An ancipital form is a form [a, b, c] in which b= 0 or b= a; these fall into pairs of associates: [a, 0, c] and [c, 0, a] (type 1), and [a, a, c] and [4c–a, 4c–a, c] (type 2). The set of discriminantal divisors of discriminant d is formed by choosing, from each pair of primitive associate ancipital forms of discriminant d, exactly one of the two leading coefficients. In this article we study representations of discriminantal divisors of a given discriminant by binary quadratic forms of that discriminant, previously studied by the author and by G. Pall. We are concerned here with discriminants d= 4kpq, where k ≥ 1, p = 1, q = 3 (mod 4) are primes, and d = 4kp, where k ≥ 1 and p is an odd prime. This investigation arose in connection with the search for integral solutions of x2 –Dy2 = – 1.

In this paper we use the Bourbaki conventions for rings and modules: all rings are associative but not necessarily commutative and have an identity element; all modules are unital.

Numerous formulae have been given which exhibit the asymptotic behaviour as t → ∞solutions of

where F(t) is essentially positive and Several of these results have been unified by a theorem of F. V. Atkinson [1]. It is the purpose of this paper to establish results, analogous to the theorem of Atkinson, for the third order equation

and for the fourth order equation

We give a simple proof of the lemma of Ford [1] on the existence of self-adjoint square roots in Banach*-algebras in which continuity of the involution is not assumed.

Bellman [1], [2, p. 116] proved that, if all solutions of the equation

are in L2, ∞) and b(t) is bounded, then all solutions of

are also in L2(a, ∞). The purpose of this paper is to present conditions on the function f that guarantee that all solutions of

be in the class L2(a, ∞) whenever all solutions of the equation

have this property. It is assumed that r(t) >0, r and qare continuous on a half line (a, ∞) and f is continuous. Actually the continuity assumptions may be weakened to local integrability and L2 (a, ∞) may be replaced by Lp(a, ∞) for any p > 1.

If p and q are cardinal numbers and E is a topological space, then the following property may or may not hold:

Every cover of Eby fewer than q open sets has a subcover by fewer than pof them. (1)

Clearly these properties, for various numbers p and q, are far from being independent; in this paper, we investigate some of the interrelationships between them.

Djabali [1] has proved that, if R is a right and left noetherian ring with an identity and if the proper prime ideals of R are maximal, then R has a right and left artinian two-sided quotient ring. Robson [5, Theorem 2.11] and Small [6, Theorem 2.13] have proved independently that, if Ris a commutative noetherian ring, then Rhas an artinian quotient ring if and only if the prime ideals of Rthat belong to the zero ideal are all minimal. We shall generalise these results by proving the

Theorem. Let R be a right and left noetherian ring with a regular element. Then R has a right and left artinian two-sided quotient ring if and only if each prime ideal of R consisting of zero–divisors is minimal.

Let Rbe an associative ring, Jan infinite index set, and Rj the ring of all J × J row-finite matrices over R. The Jacobson radical of Rwill be denoted by Γ(R).

There are two well-known methods to build up algebras from given algebras, the direct and inverse limits, and a systematic account of these constructions may be found in [2]. It is known that every algebra can be represented as a direct limit of finitely generated algebras although in some cases the representation is trivial. Furthermore, Haimo [3] has established a certain inverse limit representation for the class of all infinite Boolean algebras which generalises, in actual fact, to the class of all infinite lattices with 1. The purpose of this note is to exhibit a certain nontrivial inverse limit representation which is peculiar to the class of infinite, complete Boolean algebras.

Let T = (tmn) be a regular matrix, and CTbe its bounded convergence field. Necessary and sufficient conditions for CT to contain the space of almost convergent sequences are well known. (See, e.g., [7, p.62]). G. M. Petersen has suggested as a problem for research the discovery of necessary and sufficient conditions for the reverse inclusion: When is CT contained in the space of almost convergent sequences? [7, p. 137, research problem 9]. In this paper we deal with this question in a more general context. First we need some notation.

It is sometimes possible to reconstruct semigroups from some of their homomorphic images. Some recent examples have been the construction of bisimple inverse semigroups from fundamental bisimple inverse semigroups [9], and the construction of generalized inverse semigroups from inverse semigroups [12].

To find a “ description of the structure of bands which is complete modulo semilattices ” (from page 25 of [1]) seems to be a very difficult problem. As far as the author is aware, the only class of bands (except for rectangular bands) for which this problem has been solved (see [4] and [3]) is the class of all bands satisfying a generalization of commutativity, namely the condition that efgh = egfh for all elements e, f, g and h.

Using the theory of group extensions, A. Leutbecher [1] proved this

Lemma. Let G bea group, and w some 2-cocycle of a trivial G-module M. The cohomology class ofw will contain symmetric cocycles if and only if w is semisymmetric.

Here we have called w symmetric or semisymmetric according as w(h, g) = w(g, h) for all g, h ∈G or only for those with hg = gh. In one direction, the proof reduces to observing that 2-coboundaries of trivial G-modules are semisymmetric. The nontrivial part of the lemma also admits of a straightforward proof, as follows.