]\gt ; On Trigonometrical whose Cesuro Partial Summations Oscillate Finitely .
By W. H. YOUNG , Sc. I ) .
, F.R.S. , Professor of Mathematics in the University of Liverpool .
( Received May 21 , \mdash ; Read June 19 , 1913 .
) S1 .
Riemann 's remarkable theorem , which , in the extended form given to it by Lebesgue , by virtue of the use of the concept of generalised integration , asserts that a trigonometrical series is a Fourier series if it converges everywhere , except at a reducible set of points , to a function which is summable and has a value , or values , everywhere finite , has been discussed , and still further extended , by a relatively large number of writers .
* The object of the present communication is to state and prove certain results which include all those at present known .
They are as follows:\mdash ; I. the upper lower fumtions of the succession of the Cesaro partial , index not them unity , of a trigonometrical series which is such that and as , are summable , and everywhere finite except possibly at a set of points which no perfect sub-set , then the series Fourier series .
II .
If the upper lower functions of the succession of the Cesaro partial summations , index , of a trigonometrical series , are summable and everywhere finite , then the series is a Fouriseries .
S2 .
In proof of these results all turns on the use of upper and lower derivates of a generalised character .
If denote the function , it is open to us to consider , not merely the limits of It will be sufficient for the purpose of the present communication to refer to the following:\mdash ; W. H. Young , " " On the Conditions that a ourier Series should have the Fourier Form 'Lond .
Math. Soc. Proc 1910 , Series 2 , vol. 9 , pp. 421-433 ; Marcel Riesz , " " IJeber summierbare trigonometrische Reihen 'Math .
Ann 1911 , vol. 71 , pp. 54-75 ; De la Vallee Poussin , " " Sir l'Unicit6 du Developpement Trigonometrique ' Bull .
de l'Acad .
Royale de Belgique , ' 1912 , pp. 702-718 , and , 1913 , pp. 9-14 .
We shall also have occasion to refer to L. Fejer , " " Untersuchungen uber Fourier'sche Reihen ' Math. Ann 1903 , vol. 58 , pp. 51-69 .
for example .
We may , as is well known , consider with advantage the limits of the expression obtained from this formally by integrating numerator and denominator with respect to separately between the limits and .
The expression so obtained , though not conceived of in this manner , has indeed formed the subject of considerable discussion .
No one appears , however , to have seen the advantage that could be obtained from the consideration of the expressions obtained when the process in question is repeated .
L. Fejer , in his very valuable paper , *employs in fact an entirely different expression which has not the simple properties of those here referred to .
By a known theorem in the theory of indeterminate forms such expressions as those made use of in the present communication have at each stage limits lying between the limits of each of the expressions obtained at a lower This fact enables us to see that the higher expressions share the striking properties shown by de la Vallee Poussin to be possessed by A modification of Feje'r 's reasoning is then all that is required to obtain the main theorem from which the first of the results iven in S1 immediately follows .
To obtain the second of the results we have then merely to make use of certain simple facts in the theory of Cesaro summation , due essentially to Knopp and Marcel Riesz , though not as far as I am aware explicitly stated by them .
S 3 .
As ards the importance of the results here exposed , this does not consist merely in the fact that they include previous results as special cases .
This will be evident if we reflect on the part played by Cesaro summation in the theory of Fourier series .
We know practically about the mode of convergence , or oscillation , of a Fourier series when summed in the ordinary manner , except in one or two isolated instances .
On the other hand , all such series converge almost everywhere when summed in any Cesaro manner .
Thus , the upper and lower functions of such Cesaro partial summations have the same degree of summability as the function corresponding to the Fourier series , differing , in fact , from it only at a set of content zero .
That the most general results hitherto obtained , those due to de la Vallee Poussin , remarkable as they are , are not final , is then at once evident .
It should not be necessary to have information with to the ordinary upper and lower functions , in order to say whether a .
cit. , supra .
Marcel Riesz 's interesting investigations loc. closely connected with .
cit. , supra .
Prof. W. H. Young .
Series whose trigonometrical series is a Fourier series ; it should be sufficient to have the corresponding information with regard to the Cesaro partial summations .
It is precisely this step which the results of the present communication enable us to take .
Moreover , in this way the investigations which took their rise in Riemann 's paper take a more or less final form which brings them into formal analogy in some sort with the necessary and sufficient conditions I have shown to hold that a trigonometrical series should be the Fourier series of a function of a certain type .
These necessary and sufficient conditions are , in fact , expressed in the language of Cesaro summations .
The interest is further enhanced if we appeal to the analogy between the properties of Fourier series when summed in the Cesaro manner with the properties of derivates .
The analogy is a striking one .
Thus we have the pair of theorems:\mdash ; If fcmction us an integrat , ?
upper and ivates agree except at a set.of content If a series converg to integral , upper and lower functions of its derived series , when summed any ro manner , agree except at a set of content zero .
's Theorem remains true when for an integral we write a function of bounded variation , and I have recently shown that the same is true of the theorem .
We are thus naturally led to inquire whether the counterpart of Lebesgue 's Theorem , also proved by him , has its correspondent in the tpeory of Fourier series .
The present paper enables us to answer this question in the affirmative , and in this way to complete the analogy .
We have , in fact , the following pair of theorems:\mdash ; If the of a and finite , the fnnction is an .
