I. In this note, I consider an inequality bearing a formal resemblance to that of HSlder, and I derive from it new conditions for the existence of a Stieltjes integral, and for passage to the limit under the integral sign. The conditions for limits under the integral sign differ from any previously known, in that, for the first time, absolute integrability is not required. They throw some light on problems of convergence of Fourier series.The first proof of the inequality is due to ~I ~ E. R. Love, who studied it at my suggestion. In a joint paper, elsewhere, we propose to consider further questions connected with it.2. We begin with a simple lemma.If as,... , an and b~ .... , bn are two ordered sets of n complex numbers, and p, q > o, then there is an index k (o < k --< n), such that L.C. Young. separating consecutive terms of a finite sequence a=(a,,..., an) may be termed a partition P. The result of the operation is a finite sequence Pa =-x = (xl, ..., xm) in which each x~ is a corresponding sum of at, and, of course, m ~< n. And if dO(a, b) is a function of a pair of sequences a---~ (al, ..., a~) and b ~ (bl,..., b~) the expression dO(Pa, Pb) may be said to be derived from dO(a, b) by the partition P.It is with the expressions thus derived by partition from the product that we shall be concerned.
A.5. The inequality for finite sequences. Let 8p, q(a, b) be the largest of the values of the product for which xl,.., xm and Yl,..., Y~ are the result of a same partition applied to the finite sequences a=al,...,a~ and b=bl,...,bn.