In a highly interesting critical account of the mathematical work of James Gregory (1638–1675), written for the Proceedings of the Edinburgh Mathematical Society, (1) 41 (1923), 2–25, by the late Professor G. A. Gibson, there occurs at p. 8 something of a mathematical puzzle. On that page a pair of formulae are quoted, which certainly are striking examples of the analytical power of Gregory, and which run as follows:—

Gibson adds that “there is another formula (Rigaud, p. 207), but it is of a very complicated character and I do not reproduce it.” It will be convenient to refer to the above pair as formulae A′ and B′, and to the more complicated but analogous series as formula C, and to the original series, from which the above were transcribed, as formulse A and B. I am indebted to Mr A. Inglis for drawing my attention to the problem.

The object of this paper is to introduce the differential operator, ▽, generalised for a Riemannian space Vn immersed in a flat space Vp, and then to discuss the general small deformation of Vn.

Formulae of interpolation in terms of given central differences might be regarded as falling into two groups, A and B. In group A, the simplest cases are those in which each given difference is one of the two which in the difference table lie nearest to the preceding given difference; the differences are all natural differences (i.e., are not mean differences), and are all expressed in the centraldifference notation. Any such formula can be a central-difference formula for a certain range of the variable: but that is a matter with which we are only incidentally concerned. What I have to do is to examine the formula as determined by the series of differences given. I have then to see how the formula is affected when an ordinary difference is replaced by a mean difference. This brings us to group B, which comprises two formulae only: the Newton-Stirling formula, which expresses the required quantity in terms of a tabulated value and its central differences; and the Newton-Bessel formula, which expresses it in terms of the mean of two tabulated values and the central differences of this mean.

In this section we extend the definition of an E-set, so that it includes sets of the type

where the only restriction on the Ei is that they be non-singular. We now consider matrices of the type

where each ei takes independently the values 0, 1, …, n − 1, while the a(ei) are either complex numbers or else matrices of order r, the product a(ei) E(ei), in the latter case, being interpreted as the direct product of the two matrices a(ei) and E(ei).

§ 1. In space of three dimensions the properties of self-conjugate tetrads, pentads and hexads with regard to a quadric are well known (see Baker's Principles of Geometry, vol. iii). The general theorem in space of n dimensions Sn is to establish the existence of a set of n + p + 1 points A0, A1, …, An+p (0 ≤ p ≤ n−1) such that the pole, with respect to a given quadric, of the (n−1)-flat determined by any set of n of the points lies in the p-flat determined by the remaining p + 1 points.

Absolute summability according to Cesàro's method has been defined by Fekete for positive integral orders, as follows:—

Denoting the rth partial sum of a series Σun by and its rth mean, namely2, by we can regard as the sum of the series

.

A period of a function f(z) is defined to be a number w(≠0) such that

is identically zero; and it is not a difficult matter to show that an integral function may either have no periods or else a single sequence kλ, k = ±1, ±2, ….

The object of this paper is to discuss conditions for a maximum or minimum of functions of integrals of the type

employing the methods of the Calculus of Variations.

When a given frequency distribution is to be graduated, it is customary to express the constants of the fitted curve in terms of the moments of the frequency distribution. The rth moment of the distribution, in which the relative frequency of a measure x is φ(x) or, in the case of a continuous variable, the differential of frequency is φ(x)dx, is defined in the respective cases by

the summation or integration being taken over the whole range of possible values of x. In the present paper we make use of another kind of moment, the factorial moment, which has already been considered by several writers, and which is specially suited to the case when the frequencies of the distribution are given for discrete, equidistant values of the variable. The (r + 1)th factorial moment, for the case where x, measured from some arbitrary origin, can increase by increments h, 2h, 3h,.…, will be defined to be

where the summation extends over all possible values of x; it will be denoted by m(r+1). By a suitable choice of scale the increments of x may be taken as equal to unity in any given case.

