearly they are uniquely defined and T < T". In order to fix the idea we remark that the Jacobson = radical of a ring is the transfer ideal of the group of units. Suppose that a noetherian ring R has a right quotient ring Q and let T, T" be the transfer ideals of R. It is a consequence of Theorem (2.3) that

JnR=T = T',
where J is the Jacobson radical of Q. It follows that T is a semi-prime ideal of R. This may, of course, be true in a ring which does not have a quotient ring but at present the best result known is that T 3 N, see Theorem (2.1). We conclude our remarks on the quotient problem with a reference to Small [107] which gives an example which is a right and left noetherian ring but does not have a quotient ring on either side. This ring satisfies a polynomial identity. For relief we next turn to an easy case, the principal ideal rings. A ring R with unit element is a pri-ring when its right ideals are principal (single generator). The following theorems are due to Goldie [39]. See also Procesi [96] and Johnson [68] for a different treatment.
THEOREM

(2.3). (1) A semi-prime pri-ring is a finite direct sum of prime pri-r ings. (2) A prime pri-ring is isomorphic to a full ring of matrices over a right Ore domain.

THEOREM (2.4). A pri-ring which is left noetherian is the direct sum of a semiprime pri-ring and an artinian pri-ring. It has a right quotient ring which is an artinian pri-ring.

Difficulties arise with pri-rings which are not left noetherian, see the example in Section 1 and also Jategaonkar [62]. A number of other problems have been studied from time to time. An old conjecture due to Jacobson enquires whether f] J" = 0, where J is the Jacobson
n=l
00

radical, holds in a right noetherian ring. A counter-example is given in Herstein [52], following an idea of Lance Small. The problem is still open for right and left noetherian rings. An ideal T of a ring R is a primary ideal if AB a T, where A, B are ideals of R, implies that either i c T or Bk c T, together with the corresponding property when A and B are interchanged. A strongly-primary ideal T is a primary ideal such that R/T has an artinian quotient ring. Is a primary ideal always strongly-primary? This seems unlikely and a counter-example should throw light on the nature of the quotient process. Many difficulties stem from the lack of a representation of an ideal as an intersection of primary ideals. It is fashionable to avoid this property in commutative work, but we cannot indulge this whim. Indeed primary decomposition does not

SOME ASPECTS OF RING THEORY

137

hold in algebras, for if it held in group algebras then allfinitegroups would be soluble. The only decomposition theory which has anything to offer in the general case is the tertiary theory. There have been many attempts to deal with the questions of uniqueness and existence, most theories succeed in one place and fail in the other. The tertiary theory succeeds in both respects. Unfortunately, it is very difficult to relate to other aspects of the ring structure. We refer the reader to Curtis [20], for earlier work and Lesieur-Croisot [75] for the tertiary theory. Here we restrict our summary to the case of a right noetherian ring R having unit element, which has some simplifying features. Let M be a right K-module. Index its essential submodules as Ma (a e A) and define the radical of M to be rad M = (xeR; Ma x = 0, some a e A). Clearly rad M is an ideal. When M isfinitelygenerated, rad M is a finite intersection of prime ideals. Let / be a right ideal then the tertiary radical of / is r(/) = rad(K-J) when / is an ideal of R we have r(I) 3 I, but this need not hold for right ideals. Let R b e a simple artinian ring, / = eR, e2 = e ^ 0, 1, then r(I) = 0. The definition can be rephrased for left modules and left ideals and indeed for bimodules. Thus for an ideal T we can define a radical R(T) by the rule: R(T) is the set of elements aeR such that every ideal B c T contains an ideal B' $ T with B'a c T. This definition can be rephrased on the left as well. It follows that an ideal T has four tertiary radicals; as a right or left ideal and on the right or left as a (twosided) ideal. These can differ, certainly R(T) ZD r(T), but equality occurs for artinian rings. This is a nice exercise in socle play. Whether R(T) = r(T) for noetherian rings is not yet settled. In order that the reader can keep his bearings, we mention that in a commutative ring these radicals all reduce to VT = (xeR; x"eT for some n > 0). A right ideal / is tertiary when bRa d , b$I, implies that aer(I). Then r(I) is a prime ideal. A right ideal V is irreducible when R V is a uniform right .R-module. Then V is a tertiary right ideal. It follows that every proper right ideal is a finite intersection of tertiary right ideals.
THEOREM (2.5). Let I be a proper right ideal of the right noetherian ring R and I = It n ... n Ik where the Ij are tertiary right ideals with associated prime ideals Pjlj = l,...,k. Then

The intersection of a finite set of tertiary right ideals, each having the same associated prime ideal P, is again P-tertiary. This enables a decomposition to be

138

A. W. G0LD1E

brought into reduced or normal form (the associated primes are distinct), as in the commutative theory. A uniqueness theorem now follows.
THEOREM (2.6). Let I = Iln...nIh = Jln...r\Jk, where the decompositions are reduced, the Ia, Jp being tertiary right ideals. Then h = k and the two sets of associated prime ideals coincide.

These results can be carried out for left ideals and repeated for two-sided ideals T, using R(T) instead of r(T). In the formal sense the theory is entirely satisfactory and indeed is the only possible theory for which these theorems hold, see [101]. Nevertheless, it has proved difficult to apply to the study of the structure of noetherian rings, because the nature of a tertiary ideal is difficult to understand. Moreover the tertiary radical has to destroy the partial order even for two-sided ideals. For example, r(P) = P for all prime ideals P of R. Since (0) c P, preservation of partial order would imply that r(0) c= P for all primes P, which would mean that r(0) is the nilpotent radical. This is certainly not the case as it would lead to the existence of a primary decomposition for ideals. 3. Localisation Non-commutative local rings occur surprisingly often, although it is difficult to provide a natural construction of a local ring in the general situation and its value as a method is as yet uncertain in non-commutative rings. A local ring is defined in Cartan-Eilenberg [15] to be a ring R with unit element such that the elements which do not have a left inverse form a left ideal M. Then M is a two-sided ideal which contains all proper left and right ideals; in particular, R/M is a division ring. In practice the definition is found to be too restrictive and is modified as follows. A local ring is a ring R with unit element such that the Jacobson radical J is a maximal ideal and R/J is an artinian ring. Evidently J is the only maximal ideal of R. A full ring of n x n matrices over a Cartan-Eilenberg local ring is a local ring in the new sense, although it is unlikely that all local rings have this form. A more general construction is studied in Gabriel [33], see also Bourbaki [14] and Murdoch [89], proceeding in a context much more general than is considered here and in effect using partial endomorphisms. These do not lead, as far as can be seen, to local rings of our sort except in special cases. All methods reduce to the usual construction for commutative rings. A set 3F of right ideals of a ring R is topological provided that (1) & is a filter, (2) aeR, Fe^ implies that a'^F = (xeR; axeF)e^.

