Let G be a compact connected Lie group with H a connected subgroup of maximal rank. Suppose there exists a compact connected Lie subgroup K with H ⊂ K ⊂ G. Then there exists a smooth fiber bundle G/H → G/K with K/H as the fiber. (See for example [13].) This can be incorporated into a diagram involving the classifying spaces as follows:

(1)

Here π, ϕ, ϕ1 and ϕ2 denote fibrations. We also know that the homogeneous spaces and the Lie groups, which are homotopy equivalent to the loop spaces of their respective classifying spaces, are homotopy equivalent to connected finite complexes.

Now suppose H is a maximal subgroup. Can there still exist spaces, which we will call BK, K/H, and G/K, and fibrations so that diagram (1) is still valid? This paper will show that in many cases either G/K or K/H will be homo topically trivial.

Throughout this article c0, c1, c2, … will denote effectively computable positive absolute constants. Denote the cardinality of a set X by |X|. Let N be a positive integer and let A and B be non-empty subsets of {1, …,N}. Put

In [3], Balog and Sá;rközy proved that if N > c0 and

(1)

then there exist a0 and b0 with a0 ∊ A0 and b0 ∊ B0 and a prime number p such that

and

(2)

If follows from this result that if |A| ≫ N and |B| ≫ N then there exist a in A and b in B and a prime p such that p2|(a + b) with

Le but de ce travail est de définir un certain nombre d'objets de Fq((1/t) et de fonctions

qui sont analogues à des objets classiques de R(ou C) et à des fonctions classiques R → R (ou C → C) : par exemple ζ(l), ζ(2), etc., fractions continues, domaine fondamental du “demi-plan” de Poincaré, séries d'Eisenstein, fonctions multiplement périodiques.

Un tel travail avait été entrepris autrefois par E. Artin [2], mais dans un autre esprit, et la fonction zéta que j'ai définie dans [7] n'est pas celle d'Artin, mais celle de Carlitz [3].

L'intersection de la thèse d'Artin et de ce travail n'est donc pas vide, mais elle n'est pas non plus très étendue bien que je me sois limité aux définitions et aux propriétés les plus élémentaires. En revanche, par son esprit, ce travail se rapproche davantage du point de vue de Carlitz.

In [14] we began a study of C*-algebras corresponding to projective representations of the discrete Heisenberg group, and classified these C*-algebras up to *-isomorphism. In this sequel to [14] we continue the study of these so-called Heisenberg C*-algebras, first concentrating our study on the strong Morita equivalence classes of these C*-algebras. We recall from [14] that a Heisenberg C*-algebra is said to be of class i, i ∊ {1, 2, 3}, if the range of any normalized trace on its K0 group has rank i as a subgroup of R; results of Curto, Muhly, and Williams [7] on strong Morita equivalence for crossed products along with the methods of [21] and [14] enable us to construct certain strong Morita equivalence bimodules for Heisenberg C*-algebras.

In [5] C. Akemann and G. Pedersen defined four concepts of semicontinuity for elements of A**, the enveloping W*-algebra of a C*-algebra A. For three of these the associated classes of lower semicontinuous elements are , and (notation explained in Section 2), and we will call these the classes of strongly lsc, middle lsc, and weakly lsc elements, respectively. There are three corresponding concepts of continuity: The strongly continuous elements are the elements of A itself, the middle continuous elements are the multipliers of A, and the weakly continuous elements are the quasi-multipliers of A. It is natural to ask the following questions, each of which is three-fold.

(Q1) Is every lsc element the limit of a monotone increasing net of continuous elements?

(Q2) Is every positive lsc element the limit of an increasing net of positive continuous elements?

(Q3) If h ≧ k, where h is lsc and k is usc, does there exist a continuous x such that h ≧ x ≧ k?

The transformations we will discuss in this paper are the Hankel transformation Hυ defined for f ∊ C0, the collection of continuous functions compactly supported in (0, ∞), by

(1.1)

and the and transformations defined for such f by

(1.2)

and

(1.3)

where Jv >and Yv are the Bessel functions of the first and second kinds respectively, and Hv is the Struve function; for the theory of these functions see [1, Chapter VII].

These transformations were studied extensively by one of us in [5] and [6] on the spaces defined in [7; Sections 1 & 5]. In those papers the boundedness of the three transformations was fully given on the spaces for 1 < p < ∞, but not for p = 1. Also inversion formulae were given for the transformations only for portions of their respective ranges of boundedness.

Let Bτ denote the class of entire functions of exponential type τ (>0) bounded on the real axis. For the function f ∊ Bτ we have the interpolation formula [1, p. 143]

1.1

where t, γ are real numbers and is the so called conjugate function of f. Let us put

1.2

The function Gγ,f is a periodic function of α, with period 2. For t = 0 (the general case is obtained by translation) the righthand member of (1) is 2τGγ,f (1). In the following paper we suppose that f satisfies an additional hypothesis of the form f(x) = O(|x|-ε), for some ε > 0, as x → ±∞ and we give an integral representation of Gγ,f(α) which is valid for 0 ≦ α ≦ 2.