]\gt ; On Classes of Functions and their Four Series .
By Prof. W. H. YOUNG , ( Received May 30 , \mdash ; Read June 27 , 1912 .
) S1 .
Functions which are summable may be such that certain functions of them are themselves summable .
When this is the case they will possess certain special properties additional to those which the mere mability involves .
A remarkable instance where this has been nised is in the case of summable functions whose squares also are .
The\mdash ; in its formal statement almost self-evident\mdash ; Theorem of Parseval which asserts tha6 the sum of the squares of the coefficients of a Fourier series of a function is equal to the of the square of , taken between suitable limits and ultiplied by a suitable constant , has been recognised as true for all functions whose squares are summable .
Moreover , not only has the converse of this been shown to be true , but writers have ) led to develop a whole theory of this class of functions , in connection more especially with what are known as integral equations .
That functions whose power is summable , where , but is not necessarily unity , should next be considered , was , of course , As was to be expected , it was rather the of such functions the functions themselves properties were required .
Lebesgue had already given the necessary and sufficient condition that a function should be an integral of a summable function .
F. Biesz*then showed that the necessary and sufficient condicion that a function should be the integral of a function power is summable had a form which constituted rather the genelalisation of the expression of the fact that such a nnction has bounded variation , than one which included the condition of as a icular case .
S2 .
I myself was led to consider functions in connection with the problem of the integration of Fourier series , and , in a paper published in the ' Proceedings ' of the Society , I have proved that term-by-term integration of the Fourier series of such a function is allowable after it has been multi- plied by a function belonging to the complelllentaty provided only the summation is carried out in the Cesaro manner .
The method I employed in the paper in question has the great ' ' Systeme Integrierbarer Funktionen ' Math. Ann 1910 , vol. 69 , pp. 449\mdash ; 497 .
in the Theory of Fourier Prof. W. H. Young .
On of Summable [ May as regards the details of the work that it is based on an extremely simple inequality .
From this point of view it is far superior to the other methods of proof , almost intuitive as they are , when the necessary preliminary theorems have been proved , that I have indicated elsewhere .
* Instead of requiring the generalisation of Schwarz 's well-known inequality , it suffices to be acquainted with the generalisation of the relation 2 , namely , that In the prosecution of research one is often hampered by the difficulty of generalising a known formula , owing to the very simplicity and obviousness of its statement .
The inequality just written down possesses , however , the great advantage over Schwarz 's inequality that its generalisation is almost immediate .
It is as follows:\mdash ; If a positive monotone increasing function of a positive real variable possessing a differential coefficient is positive , and the inverse function such that , then , ( 1 ) where .
In fact the function of represented by has a negative second differential coefficient , its first differential coefficient vanishes when , that is when .
It has accordingly a maximum whose value is , that is ; or , what is the same thing , .
Hence the inequality in question follows .
S3 .
With this inequality in our minds we may conceive of summable functions grouped in pairs of classes in such a manner that the product of twc functions , one from each dass of such a pair , is summable .
If , for example , we put we see that if and are summable functions of , where and are themselves functions of , then is also a summable function of " " Successions of Integrals and Fourier Series S26 seq. , 'Proc .
L.M.S. , ' 1911 , Series 2 , vol. 11 , pp. 43\mdash ; 95 .
1912 .
] Functions and their Fourier Series .
Moreover we have to retrace the reasoning of SS7 seq. of the paper cited on ' A Class of Parametric Integrals\ldquo ; to see that is a continuous function of , provided and to such a pair of classes of summable functions , and that accordingly we may assert that if we multiply the Fourier series of one of them by the other and intra term by term , the result obtained the integral of the product of the two functio between the limits of integration , provided only we the summation in the Cesdro manner .
In particula , and , denote the typical Fourier constants of two such functions we the summation being formed in the manner .
S4 .
But further we are naturally led , in virtue of the great generality of the theorem just enunciated , to attempt to extend to such new classes of functions the general theory of the functions whose powers are summable .
In particular we require the necessary and sufficient condition that a function should be the function belonging to such a class .
For this purpose it will be found that the inequality is sufficient , ( 2 ) where is the indefinite integral of a positive function whose differential coefficient is positive .
This inequality follows in fact by the theory of integration from the simpler inequality which is obtainable at once by making the right hand side a subject to the condition that should be constant .
We thus obtain the necessary and sufficient condition that a function should be such that , where possesses a positive differential coefficient , is a summable function of that where the integral of , is constant , ) endent of , and of the set of non-overlapping over which the summation is extended .
Prof W. H. Young .
On Classes of Sum [ May 30 , The necessity of the condition is , in fact , an immediate consequence of the inequality ( 2 ) , while its sufficiency follows from reasoning similar to that employed by ] .
Riesz in the special case in which S5 .
The theory of sumnlable functions is interesting from the abstract side from two points of view .
As compared with the theory of the con .
vergence of series we have the great that necessary and sufficient conditions exist , and can be easily formulated .
Just as , however , given a convergent series , we can always find a series which is less , and one which is more convergent , so , iven any summable , we can find whole classes of functions which are less summable , as well as classes of functions which are more summable .
It would appear , indeed , that , given any summable function , a function exists , of the type considered , such that is also summable .
If this be so , the fact that , if we put , for example , in the condition of F. Biesz , above referred to , and make the substitutions in the various conditions of which the general type has been written down , we get the condition for bounded variation , finds its explanation .
The theory of summable functions is also of intel.est when we are concerned with potency and correspondence , and compare the ensemble of all summable functions with , for example , that of the continuum .
If we were to confine our attention to functions some power of which is summable , 01 rather to the grouping into classes determined such considerations , we should , of course , have -correspondence between the classes and the continuum .
In point of fact , however , we have functions which , for example , have their squares summable , and no higher power ; while we also have functions whose squares are not summable , but which , in a more than continuously infinite manner , have all powers less than the second summable .
S6 .
I do not propose on the present occasion to enter into other generalisations for the wider classes here considered , of the properties of known classes of summable functions .
I may say , however , that the results , for example , iven in the paper by F. iesz , already cited , are readily generalisable , as well as the cognate results of Weyl and others .
To obtain the complete generalisation in the case of certain theorems , if not of the results at least of the methods , employed by these authors , it will be necessary to obtain the most general form for the purpose in hand of the inequality of Schwarz .
The nature of this generalisation may perhaps be inferred from the following special case , which is deducible bylimiting processes from Riesz 's form , or may be proved independently by considerations of maxima and minima:\mdash ; and being positive functions .
1912 .
] Functions and their Series .
In the same connection the following formula will be found of Such inequalities will be found useful for example , in extending the results on Fourier series contained in my paper on " " Successions of and Fourier series S7 .
We thus obtain , for example , theorems of the following type : If denote the -th Cesaro partial sumrnation of the Fouri series of positive function , which ?
such that is denote th -th partial surnmation of the Fourier series of the function , then .
From this also we deduce that , if denote the n-th partial mation of the Fourier series of a function [ which is such that is , then has the double limit ; and that Finally we may remark that from analogous considerations we may establish the truth of the theorem on double successions:\mdash ; If the summations of the first }of the series of two funwtions , belonging to a paiir of classes of summfunctions , such as here sidered , Negative After-Images Successive with Pure Spectral Colours .
By A. W. PORTER , B.Sc. , F.RS .
, and F. W. EORIDGE-GBEEN , , F.R.C.S. ( Received May 28 , \mdash ; Read June 27 , 1912 .
) [ This paper is published in ' Proceedings , ' Series , No. 581 .
] VOL. LXXXVII .
It