R is thus regarded as a topological ring having a basis of neighbourhoods of zero consisting of right ideals.

SOME ASPECTS OF RING THEORY

139

When M is a right .R-module, define the ^-singular submodule to be

= (meM; r(m) e &), where r(m) = (xeR; mx = 0).

A closure operation is defined on submodules N of M by cl^ N = (m e M; mF cz N, some F e &). Topological sets 3F, $ of right ideals can be multiplied together, the set comprises those right ideals / for which cl^/ belongs to 2P'. The set tpy is again Ge&. topological and contains the ideals FG\Fe&r, A topological set $F is idempotent when $F2F = !F'. This is the important case and occurs when J5" satisfies the additional axiom: (3) c l , / e & if and only if / e 3F. The useful construction of an idempotent topological set is obtained by taking a multiplicatively closed subset S of elements of R which does not contain zero and defining 3F(S) to be the set of right ideals F of R such that a~lF meets S for every aeR. It is readily verified that ^(S) satisfies axioms (1), (2), (3). A multiplicatively closed subset S cz R which does not contain zero, satisfies the right Ore condition when, for given aeR, seS there exist a^ e R , si eS with asx = sav Then the set of right ideals which contain a right ideal sR, seR, is an idempotent topological set and is a subset of #"(S). When S has only regular elements, the right Ore condition is essentially that considered in Section 1 and is equivalent to the existence of a classical right quotient ring Rs, with elements of the form

as'1, aeR, seS.

It is now assumed that we are dealing with a (left and right) noetherian ring R with unit element. Let P be any prime ideal of R; P ^ R. We know from Theorem (1.2) that < = >cti, where

# = W(P) = {ceR; cxeP implies that xeP}

l(

= '<#(P) = {ceR; xceP implies that xeP}

This depends on knowing that the prime ring R/P has a quotient ring on left and right. # is multiplicatively closed. Moreover cR + Pe^ffi, whenever c e # . 5 Now set J " = &{$>) and define a corresponding topological set of left ideals
<g = <$(<$). For example Rc+Pe<&.

Now define a closure operation for any two-sided ideal A or R by

k(A) = (xeR; GxF a A, some Ge&, Fe&).

k(A) is an ideal containing A and k(kA) = kA. This operation enables symbolic powers of P to be defined; there are two sorts; left symbolic powers Hn and right symbolic powers Kn where n = 1, 2,... given by P; Hn+l = k(PHn); Kn+i = k(KnP).

140

A. W. GOLDIE

The following theorems were given in Goldie [41].

THEOREM (3.1). Let R be a right and left noetherian ring, P a prime ideal and Hn, Kn defined as above. Then

(1) HrHs c Hn, whenever r+s = n, n = 1, 2 , . . . , (2) R/Hn satisfies the left and right Ore conditions with respect to the set # = {c+Hn\ ceW} and # is a set of regular elements, (3) Hn = Kn, (4) <6n is the set of all regular elements of R/Hn, (5) Hn is a P-primary ideal. Properties (2) and (4) show that R/Hn has a right and left quotient ring which we denote by Q(ri).
THEOREM

(3.2). Under the conditions of Theorem (3.1),

(1) the quotient ring Q(ri) of R/Hn is an artinian primary ring, (2) there is a canonical epimorphism Q(n + \)~* Q(n), (3) the inverse limit Q, of the rings Q(n),n = 1,2,... under the canonical epimorphisms is a complete local ring, (4) Q is a full kxk matrix ring over a complete local ring L, (5) L is a local ring in the sense of Cartan-Eilenberg. Q is complete in the topological sense, the topology being that arising from the powers of the maximal ideal $ of Q; this corresponds to the Zariski topology for
00

commutative rings. The topology is hausdorff, because the property f] $" = 0

n=l

holds. The integer k of part (4) is obtained from the quotient ring Q(l) of the prime ring R/P. Q(l) is isomorphic to the full matrix ring Dk, D a division ring. Note that g/i& ^ Dk. Since the classical (or commutative) method of localisation is not possible in general the " usual" local ring has to be sought in another way. An existence theorem is the following. THEOREM (3.3). Let {(^l a e A} be the family of local rings with corresponding maximal ideals {Ma; a e A} such that Ra&czQ; Pac=Macijfr; aeA. Then for all aeA ( l ) M a = jfrnQ a , (2) {R,V}<^Qa, = 1,2,..., (3)Hn = MannR, (4) QJMa s g/jft. Let Q = (f)Qa; a e A ) and M = ((]Ma; aeA). Q is a local ring with maximal ideal M. Q is a member of the family.

SOME ASPECTS OF RING THEORY

141

The local ring Q can be derived by construction as follows.

