Minkowski [1] first proved that the surface of a convex body in E3 can be approximated by a level surface of a convex analytic function. His proof is strikingly simple. His proof is also presented for En by Bonnesen-Fenchel [2, pp. 10–12[. We here prove that the same kind of result is achieved using level surfaces of convex non-negative polynomials. We give two types of approximation, one based on finite sums as Minkowski did, and the other using integration. Since these approximations may be used for other applications we also extend them and give special formulae when the surface is centrally symmetric.

The theorem proved in this paper is basic for a general intersection theory applicable to multigraded polynomial rings. When the polynomial ring is graded by the non-negative integers the facts, in one form or another, are well known, but on passing to more general gradings fresh complications appear. These are not wholly trivial and, as the author was unable to find an account of these matters in the literature, it may be of interest if one is given here.

The first and second boundary value problems of plane elastostatics are solved for the interior of a parabola. A conformal transformation is used to map the interior of the parabola onto an infinite strip. An analytic continuation technique reduces the boundary value problem to the solution of a form of differential-difference equation. This is solved by a Fourier integral method. The resulting integrals are evaluated by residues to give eigenfunction expansions for the complex potentials.

Erdös, Kestelman and Rogers [1[ showed that, if A1, A2,… is any sequence of Lebesgue measurable subsets of the unit interval [0, 1] each of Lebesgue measure at least η > 0, then there is a subsequence {Ani} (i = 1, 2,…) such that the intersection contains a perfect subset (and is therefore of power ). They asked for what Hausdorff measure functions φ(t) is it possible to choose the subsequence to make the intersection set ∩Ani of positive φ-measure. In the present note we show that the strongest possible result in this direction is true. This is given by the following theorem.

Based on the method of analytic continuation of Buchwald and Davies [1[, [2], the first boundary value problem of an elliptic plate is solved. The results are in agreement with the more complicated solution of Muskhelishvili [3]. The solution of the second boundary value problem is also obtained.

The following problem was proposed by Professor N. J. Fine: to prove that there do not exist rational functions F1, F2, F3 of x1, x2, x3, x4, with real coefficients, such that

identically. In the present note I give a proof.†

Let Q denote the closed unit cube in Rn. The elementary area A(f) of a Lipschitz function f on Q is given by the formula.

We consider the following problem. Calls arrive at a telephone exchange at the instants t0, t1, … tm …. The telephone exchange contains a denumerable infinity of channels. The holding times of calls are non-negative random variables distributed independently of the times at which calls arrive, independently of which channel a call engages and independently of each other with a common distribution function B(x). Takacs [3Τm = tm+1 − tm, m ≧ are identically and independently distributed non-negative random variables with common distribution function, A (x). Finch [1] has studied the transient behaviour in the case of a recurrent arrival process and exponential holding time, that is when the common distribution of holding time is given by In this paper we make no assumption about the arrival process {tm}. The underlying principle of this paper is the same as that of Finch [2]. We consider the instants of arrival t0, t1,…, tm,… as given and determine various probabilities of interest conditionally as functions of the inter-arrival intervals Τ1,…, Τm,… When the arrival process is a stochastic process we can then determine the relevant unconditional probabilities by integration.

Let L1 denote temporarily the usual Lebesgue space over the circle group (equivalently: the additive group of real numbers modulo 2π), and let H1 denote the Hardy space comprised of those f in L1 whose complex Fourier coefficients vanish for all negative frequencies, so tha D. J. Newman has settled a conjecture by showing [1] that there exists no continuous projuction of L1 onto H1, i.e. that there exists in L1 no topological complement to H1.

One proof of Minkowski's fundamental theorem on lattice points in an n-dimensional parallelopiped depends upon a multiple Fourier expansion. Its terms involve multiple integrals which are easily evaluated and so are shown to have positive values. Then the expansion takes only positive values and the desired proof follows at once. It seems of interest to note a multiple integral which assumes only positive values.

We consider a single server queue for which the interarrival times are identically and independently distributed with distribution function A(x) and whose service times are distributed independently of each other and of the interarrival times with distribution function B(x) = 1 − e−x, x ≧ 0. We suppose that the system starts from emptiness and use the results of P. D. Finch [2] to derive an explicit expression for qnj, the probability that the (n + 1)th arrival finds more than j customers in the system. The special cases M/M/1 and D/M/1 are considerend and it is shown in the general case that qnj is a partial sum of the usual Lagrange series for the limiting probability .

A queue at which arrivals occur randomly in batches of fixed size r and for which the service times are independent negative exponential variates is considered. Expressions are obtained for the moments of the transient waiting time distribution and the distribution of the number of customers in the system just before the nth batch arrives. The distribution of the number of customers served in a busy period is also determined.

Huta [1], [2] has given two processes for solving a first order differential equation to sixth order accuracy. His methods are each eight stage Runge-Kutta processes and differ mainly in that the later process has simpler coefficients occurring in it.

If G is a finite linear group of degree n, that is, a finite group of automorphisms of an n-dimensional complex vector space (or, equivalently, a finite group of non-singular matrices of order n with complex coefficients), I shall say that G is a quasi-permutation group if the trace of every element of G is a non-negative rational integer. The reason for this terminology is that, if G is a permutation group of degree n, its elements, considered as acting on the elements of a basis of an n-dimensional complex vector space V, induce automorphisms of V forming a group isomorphic to G. The trace of the automorphism corresponding to an element x of G is equal to the number of letters left fixed by x, and so is a non-negative integer. Thus, a permutation group of degree n has a representation as a quasi-permutation group of degree n.

We consider a single server queueing system in which customers arrive at the instants t0, t1, …, tm, …. We write τm = tm+1 − tm, m ≧ 0. There is a single server with distribution of service times B(x) given by where k is an integer not less than unity.

Let G be an n-generator group and let g = (g1, g2, …, gn) be an ordered set of n elements which generate G, then g is called a generating n-vector of G. Let denote the set of all generating n-vectors of G.

Throughout this paper E, F and G denote separated locally convex spaces, F C G, the injection i: F → G being continuous (i.e. the topology on F is finer than that induced on it by the topology on G). E′, F′ and G′ denote the respective duals of E, F and G. i′ is the adjoint map of G′ into F', which is defined by restricting linear forms on G to F C G.