In this paper we describe a way of representing varieties of algebras by algebras. That is, to each variety of algebras we assign an algebra of a certain type, such that two varieties are rationallv equivalent if and only if the assigned algebras are isomorphic.

The BFC number of a group G is defined to be the least upper bound n of the cardinals of the conjugacy classes of G, provided this is finite, and we then say that G is n-BFC. It was shown by B. H. Neumann [2] that the derived group G′ of such a group is finite, and J. Wiegold [5] proved that.This bound was sharpened by I. D. Macdonald [1] to, and P. M. Neumann has recently communicated the (unpublished) result that G′ ≦ nq(n) with q(n) a quadratic in log2w, an immense improvement on the above. J. A. H. Shepperd and J. Wiegold [4] improved the bound in two special cases, showing that if G is soluble, G′ ≦ np(n) with p(n) a quintic in Iog2n, and that if G is nilpotent of class 2, , It is conjectured that for any n-BFC group G, , Wiegold [5] having shown that this bound is attained by certain nilpotent groups of class 2.

The theme of this paper is a conjecture whose origins go back to work of B. H. Neumann ([5], Theorem 3·1) and his former students I. Wiegold and I. D. Macdonald ([7], [4]). B. H. Neumann proved that if there is a bound to the sizes of conjugacy classes in the group G. that is. if G is a BFC group. then the derived group G′ is finite; Wiegold. and later Macdonald. produced explicit upper bounds for G′ in terms of the maximum n of the sizes of the conjugacy classes in G. The coniecture is that where l(n) should be the arithmetic function1 λ(n) for a best possible bound, or l(n) may be interpreted as log2n for a smooth, monotonicorderof-magnitude estimate. However, the bounds produced in [7] and [4], and even the vastly better upper bounds proved bv Sheooerd and Wieeold [6] for soluble groups, are very much bigger than , and it is mv aim in this paper to take a first step towards closing the gap. This first, step is study of BFC p-groups: in a sequel I hope to show how the results proved here can be used to obtain improved bounds for arbitrary BFC groups.

Let ℜc denote the variety of all nilpotent groups of class ≦ c, that is, ℜc is the class of all groups satisfying the law, where we define, as usual, and, inductively, . Further, let Fk(ℜc) denote a free group of ℜe of rank k. In her book Hanna Neumann ([4], Problem 14) poses the following problem: Determine d(c), the least k such that Fk(ℜc) generates ℜc. Further, she suggests, incorrectly, that d(c) = [c/2] + l. However, as we shall prove here, the correct answer is d(c) = c—1, for c ≦ 3. 2 More generally, we shall prove the following result.

We refer the reader to Hanna Neumann [7] for notation and other undefined terms. Let and denote the varieties of groups defined by the laws (xy)n=xnyn, [x, y]n=1 and [x, yn]respectively, where n is an integer. (n)-groups were termed “n-abelian” by R. Baer [1] and have been a subject matter of investigation by various authors (see [3], [5], [6] and the references therein). Recently KaluŽnin [5] has shown that , thus clarifying the relationship between U(n) and the familiar varieties. From the elementary inequalities(n ≠ 0,1)it is easily deduced that(see for instance [5]). If G = Cm Wr C∞, then cleary but for any Thus and . It is also easy to see that in general (see for instance [6] § 5.1) and we are led to ask

Given a group G and a partial automorphism μ of G, i.e. an isomorphism mapping a subgroup A of G onto another subgroup B of G, then it is known [3] that μ can always be extended to a total automorphism, in fact an inner one, of a supergroup of G; that is there exists a group G* ⊇ G with an inner automorphism μ* whose effect on the elements of A is the same as that of μ. Also any number of partial automorphisms μσ, where a ranges over some index set Σ can be simultaneously extended to inner automorphisms of one and the same group [3, Theorem II].

G(n;l) will denote a graph of n vertices and l edges. Let f0(n, k) be the smallest integer such that there is a G (n;f0(n, k)) in which for every set of k vertices there is a vertex joined to each of these. Thus for example fo = 3 since in a triangle each pair of vertices is joined to a third. It can readily be checked that fo = 5 (the extremal graph consists of a complete 4-gon with one edge removed). In general we will prove: Let n > k, andthen f0(n, k) = f(n, k).

The quadrature formula of order m using values of derivatives up to the m — 1st order with the best possible bound in is derived. Using certain properties of the polynomials minimal in Lq norm, it is proved that the optimal formula not use the derivatives of m — 1st order if m is even.

Let S be a poset with a greatest element 1. We denote order in S by ‘≦’ and, whenever they exist in S, l.u.b and g.l.b by ‘∨’ and ‘∧’ respectively. An orthocomplementation of S is a bijection w: S → S such that x ∨ xω exists for each x in S and (i) xωω = x, (ii) x ≦ y implies yω ≦ xω and (iii) x ∨xω = 1. If a poset S admits an orthocomplementation ω we call the pair (S, ω) an orthoposet.

It is the purpose of this note to illustrate by example the results obtained in Finch [1]. All our definitions and notation are as in that paper, and we use them without reference. In his paper, Finch considered orthoposets and, in particular, orthomodular posets that had either the A-property or the B-property.

The object of this note is to exhibit a certain class of distribution spaces as being c-admissible in the sense of [1] and [2]. Throughout the terminology and notation are the same as in [1] and [2]. The only addition to this is that if x ∈ Rn then τx will denote the translation operator which carries each distribution u onto the distribution ux.

In the list of known finite non-abelian simple groups there are infinitely many pairs of non-isomorphic simple groups which have the same order. The smallest known example of two such groups are the simple groups A8 and PSL(3, 4), of order 20,160.

In a previous communication [5] the author has dealt with the congruence properties of some restricted partition functions. The general category of such functions may be denoted.

Let {Xn: n ≧ 1} be a sequence of independent random variables and write Letand let . Suppose that converges in law to the standard normal distribution (see [5, 280] for necessary and sufficient conditions). Let {xn} be a monotonic sequence of positive real numbers such that xn → ∞ as n → ∞. Then as n → ∞ for all ε > 0. [6] Rubin and Sethuraman call probabilities of the form probabilities of moderate deviations and obtain asymptotic forms for such probabilities under appropriate moment conditions.

The methods developed up to now in the theory of plates do not lend themselves easily to the solution of problems in the case of plates of arbitrary shape. Such methods lend themselves only to the solution of problems for plates of certain special simple shapes. Moreover, because of the complicated mathematical tools involved in these methods, the methods are not suitable for engineers. Hence there arises the need for the formulation of an approximate theory which may appear quite suitable for attacking problems on plates of arbitrary shape. In the present paper the author develops a new method which aims at the solution of the problem quite independently of the shape of the plate.

In [2] Evans and Neumann raised the question of whether diassociativity in groupoids or loops is equivalent to a finite set of identities and in [3], Neumann still lists the problem as unsolved. It is the purpose of this paper to show that the answer to the question for groupoids is negative.

The necessity of accurate numerical approximations to the solutions of differential equations governing physical systems has always been an important problem with scientists and engineers. Hammer and Hollingsworth [11] have used Gaussian quadrature for solving the linear second order differential equations. This method has been further developed by Morrison and Stoller [3], Korganoff [1], Kuntzman [9], Henrici [12] and Day [7, 8]. Quadrature methods based upon Lobatto quadrature formulae have recently been considered by Day [6, 8] and Jain and Sharma [10] and seem to give better results.