THEOREM

(3.4). Define a sequence of rings Q(n), n = 1,2,..., where RczQ' czQ"c ...c(2 ( f l ) c . . . e g

with Q' = {R,^-1} and Q(n+1) = { Q ^ , ^ " 1 } ; M(n) = til n Q(n); and ^ (ll) m rte n# Q(fl). Set Q* = UQ(n)> ^ M* = UM(n), = 1,2,.... 7%i Q* = Q and M* = M, as defined in Theorem (3.3). The ring {R, # - 1 } is the subring of (5 generated by R and the inverses of elements of (. However it should be noticed that in these theorems the symbol R has been used to denote R/H, where H = f]Hn, n = 1,2, It is R/H which is embedded in (5. This accords with the commutative case, since there we have 0P (n) , n = 1, 2, ... = (xeR; xc = 0 some ce<tf(P)). The usual procedure for clearing out zero divisors before localising has to be avoided, because (xeR; xc = 0, some ce^(P)) is no longer an ideal. See McConnell [83] for an interesting example. The classical construction occurs in a number of cases and its non-commutative form is as follows. To localise on the right it is necessary to suppose that the elements of #(P) satisfy the right Ore condition. For each pair a e R, ce #(P) we need the existence of a pair at eR, c1 6 ^(P) with acy = cat. The ^-component of zero is (0 :#) = (xeR; xc = 0 some c e # ) . Then (0 : #) is a two-sided ideal of R. Moreover if the elements of %> are left regular (xc = 0, x = 0) then they are regular. It follows that on passing to the ring R , . = R, the image # of # is a set of regular elements which satisfy the right (U: <r) ^ Ore condition. Form the right quotient ring of JR with respect to #; this ring is a local ring RP. The procedure works whenever R is right noetherian and one finds that RP is also right noetherian. The restriction imposed by the Ore condition is on the prime ideal P rather than on R itself because act = cat has only to be verified when ae P. See Goldie [41] for details or an account of a similar type in Gabriel [33]. It is quite likely that in a given ring R, classical and non-classical localisations may both exist, the particular sort dependent on whether we have a " good " or " bad " prime ideal to deal with. The serious weakness of the general construction lies in that it is not known whether Q or Q are noetherian or otherwise. Many other problems arise with local rings. Is every factor ring of a local ring, even when noetherian, also a local ring? This is a question of whether two-sided ideals are closed in the Zariski topology and can be asked for one-sided ideals also. A discussion of the relationship between these and classical localisation is given in [41].

142

A. W. GOLDIE

This question is generalised in McConnell [84] and Hinohari [57]. Let R be a ring with a two-sided ideal / such that f]I" = 0, n = 1,2,..., and suppose that (i) R/I is a right artinian ring, (ii) / is finitely generated as a right ideal, (iii) R is complete in the /-topology. For a ring with these properties we have (3.5). Let E be a hausdorjf right R-module ((]EIn, (n = 1,2,...) = 0) and F a submodule of E. Consider the conditions:
THEOREM

(i) F is a finitely generated R-module, (ii) For each k > 0, then exists n > 0 with F n El" < FIk, = (iii) F is a closed submodule of E, viz F = f](F+EIn); n = 1,2,....

Then (i) implies (ii) and (iii). / / E is finitely generated then (ii) implies (i) and the induced topology on F coincides with the l-topology. The implication (i) -(iii) is due to Hinohari; the remainder to McConnell. 4. Universal enveloping algebras Suppose that L is a Lie algebra of finite dimension over a field k and let U(L) be the universal enveloping algebra of L. Refer to Section 1 and Jacobson [61; Chapter 5]. We give here a brief description of properties of U(L) which are related to the ideas given so far. This is but a small part of the extensive literature on Lie algebras and is restricted to the cases of nilpotent or solvable algebras. We need to define the algebras An = Alt(k), n 1,2,.... These are given in Hirsch [58] and D. E. Littlewood [81]. An is the algebra generated over k by 2 generators phqx where / = 1,..., subject to the equations [Pi, qi\ = Pi qi-qiPi = i; \j>i, qj\ = [Pi,pj\ =fo,>qjl = o (/ # ; ) . The elements p^'q^' pninqnJn (ia,j* > 0) are a basis of the vector space An over k. The centre of An is isomorphic to k and An = Ax...A1.

An is a left and right noetherian domain; it is a simple algebra when k has characteristic zero.
THEOREM (4.1). Let L be a nilpotent Lie algebra of finite dimension over a field k of zero characteristic. The following properties of a two-sided ideal IofU = U(L) are equivalent:

(1) the centre of U/I is afield, (2) U/I is isomorphic to An for some n, (3) / is a maximal ideal of U, (4) / is a primitive ideal of U.

SOME ASPECTS OF RING THEORY

143

This theorem is due to Dixmier [26] when k is algebraically closed and nondenumerable. He also proved most of the statements in the general case except for (4) => (3) which is due to Quillen [100] and (3) => (2) which is due to Gabriel and Nouaze" [34]. The latter paper introduces some new methods of a functorial nature.
THEOREM (4.2). Under the circumstances of Theorem (4.1), the following properties of an ideal IofU are equivalent,

(1) I is a prime ideal of U, (2) U/I is an integral domain, (3) the centre of R/l is an integral domain. A prime ideal / with property (2) is said to be completely prime. In the case of non-zero characteristic there is a counter-example due to McConnell. Let L be the nilpotent Lie algebra over a field k of characteristic 2, with a basis x, y, z such that [x, y] = z and [x,z] = [y,z] = 0. There is a chain of prime ideals in U, Pl = (0)aP2 = (x2)(zP3 = (x,y,z). Also P l 5 P3 are completely prime but R/P2 has a quotient ring which is a 2 x 2 simple matrix algebra. It is interesting to note that localisation is classical at each P,-. Part of these theorems still holds for solvable Lie algebras and is again due to Dixmier [27].
THEOREM (4.3). Let L be a solvable Lie algebra over a non-denumerable, algebraically closed, field k of zero characteristic. Then every prime ideal P of U is completely prime. Also P is a primitive ideal if and only if the centre of the quotient rint of U/P is k. THEOREM (4.4). With conditions on the field k as in Theorem (4.3) and L any Lie algebra of finite dimension, the following properties are equivalent

(1) L is solvable, (2) every primitive ideal of U is completely prime, (3) every prime ideal of U is completely prime.
Let R be any ring with 1 and / a two-sided ideal of R. A set xu ...,xneR centralising set of generators of / when / = (xu ...,xn) a n d
n

is a

x, e centre R; Xt+Cx,, ...,X;_,)e centreThe property can be generalised by using the subset of R, JV(K) = (xeR; xReRx).