( Lebesgue .
) If the lower functions the dprived series of a trigonometrical series , uhen sumrned some Cesara of positive index less than unity , are snmmabl and finite , the trigonometrical series converges to an integral .
Moreover , corresponding to the known extension of Lebesgue 's Theorem to which I have called attention , in which exceptional points are permitted , we have a corresponding extension of its analogue in the theory of Fourier series .
CesSummations Oscillate Finitely .
S4 .
We shall now obtain the theorem which corresponds to that proved by Feje it in a which we can at once apply .
Theorem.\mdash ; If siu and .
scrics , pllld the second , of subsequent condition , if denote of th succession of tho modnli of ations o}}in .
thr Cesaro , then of as , arc ricall .
, ?
is a constant of it bjing that We have whence ( rh ) , say , so that Now write Hence ' since , by S5 , ?
is bounded , and ?
for fixed vanishes as .
Thus .
cit. , pp. ) .
theoretll states that " " if converges ( C1 ) and , and if when then ' ( nt ) converges for every positive value of , and if , the limit of its sum when is the Cesaro sum of the series Prof. W. H. Young .
Trigonometrical Series whose Now has finite upper and lower bounds at the point considered , while ( mh ) and have each the unique limit zero as , therefore the final term on the of the last expression vanishes as .
Hence Now we can find a point internal to the interval so that Therefore where is a number independent of , but depending on , such that , for values of lies between and , whence , denoting by the yreater of and , and is such that We can confine our attention to values of so small that , and therefore Now Therefore Hence for all values of , say , while , since we see that , for is numerically less than a certain corresponding number .
In particular , for values of , since , we have , provided .
Hence Cesaro Partial Finitely .
The first summation on the has , since is independent of , the unique limit zero as .
The second summation is , and has , therefore , the limit .
Finally , since the third smmmation is less than .
Hence , finally , , where vanishes with , for values of less than 1/ 3 , and less than where NI depends on , and is as small as we please .
Thus where is independent of .
If we next make the hypothesis that and have zero as limit when / , it at once follows that the series of which is the sum converges uniformly , and that accordingly is a continuous function .
It also follows that , as iemann has shown , has zero as limit , as Further , as we have already remarked , the upper and limits of lie between those of Hence , on the hypothesis stated in the first of the theorems stated in S1 , it follows that , if be any one of the limits of , it is one of the limits of the latter expression , and is slunmable and finite except at the points of a set which has no perfect sub-set .
In fact , by the theorem of S4 it is numerically less than a metion h these properties .
We have therefore only to apply Theorem / de la Vallee Poussin 's first paper , above cited , to see that our theorem is true .
* * Here we may conveniently make of the theorem that a trigonometrical series is a Fourier series , if its second integrated series converges to a repeated integral .
Moreover , the function of which it is the Fourier series is the second differential coefficient of this repeated integral .
See W. H. Young , .
cit. 156 Prof. W. H. Young .
Trigonometrical Series whose $ S6 .
Now let us omit the hypothesis that and converge to zero .
Then it will no longer be possible to assert a priori that the expression approaches zero as .
Thus the reasoning of S5 will no longer apply .
We can no longer indeed assert that is continuous , unless we suppose , which is the case in the enunciation of our second theorem , that the loriven trigonometrical series is summable ( Ck ) , where has some value less than unity .
It is then possible , as is shown below , to prove that the series of which is the sum converges uniformly , so that is continuous .
We are thus able to use de la Vallee Poussin 's Theorem 6 , and so obtain a proof of the latter of our two main results .
In our ument we make the tacit assumption that , if the upper and lower functions of a series are finite at an crned point , the same is true of the upper and lower functions ( Ck ' ) , where .
Moreover if the former are summable , so are the latter , these in fact less in absolute value than a numerical multiple of the former .
S7 .
It remains then to show that if the series oscillates boundedly ( Ck ) , and denote the ordinary partial summation of the series , then is bounded , while the series obtained by each term by the power of its place in the series oscillates boundedly when summed in the ordinary manner .
* Denote by the Cesaro partial unmation .
Then , in the case in which , the required result follows immediately from the equation and the equation When is not unity we write whence multiplying both sides by and comparing coefficients the second of the results follows .
The first result follows similarl by both sides by , and the fact that S8 .
We have also to obtain a relation between the partial summations From this follows that the first , and therefore the second , of the series in the enunciation of S4 converges everywhere .
Cesuro Oscillate and the partial summations .
For this purpose we have merely to use the equation .
Now write , and expand on both sides , and equate coefficients of Suppose a fixed number so that , for is less than its upper bound as a suitable small quantity has been added to the latter .
Then , for } the limits of the coefficient of on the left , taken in absolute value , are expressible in terms of the limits of the coefficient of on the -hand side , by means of an inequality .
The required result then easily .
Thus we obtain the result that the upper limit of the modulus of the Cesaro partial summations ( Ck ' ) is less than a numerical multiple , *independent of , of the upper limit of the modulus of the Cesaro partial summations .
It follows that , if the Cesaro partial unmations ( have the property hypothecated in the second theorem of S1 , the Cesaro sumnlations have the same property .
Thus the theolem of S3 is applicable .
It is of course not necessary for our purpose here to show that this numerical quantity is unity .

