]\gt ; 178 Prof W. H. coefficient of its integral , we get at once interpreted iense.emarked tnthis equation itrue wntegration isLebesgue Gvident frecedes , bearing ibeing continuous , Here tntegration refers tntegration o , and is , therefore , of a very elementary type in our scheme of.functions .
On the Formation ofConvergent Fourier Series .
By W. H. YouNG , Sc. D. , F.R.S. , Professor of Mathematics , University of Liverpool .
Received December 19 , 1912 , \mdash ; Read January 30 , 1913 .
) S1 .
Series which converge except at a set of content zero , or , using the expression very commonly adopted , series which converge usually , ossess many of the properties which appertain to series which converge everywhere .
It becomes , therefore , of importance to devise ciroumstances under which we can assert the consequence that a series converges in this manner .
The subject has recently received considerable attention .
So far as Fourier series are concerned no result of even an approximately final character has been obtained .
It may be supposed , indeed , that the results* Jerosch and Weyl were at first so regarded , but , if we examine them closely in the light of the -Fischer theorem , which was known previously to the results of these authors , it becomes evident that they are merely equivalent to the statement that the Fourier series of a function , whose square is summable , is changed into one which converges usually , if the typical coefficients and are divided by the sixth root of the integer denoting their place in the series .
Now it is to believe that the question of the usual convergence of a Fourier series can depend on the degree of the summability of the function with which it is associated , 'Math .
Anm vols .
66 and 67 .
The result due to Fatou that a series of Fourier converges usually if and converge to zsro is still more special , being of course included in Jerosch 's condition .
For Fatou 's paper , see ' Acta Mat vol. 30 .
1912 .
] Forvnation of Usually Convergent Fourier Series .
and it is still more difficult to see how precisely the sixth root of can have anything to do with it .
On the other hand Weyl 's method , which itself marks bn advance on that of Jerosch , does not obviously lend itself to any suitable modification which would secure a greater degree of generality im the result .
The mistake is frequently made of confusing theoretical interest withr practical importance in the matter of a necessary and sufficient test .
Testa which are only sufficient , but not necessary , are often much more convenient. . .
Still more frequently it is convenient to work from first principles , and not to use any test at all .
Instead of employing Weyl 's necessary and sufficient condition that a series should converge usually , I have attacked the problem directly .
The principles I have employed do not differ essentially from ! . .
those already exposed in previous communications to this Society , but the ' generality and interest of the results obtained in the matter in hand seem to justify a further communication .
These results are as follows:\mdash ; The Fourier series of any function uhatever , of any degree of summability , and its allied series , are both of them ( jhanged into Fourier which converge : usually , the coefficients and are divided by any power , however small , of the index denoting their place in the series , in other words , they are converted into such by the use of convergence factor , and by the of the convergence factor log More generally , they are converted into sueh series by the use of the convergence factor whose numerator unity , and whose denominator is , where and I have thought it .
sufficient to prove this latter theorem for the case in which .
With this before him the reader will easily be able , by induction or otherwise , to carry out the proof of the general result .
It will be remarked that , unlike Weyl , I have not considered general series of orthogonal functions .
The theory of these functions usually greater analytical difficulties , but not a greater wealth of ideas , and I have thought it best , with so many problems in the theory of Fourier series still unsolved , to confine my attention to this simpler class of seriesThe detailed study of the theory of Fourier series seems to me to form the best possible preparation for the larger theory .
Finally , it should be noticed that the results arrived at do not in any way strengthen the probability suggested by Lusin that the Fourier series of a function whose square is summable necessarily converges usuallyVOL LXXXVIIJ.\mdash ; A. 180 Prof. W. H. Young .
lusin 's statement that this is infinitely probable seems based that in any succession of partial summations of such a series a sub-succession $3 can be found which usually converges to the function .
This is , however , quite consistent with , for example , finite oscillation except at a set of content zero .
On the other hand , the results perhaps do suggest the possibility , though scarcely the infinite probability , of this latter circumstance presenting itself for all Fourier series .
If this be true , every monotone succession of constants , having zero as limit , would have the effect of the special series of constants employed in the theorems of the present communication .