- ; i > 1.

A set Xj,..., xn E R is a normalising set of generators of the ideal / when I = (xlt ...,xn)

144

A. W. GOLDIE

and
CieN(R);

/
\xi>

n
'>xi-i

These definitions and the following theorem are due to McConnell [82], [83].
THEOREM (4.5).

(1) Let L be a nilpotent Lie algebra over any field. Then any two-sided ideal of U(L) has a centralising set of generators. (2) Let L be a solvable Lie algebra over an algebraically closed field of zero characteristic. Then any two-sided ideal of U(L) has a normalising set of generators. When this theorem is applied to the ideal M of U(L), which is generated by L, then a theorem due to Kaplansky [70], or its extension due to Fields [32], shows that the global dimension of U is ^ dim L. It follows that these dimensions are equal, by a theorem due to Roy [105]. This theorem has relevance to the question of localisation in U(L). Refer to Section 3 and to McConnell [83].
THEOREM (4.6) Let R be a right and left noetherian ring with unit element and P be a prime ideal of R. If either

(i) every ideal of R has a centralising set of generators, or (ii) P is completely prime, H = 0 and every ideal of R has a normalising set of generators, then Hn = (xeR; xceP") = (x; cxeP"), some ce<$(P).
COROLLARY Let R and P satisfy one of the conditions (i), (ii). / / P has the AR property then Rp is classical. This is true for a nilpotent Lie algebra L.

An ideal / of R is said to satisfy the AR-condition when for each right ideal E of R there exists n > 0 such that EnFczEI. The following theorem along these lines appears in Nouaz6-Gabriel [34].
THEOREM (4.7). Let R be a right noetherian ring with unit element and I be an ideal of R having a centralising set of generators. Let E be a finitely generated Rmodule, F a submodule of E. Then for each k > 0, these exists n > 0 such that

FnEFc

FIk.

Theorems on chains of primes are available in U(L), these are due to Gabriel and Rentschler [35]. (4.8). Let L be any Lie algebra of dimension n. Then any chain of primes in U(L) including (0) but excluding R, has length ^ n + l.
THEOREM

SOME ASPECTS OF RING THEORY

145

5. Rings with polynomial identities and related rings The centroid C{R) of a ring R is the ring of those endomorphisms co of the additive group R+ with the property w(xy) = (cox)y = x(coy), x, yeR. Let {xj} be a set of non-commuting indeterminates and consider a polynomial p[x]=p[xl, ...,*] = ScoO")*,-, ...*,-k,

which has coefficients co(i) in C(R) and (/) = (/ l5 ..., ik), the summation being over all ilt..., ik e 1,2,...,. The polynomial /?[*] is a polynomial identity for i? if /*[ri >rn] = 0 f r a ^ ri 5 r,,e# and co(i)R+ ^ 0 for at least one (/). For brevity we say that R is a Pl-rmg. The theory was initiated by Kaplansky in [69] where the important Theorem (5.1) is proved. The early development is given in Jacobson [60], much of its direction is towards the solution of the Kurosch problem for Pi-algebras over a field. For Theorem (5.1) we follow Amitsur [6]. (5.1). Let R be a right or left primitive ring satisfying a non-trivial identity p[x] = 0 of minimal degree d. Then R is a central simple algebra of dimension n2 over its centre and d = In.
THEOREM

Conversely, a central simple algebra of dimension n2 over its centre satisfies a polynomial identity of minimal degree In. The identity satisfied in Theorem (5.1) will be unique to a scalar multiple (from C(R)) and is the standard identity

with summation over all a = (iu ..., i2n) belonging to the symmetric group on (l,...,2n). We are concerned here with the effect of the following theorem due to Posner [94], which caused a decisive shift in the theory of Pi-rings.
THEOREM (5.2). Let R be a prime ring which has a polynomial identity of minimal degree d. Then R has a right and left quotient ring Q which is a central simple algebra of dimension n2 over its centre and d = In. Let C be the centre of Q, then RC = Q. Moreover Q satisfies the same identities as R.

The theorem was proved by Posner for prime algebras and extended in Amitsur [6] to the general case. Amitsur made use of the theory of ultra-products, but this can be avoided without affecting his clear approach to the theorem. See Goldie [44]. The following lemma is fundamental.
LEMMA (5.3). Let R be a prime ring with a polynomial identity and c be a regular element. Then cR ZD Rd where d is a regular element.

This lemma shows that R satisfies a strong form of the right Ore condition since act = ca^ holds with cx = d independent of the choice of the element a. No chain
BULL. 2 2

146

A. W. GOLDIE

conditions on right or left ideals are assumed in Theorem (5.3), because the PIhypothesis guarantees them. See Posner [94]. The extension of these results to semi-prime rings with PI needs care, since a polynomial identity p[x] may become trivial, thus all coefficients are zero, after passage to a factor ring. In particular, a prime ideal P of R is trivial with respect to p[x] when coR c P for all its coefficients co. The following theorem from [6] copes with this difficulty.
THEOREM (5.4). A semi-prime ring R satisfies an identity p[x] = 0 of degree d if and only if R is a subdirect sum of two semi-prime rings Rl}R2 such that p[x] is trivial on R2 and Rt Hn, where n < [d/2] and H is a commutative semi-prime ring.