A common method of solving a linear differential equation consists in expressing the differential operator as a product of factors. The possibility of doing so has been studied extensively by Vessiot, following the work of Picard and Drach, on the lines of the Galois theory of algebraic equations. The analogous process of resolving a, linear differential system, consisting of an equation together with boundary conditions, into two or more systems of lower order does not seem to have been investigated. Such a resolution is not always possible, even in cases where the differential equation can be factorised. Thus the system

is equivalent to the systems

but the system

cannot be so resolved.

Much has been written, from the algebraical as well as from the geometrical standpoint, on the subject of pencils of quadrics: algebraically the problem consists of the canonical reduction of a pencil of quadratic forms, and the classical paper on the subject is by Weierstrass. But among the different kinds of pencils of quadratic forms there is the “singular pencil,” in which the discriminant of every form belonging to the pencil is zero; interpreted geometrically this means that every quadric belonging to the pencil is a cone. This case was expressly excluded from consideration by Weierstrass, and the canonical reduction was only accomplished later by Kronecker. But, although Weierstrass and Kronecker together solved completely the problem of the canonical reduction of a pencil of quadratic forms, a much clearer insight into the nature of the problem was gained when Segre gave the geometrical solution. He published two papers, one dealing with the non-singular pencils and the other with the singular pencil.

In a paper entitled “Sets of anticommuting matrices” Eddington proved that if El, E2, …., Eqform a set of q four-rowed square matrices satisfying the relations,

,

where E is the unit matrix, then the maximum value of q is five. Later Newman showed that this result is a particular case of the general theorem that ifE1, E2, …., Eqform a set of q t-rowed square matrices satisfying (1), where t = 2Pτ and τ is odd, then the maximum value of q is 2p + 1.

§1. The coordinates considered are linear, i.e. in a plane the equation of a straight line, and in space the equation of a plane, is linear in the coordinates. We shall first consider point-coordinates in plane geometry, taking elliptic geometry as typical, with space-constant unity.

§1. Introduction. Care is needed in dealing with determinants whose elements are subject to experimental error, particularly when a determinant itself is small compared with its first minors. For, as these examples show, a relatively tiny error in one element may be responsible for a large error in the determinant

The present paper describes briefly a notation for representing continued fractions in many dimensions, which has the advantage providing a direct method of attack and of rendering intuitive, results which are usually proved by induction. The notation is the outcome of a generalisation which I previously made [1] in connection with the solution of certain difference equations. Only formal theorems are considered here. For a discussion of convergence reference may be made to the works [2, 3, 4, 5] cited at the end. The paper by Paley and XJrsell is particularly important since these authors discuss very fully the non-cyclic simple continued fraction

In a recent series of papers, Straneo has introduced into Riemannian space-time a teleparallelism defined by an asymmetric connection, and in this manner attempted to develop a unified theory of gravitation and electricity. The essential idea underlying his work is, therefore, similar to that of the Einstein-Mayer theory of 1929–31.

§ 1. Introduction. Little is known concerning the theory of resultants of equations other than in the complex number system. The cyclic number systems provide a simple example which is not a division algebra. In such a system with n units er. any number y ≡ y0 + y1e1 + y2e2 + … + yn−1en−1 has coefficients yr drawn from a field, and the units satisfy the product law:

In a very remarkable work on the operational Calculus, Dr Balth. van der Pol has introduced a new function, playing with respect to Bessel function of order zero the same part as the cosine- or sine-integral with respect to the ordinary cosine or sine. He showed that this function—which he called Bessel-integral junction—can be used to express the differential coefficient of any Bessel function with respect to its index. But he did not investigate the further properties of his new function. I propose to give here some of them, which appear to be interesting, and to introduce and study the functions connected, in the same way, with Bessel functions of any order.

One of the many interesting problems discussed by Ramanujan is concerned with the effect of truncating at its maximum term nn/n! the exponential series for en, where n is a positive integer. When n is large, the sum of the first n terms is, roughly speaking, half the sum of the whole series.

§ 1. Introduction. The algebra of quantum mechanics is characterized by the fact that the variables p, q obey all the laws of ordinary algebra except that multiplication is non-commutative and instead there exists a relation of the form

where c is a real or complex scalar constant and is thus commutative with both p and q.