One would expect Rt to have a quotient ring, but this is unsettled, except under additional conditions, which are such that Rt has a semi-simple quotient ring which is a semi-simple algebra of finite dimension over its centre. This had been proved earlier in Small [108]. Posner's theorem provides a foundation for a non-commutative affine geometry developed in papers by Amitsur and Procesi [7, 97, 98]. In [7] a form of the Hilbert nullstallensatz is proved for Pi-factor rings of the free algebra J F ^ , ..., xn], F a field, xu ...,xn non-commuting indeterminates. One deduces that finitely generated algebras over F which have a polynomial identity are Jacobson rings. This latter property asserts that every prime ideal is an intersection of maximal ideals. The result is extended in the following theorem; see [7]. THEOREM (5.5). Let Q be a commutatuve Jacobson ring and R = O[A-19 ..., xn] be a finitely generated Q-ring in the centre of R, which satisfies a polynomial identity. Then (1) R is a Jacobson ring, (2) if M is a maximal ideal of R then QnM is a maximal ideal of Cl and R/M is a finite-dimensional algebra over the field Cl/Cl n M, (3) if Q is an F-Hilbert algebra then so is R. An algebra A over F is an .F-Hilbert algebra if every simple factor algebra is finite-dimensional over F. The identities which hold in R are assumed to have their coefficients in Q, and to be non-trivial. This theorem has been improved by Procesi [98] with the aid of an interesting form of ring extension. Let R be a subring of a ring S and set CS(R) = (x e Sx xr = rx, all x e R) the centraliser of R in S. Then S is an extension of R when S = RCS(R). S is a finitely generated extension of R when S = R[xlt ...,*] with x,eC s (i?). Notice that the xt need not commute among themselves.

SOME ASPECTS OF RING THEORY THEOREM

147

(5.6). / / S is a finitely generated extension of a Jacobson ring R and has a polynomial identity, then (1) S is a Jacobson ring, (2) if M is a maximal ideal of S, then RnM has finite length over R/R n M, (3) if R is an F-Hilbert algebra, then so is S. Such extensions of noetherian rings are not noetherian; see [98] for a counterexample. Procesi does prove the following results. (5.7) A finitely generated Fl-algebra over a commutative noetherian ring satisfies the ascending chain condition for semi-prime ideals.
THEOREM THEOREM (5.8). Let S be a finitely generated extension of a ring R and assume that S satisfies a polynomial identity with coefficients 1. Then

is a maximal ideal in R and S/M

(1) if R has ascending chain condition on prime ideals then so has S, (2) if R has descending chain condition on prime ideals then so has S, (3) if R has finite rank on primes then so has S. The rank of R on primes is the maximum length of a chain of prime ideals of R. A restricted type of finitely generated algebra is studied in Small [108] and earlier in Curtis [21]. Let A be a commutative noetherian right with unit. An 4-algebra R with unit is finitely generated if it is finitely generated as an i?-module. In these circumstances R is a right and left noetherian ring, as also is its centre. Such algebras include group algebras of finite groups over A and the ring of endorphisms of a finitely generated module over A. Small proves that finitely generated algebras can be embedded in artinian rings and later gives a counter-example to show that this does not hold for general noetherian rings. There are also theorems on chains of prime ideals which anticipate the generalisations of Theorem (5.8). He also obtains a successful solution of the Jacobson conjecture, namely (f) J"; n = 1,...) = 0, where J is the Jacobson radical of R. This is not true for one-sided noetherian rings as an example in Herstein [52] shows, but finitely generated algebras are left and right noetherian rings and the question is still open for these. The study of nil and nilpotent rings is part of the area of polynomial identities. There is an interesting case in Higman [56] which deals with rings which are nil of bounded index. In Herstein and Small [54] the relationship between nil and nilpotent rings is dealt with in the following result.
THEOREM (5.9). A nil ring with ascending chain condition on left and right annihilators is nilpotent. The same conclusion holds for a nil ring which has the condition on right annihilators and has finite rank {see Section 1).

A counter-example due to Sasiada shows that the ascending chain condition on right annihilators is not enough. However it is enough when the ring has a poly-

148

A. W. GOLDIE

nomial identity. In Kaplansky [69] we find the earlier result that any nil PI ring is locally nilpotent. 6. Simple and related rings In this section we give results obtained by a closer study of rings which have simple artinian rings as their quotient rings. One underlying problem is to determine the structure and representation of simple noetherian rings. Some special cases have been dealt with in earlier sections. Thus Kaplansky's theorem shows that a simple ring with polynomial identity is artinian and Theorem (2.3) shows that a simple principal right ideal ring is a full matrix ring over a simple right noetherian domain. But the general problem is difficult, and infuriatingly so, because results for simple rings have a tendency to be true for some wider class of rings. However this field of study has another powerful motivation, the development of a non-commutative arithmetic. This has a history which goes back to the quaternion integers and which received an early development due to the work of Artin [8], Asano [9] and Chevalley [17]. We begin with some results about simple rings. It is known that a simple right noetherian ring R which has Q as its right quotient ring has a unit element when Q has infinite centre. See Robson [103]. The following results are due to Hart [49]. (6.1). A simple ring R with unit, which has a uniform right ideal, is isomorphic to eKn e, where K is a right Ore domain and e is an idempotent element of the matrix ring kn for some positive integer n.
THEOREM

Unfortunately, there is nothing invariant here, e, n, K are all variables and we cannot even assert that K is isomorphic to End^ (U, U) for a uniform right ideal U. Note that a simple ring with unit has a uniform right ideal if and only if it has finite right rank. Hart also proved that EndR(U, U), U a uniform right ideal, is a simple ring if and only if U is projective as an fl-module. This leads to the following improvement of Theorem (6.1) in a special case.
THEOREM (6.2) Let R be a simple ring with unit, which has the ascending chain conditions on right annihilators, and which possesses a projective uniform right ideal. Then R is right noetherian if and only if R is isomorphic to eKn e, where K is a simple right noetherian domain and e is an idempotent in Kn.

In order to study arithmetical properties a number of concepts are needed which generalise those of the commutative theory. These are given in Jacobson [59] and Robson [104], and we mention only those which are needed for the statement of results. A right order R in a quotient ring Q is a sub-ring of Q such that every element of Q has the form ac~x\ a,ceR with c regular. Right orders R, S in Q are equivalent if there exist units c, d, e,f in Q with cRdczS and eSfczR. A right order R in Q is a maximal (equivalent) order if no right order which properly contains R is equivalent to it.

SOME ASPECTS OF RING THEORY

149

A right order R is bounded if any right ideal cR, where c is a regular element, contains a non-zero two-sided ideal. For example, a prime ring with polynomial identity is a bounded order because of Amitsur's lemma. A simple ring with unit is a maximal equivalent order. However the class of orders equivalent to a given one need not have a maximal member in the general situation. A right order R with unit is an Asano right order if the (fractional) i?-ideals in Q form a group under multiplication. The following theorem is due to Robson [104] and had been proved earlier by Asano [9] for bounded orders. (6.3). Let R be a right order with unit element in a quotient ring Q. The following properties are equivalent:
THEOREM

(1) R is an Asano right order, (2) JR is a maximal right order and every integral R-ideal is a projective right R-ideal, (3) For each integral R-ideal T there exists an R-ideal T * with TT* = T*T = R,

(4) The R-ideals form an abelian group under multiplication. In order to develop an arithmetic of right ft-ideals it is necessary to involve the orders which are equivalent to R and to strengthen the assumptions on R. For this purpose Robson [104] introduced the (right) Dedekind right order which is a maximal right order in which every integral right .R-ideal is projective. This also generalises an earlier idea due to Asano who had studied such orders in the bounded case. The ring of polynomials F[x, y] over a field F of characteristic zero such that xyyx = 1 is a noetherian hereditary domain and is simple. It is a Dedekind right (and left) order but is not a bounded order.
THEOREM (6.4). Let R be a Dedekind right order. The class of all right S-ideals, where S runs over all right orders equivalent to R, forms a Brandt groupoid under ordinary multiplication, wherever this is defined.

The definition and some properties of Brandt groupoids are to be found in Jacobson [59]. Thus far the quotient ring Q can be arbitrary but representations of Asano and Dedekind orders can be found when Q is a simple artinian ring; these parallel the Theorems (6.1) and (6.2) and are proved by the same methods as in Hart [49].
THEOREM (6.5). Let R be an Asano right order in a simple artinian ring. Then R is isomorphic to eKn e for some idempotent e in Kn, where K is a right order in a division right D, and n > 0.

When R has a projective uniform right ideal then K is an Asano right order in D. For Dedekind right orders this leads to a characterisation, because they are right hereditary rings.

150

A. W. GOLDIE

COROLLARY. A right Dedekind prime ring has the form eKn e, where e is idempotent in Kn, and K is a right Dedekind domain. Conversely, if K is a right Dedekind domain, then eKn e is a right Dedekind prime ring.

These rings have also been studied by Michler [88], who shows that localisations at prime ideals are classical for bounded Asano prime orders and deduces that these rings are Dedekind right orders. Robson (unpublished) has also shown that for a non-simple right Dedekind prime ring R every proper factor ring is an artinian principal ideal ring (as in the commutative case) and if the Jacobson radical of R is non-zero then R itself is a (non-artinian) principal ideal ring. Once again the case of simple rings evades analysis for the present. The theory of maximal orders in a different sense has also been studied by Auslander and Goldman [11]; this follows a line of development in simple algebras; see Deuring [22]. Here an order over a noetherian domain D, is a subring R of a central simple algebra A over the quotient field K of D such that R is a finitely generated D-module and RK = A. An order R is maximal if R is not properly contained in any other order in A. These orders are always right and left orders in the Asano sense mentioned earlier and they are bounded, because A satisfies a standard polynomial identity, being a finite dimensional simple algebra. The main theorem obtained is as follows.
THEOREM (6.6). Let D be a discrete rank one valuation ring and R be a maximal order in a central simple algebra A over the quotient field K of D. Then

(1) all maximal orders in A are conjugate to R, (2) R is a principal ideal ring, (3) R is a full matrix algebra over a maximal order in a division algebra. A localisation argument then shows that a maximal order R over a dedekind ring D (commutative) is the ring of endomorphisms of a finitely generated projective module over a maximal order in a division algebra. Thus the theory turns around and returns to Theorem (6.2). References
1. S. A. Amitsur, " The identities of Pi-rings ", Proc. Amer. Math. Soc, 4 (1953), 27-34. 2. , " Identities and generators of matrix rings ", Bull. Res. Coun. Israel, 5A (1955), 1-10. 3. , " A generalisation of Hilbert's nullstellensatz ", Proc. Amer. Math. Soc, 8 (1957), 649-656. 4. , " Generalised polynomial identities and pivotal monomials ", Trans. Amer. Math. Soc, 114(1965), 210-226. 5. , " Rational identities and applications to algebra and geometry ", / . Algebra, 3 (1966), 304-359. 6. , " Prime rings having polynomial identities with arbitrary coefficients ", Proc. London Math. Soc. (3), 17 (1967), 470-486. 7. and C. Procesi, " Jacobson-rings and Hilbert algebras with polynomial identities", Annali Mat. IV, 71 (1966), 61-72.

SOME ASPECTS OF RING THEORY

151

8. E. Artin, " Zur arithmetik hyperkomplexer zahlen ", Abh. Math. Sem. Univ. Hamburg, 5 (1927), 261-289. 9. K. Asano, " Arithmetische idealtheorie in nichtkommutativen ringen ", Jap. J. Math., 15 (1939), 1-36. 10. , " Quotient bildung und schiefringe ", / . Math. Soc. Japan, 1 (1949), 73-78. 11. M. Auslander and O. Goldman, " Maximal orders ", Trans. Amer. Math. Soc, 97 (1960), 1-24. 12. and , " The Brauer group of a commutative ring ", Trans. Amer. Math. Soc, 97 (1960), 367-409. 13. E. H. Batho, " Non-commutative semi-local and local rings ", Duke Math. J., 24 (1957), 163-172. 14. N. Bourbaki, Algebra commutative XXVIII (Hermann, Paris). 15. H. Cartan and S. Eilenberg, Homological algebra (Princeton Univ. Press, Princeton, New Jersey, 1956). 16. V. C. Cateforis and F. L. Sandomierski, " The singular submodule splits off", / . Algebra, 10 (1968), 149-165. 17. C. Chevalley, " L'arithmetique dans les algebres de matrices ", Actualitis Sci. Indust, 323 (Paris 1936). 18. P. M. Cohn, " On the embedding of rings in skew fields ", Proc. London Math. Soc. (3), 11 (1961), 511-530. 19. R. R. Colby and E. R. Rutter, " The structure of certain artinian rings with zero singular ideal", / . Algebra, 8 (1968), 156-164. 20. C. W. Curtis, " On additive ideal theory in general rings ", Amer. J. Math., 74 (1952), 687-700. 21. , " Non-commutative extensions of Hilbert rings ", Proc. Amer. Math. Soc, 4 (1953), 945-955. 22. M. Deuring, Algebren (Springer, Berlin, 1935). 23. N. J. Divinsky, Rings and radicals (George Allen and Unwin, London, 1965). 24. J. Dixmier, " Sur les representations unitaires des groupes de Lie nilpotentes I I " , Bull. Soc Math. France, 85 (1957), 325-388. 25. , " Sur l'algebre enveloppante d'une algebre de Lie nilpotente ", Arch. Math, X (1959), 21-26. 26. , " Representations irreductibles des algebres de Lie nilpotents, An. Acad. Brasil. Ci., 35 (1963), 491-519. 27. , " Representations irreductibles des algebres de Lie resolubles ", / . Math. Pures. Appl., 45 (1966), 1-65. 28. M. Djabali, " Anneau de fractions d'un /-anneau ", Canad. J. Math., 20 (1968), 182-202. 29. C. Faith and Y. Utumi, " On noetherian prime rings ", Trans. Amer. Math. Soc, 114 (1965), 53-60. 30. E. H. Feller and E. W. Swokowski, " The ring of endomorphisms of a torsion-free module ", / . London Math. Soc, 39 (1964), 41-42. 31. and , " Prime modules ", Canad. J. Math., 17 (1965), 1041-1052. 32. K. L. Fields, " On the global dimension of skew polynomial rings ", / . Algebra (to appear). 33. P. Gabriel, Des categories abeliennes (These, l'universite de Paris, 1962). 34. and Y. Nouaze, " Ideaux premiers de l'algebre envellopante d'une algebre de Lie nilpotente ", / . Algebra, 6 (1967), 77-99. 35. and R. Rentschler, " Sur la dimension des anneaux et ensembles ordonnes ", C. R. Acad. Sci. Paris, 265 (1967), 712-715. 36. E. R. Gentile, " On rings with one-sided field of quotients ", Proc. Amer. Math. Soc, 11 (1960), 380-384. 37. A. W. Goldie, " The structure of prime rings under ascending chain conditions ", Proc. London Math. Soc, 8 (1958), 589-608.

152

A. W. GOLDIE

38. A. W. Goldie, " Semi-prime rings with maximum condition ", Proc. London Math. Soc, 10 (1960), 201-220. 39. , " Non-commutative principal ideal rings ", Arch. Math., 13 (1962), 214-221. 40. , " Torsion-free modules and rings", / . Algebra, 1 (1964), 268-287. 41. , " Localisation in non-commutative noetherian rings ", / . Algebra, 5 (1967), 89-105. 42. , Lectures on non-commutative noetherian rings (Canadian Math. Congress, York University, Toronto, 1967). 43. , "A note on non-commutative localisation ", / . Algebra, 8 (1968), 41-44. 44. , "A note on prime rings with polynomial identities ", / . London Math. Soc. (to appear). 45. R. Gordon, " Rings faithfully represented on their left socle", /. Algebra, 7 (1967), 303-342. 46. M. Harada, " Hereditary orders ", Trans. Amer. Math. Soc, 107 (1963), 272-290. 47. , " Hereditary semi-primary rings and triangular matrix rings ", Nagoya Math. J., 27 (1966), 463-484. 48. R. Hart, " Endomorphisms of modules over semi-prime rings ", / . Algebra, 4 (1966), 46-51. 49. , "Simple rings with uniform right ideals ", / . London Math. Soc, 42 (1967), 614-617. 50. I. N. Herstein, " A theorem of Levitski", Proc. Amer. Math. Soc, 13 (1962), 213-214. 51. , " Sul teorema di Goldie ", Rend. Accad. Lincei, 35 (1963), 23-26. 52. , " A counter-example in noetherian rings ", Proc. Nat. Acad. Sci., 54 (1965), 1036-1037. 53. , " A theorem on left noetherian rings ", / . Math. Anal. Appi, 15 (1966), 91-96. 54. and L. Small, "Nil rings satisfying certain chain conditions", Canad. J. Math., 16 (1964), 771-776. 55. , " Nil rings " (corrigendum and addendum), Canad. J. Math., 18 (1966), 300-302. 56. Graham Higman, " On a conjecture of Nagata ", Proc. Cambridge Philos. Soc, 52 (1956), 1-4. 57. Y. Hinohari, "Notes on non-commutative semi-local rings", Nagoya Math. J., 17 (1960), 161-6. 58. K. A. Hirsch, " A note on non-commutative polynomials ", / . London Math. Soc, 12 (1937), 264-266. 59. N. Jacobson, " The theory of rings ", Amer. Math. Soc Surveys F/(New York, 1943). 60. , " Structure of rings ", Amer. Math. Soc. Colloq. Pub!., 37 (1956 and 1964). 61. , Lie algebras (Interscience, New York, 1962). 62. A. V. Jategaonkar, " Left principal ideal domains ", / . Algebra, 8 (1968), 148-155. 63. R. E. Johnson, " The extended centraliser of a ring over a module ", Proc. Amer. Math. Soc, 2 (1951), 891-895. 64. , " Prime rings ", Duke Math. J., 18 (1951), 799-809. 65. , " Representations of prime rings ", Trans. Amer. Math. Soc, 74 (1953), 351-357. 66. , " Semi-prime rings ", Trans. Amer. Math. Soc, 76 (1954), 375-388. 67. , " Quotient rings of rings with zero singular ideal ", Pacific J. Math., 11 (1961), 1385-1392. 68. , " Remarks on a paper of Procesi ", / . Algebra, 2 (1965), 38-41. 69. I. Kaplansky, " Rings with a polynomial identity ", Bull. Amer. Math. Soc, 54 (1948), 575-580. 70. , Commutative rings (Queen Mary College, Math Notes, London, 1966). 71. Kwangil Koh, " On the class of rings which do not contain non-zero singular ideals ", Amer. Math. Monthly, 72 (1965), 875-877. 72. and A. C. Mewborn, " A class of prime rings ", Canad. Math. Bull., 9 (1966), 63-72. 73. J. Lambek, " On the ring of quotients of a noetherian ring " Canad. Math. Bull., 8 (1965), 279-290. 74. , Lectures on rings and modules (Blaisdell, 1966). 75. L. Lesieur and R. Croisot, "Theorie noetherienne des anneaux, des demi-groupes et des modules dans le cas non commutatif II ", Math. Ann., 134 (1958), 458-476. 76. , " Sur les anneaux premiers noetheriens a gauche ", Ann. Sci. Ecole Norm. Sup., 76 (1959), 161-183.

SOME ASPECTS OF RING THEORY

153

77. L. Lesieur and R. Croisot, "Algebre noetherienne non-commutatif ", Mem. Sci. Math. CLIV (Gauthiev Villars, Paris, 1963). 78. J. Levitski, " A theorem on polynomial identities ", Proc. Amer. Math. Soc, 1 (1950), 334-341. 79. and S. A. Amitsur, " Minimal identities for algebras ", Proc. Amer. Math. Soc, 1 (1950), 449-463. 80. L. Levy, " Torsion-free and divisible modules over non-integral domains ", Canad. J. Math.* 15(1963), 132-151. 81. D. E. Littlewood, " On the classification of algebras ", Proc. London Math. Soc. (2), 35 (1933), 200-240. 82. J. C. McConnell, " The intersection theorem for a class of non-commutative rings ", Proc. London Math. Soc. (3), 17 (1967), 487-498. 83. , " Localisation in enveloping rings ", / . London Math. Soc, 43 (1968), 421-428. 84. , " The noetherian property in complete rings and modules ", / . Algebra (to appear). 85. A. Malcev, "On the immersion of an algebraic ring into a field", Math. Ann., 113 (1936), 686-691. 86. G. Michler, " Charakterisierung einer klasse von noetherschen ringen ", Math. Z., 100 (1967), 163-182. 87. , " Halberbliche fastlokale ordnungen in einfach artin ringen ", Arch. Math, (to appear). 88. , " Maximal Asano orders ", Proc. London Math. Soc. (to appear). 89. D. C. Murdoch, " Subrings of the maximal ring of quotients associated with closure operations ", Canad. J. Math., 15 (1963), 723-743. 90. M. Nagata, " On the nilpotency of nil algebras ", / . Math. Soc. Japan, 4 (1952), 296-301. 91. B. H. Neumann, " On ordered division rings ", Trans. Amer. Math. Soc, 66 (1949), 202-252. 92. O. Ore, " Linear equations in non-commutative fields ", Ann. of Math., 32 (1931), 463-477. 93. , " Theory of non-commutative polynomials ", Ann. of Math., 34 (1933), 480-508. 94. E. Posner, " Prime rings satisfying a polynomial identity ", Proc. Amer. Math. Soc, 11 (1960), 180-184. 95. C. Procesi, " Sopra un teorema di Goldie riguardante la struttura degli anelli primi con condizione di massimo ", Rend. Accad. LinceL, 34 (1963), 372-377. 96. , " On a theorem of Goldie ", /. Algebra, 2 (1965), 80-84. 97. , " Non-commutative affine rings ", Memorie Accad. Naz. Rend. Cl. Sci. Fis. Mat. Nat. 8 Fasc, 6 (1967), 239-255. 98. , " Non-commutative Jacobson rings ", Ann. Scuola Norm. Sup. Pisa, 21 II (1967), 381-390. 99. and L. Small, " On a theorem of Goldie ", / . Algebra, 2 (1965), 80-84. 100. D. G. Quillen, " On the endomorphism ring of a simple module over an enveloping algebra ". (Journal not known.) 101. J. A. Riley, " Axiomatic primary and tertiary decomposition theory ", Trans. Amer. Math. Soc, 105(1962), 177-201. 102. , " Reflexive ideals in maximal orders ", / . Algebra, 2 (1965), 451-465. 103. J. C. Robson, " Do simple rings have unity elements? ", /. Algebra, 6 (1967), 444-447. 104. , " Non-commutative dedekind rings ", /. Algebra, 9 (1968), 249-265. 105. A. Roy, " A note on filtered rings ", Arch. Math., (1965), 421-427. 106. L. W. Small, " An example in noetherian rings ", Proc. Nat. Acad. Sci., 54 (1965), 1035-1036. 107. , "On some questions in noetherian rings ", Bull. Amer. Math. Soc, 72 (1966), 853-857. 108. , " Orders in artinian rings ", / . Algebra, 4 (1966), 13-41. 109. , " Orders in artinian rings " (correction and addition), / . Algebra, 4 (1966), 505-507. 110. , " Orders in artinian rings II ", / . Algebra, 9 (1968), 266-273. 111. T. D. Talintyre, " Quotient rings of rings with maximum condition on right ideals ", /. London Math. Soc, 38 (1963), 439-450.

154

SOME ASPECTS OF RING THEORY

112. D. Tamari, " On the embedding of Birkhoff-Witt rings in quotient fields", Proc. Amer. Math. Soc, 4 (1953), 197-202. 113. Y. Utumi, " On quotient rings ", Osaka Math. J., 8 (1956), 1-8. 114. , "A theorem of Levitski ", Amer. Math. Monthly, 70 (1963), 286. 115. , " On prime I-rings with uniform one-sided ideals ", Amer. J. Math., 85 (1963), 583-596. 116. B. L. van der Waerden, Moderne algebra I, //(Berlin, 1930). 117. J. H. M. Wedderburn, " Non-commutative domains of integrity ", /. Reine Angew. Math., 167 (1932), 129-141. 118. J. M. Zelmanowitz, " Endomorphism rings of torsionless modules ", / . Algebra, 5 (1967), 325-341.

University of British Columbia and Leeds University.