These latter , however , are so far from being the most general successions of the monotone type that they are the Fourier cients both sine and of cosine series .
S 2 .
In the mode of investigaJion I have adopted , the following theorem and its analogue are fundamental:\mdash ; Theorem.\mdash ; liet be a summable function whose typical Fourier cosine a sine constants are and , and let , , , .
be a monotone descending sequence of constants with zero as limit , then the series boundedly at the point , provided a succession of constants can be found such that ( i ) bounded for values The series converges , ( iii ) is a function of Let denote the partial summation consisting of the first terms of 1 ) , and let .
( 2 ) Then , since , and , we get , rerarding f as periodic with period 2 Now 2 sin$u .
Formation of vergent Fourier Series .
Therefore , by and ( 4 ) , , ( 5 ) where Su \ldquo ; and is accordingly , to a constant factor pris , the n-th partial summation of " " lihe Fourier series of From ( 5 ) the truth of the theorem is at once evident , since by the hypothesis ( iu ) is a bounded function of , while the conditions ( i ) and are satisfied .
S3 .
The corresponding theorem .
for the allied series of is as follows:\mdash ; Theorem.\mdash ; Let be a summabte function , whose Fourier cosine and sine constants are and let , , , , be a monotone descending sequenee of constants with as limit , then the series ( 1 ) oscillates boundedly at the point , provided a succession of constants can be found , such that ( i ) is bounded for all values of ( ii ) The series 2converges , ( iii ) is a bounded function of Let denote the partial summation consisting of the first terms of and let Then , as in S2 , .
( 2 ) Now ( 3 ) Therefore , by ( 2 ) and ( 3 ) , ( 4 ) where ) 182 Prof. W. H. Young .
the allied series of the Fourier series of .
thesis ( bounded function oconstant factor psummation oying these theorems wrequire slementary r are bounded functions of in the convergence of series .
Lemma \mdash ; If where has posibive integral value , imjluding zero , and , then Writing we have But , which proves the lemma .
Cor.\mdash ; If , and has the same form as , with instead of , the series whose general converges .
Lemma \mdash ; If and the series 2convergent .
Writing we have ) ) .
But , whence .
Thus the series whose convergence is under discussion is not greater than the sum of two series both known to be convergent , namely , and This proves the lemma .
It should be remarked that if the index of the power of in the 1912 .
] Fonnation of Usually Convergent Fourier Series .
expression for be instead of , the series in question may be easily seen to be divergent .
Lemma and the seria 2 convergent .
Writing we have Hence , as in preceding proof , our series is the sum of series known to be convergent , which proves the lemma .
S5 .
In this and the next articles we shall prove certain properties as to the order of infinity of the partial summations of a Fourier series , and of its allied series , except at a set of content zero .
Theorem.\mdash ; Ifbe any surnmable fuwtion of , and , the integrals and are , for each value of not belonging to a certain set of zero content , bounded funetions of In fact the former integral may be written the absolute value of which is therefore where is a constant depending only on Here we have tacitly supposed that the integral just written down exists , which will certainly be the case , except for a set of values of of zero content , in virtue of a theorem that I have proved elsewhere , since and eosec are summable functions of In precisely the same way the second integlal is proved to be less in See W. H. Young , " " The gence of Certain Series Involving the Fourier Constants of a Function ' Boy .
Soc. Proc 1912 , , vol. 87 , p. 221 .
" " Sir la ralisation d theor6me de Parseval 'Comptes endus , ' 1912 , vol. 136 , p. 30 ; seance du ler juillet .
are , each value of not belonging to a certain set ' zero , bounded { $ functions of We have and therefore since llog is bounded in any closed neighbourhood of the origin , and remains bounded as increases indefinitely .
Hence since Sullog is summable in an interval enclosing the origin not containing the point the result stated with respect to the first integral is true , by reasoning similar to that employed in S2 , bearing in mind that it is clearly sufficient to prove the property in question for an interval of integration not including the point Precisely similar reasoning proves the statement with respect to the second integral , using the fact that is bounded .
S7 .
By a slightly more complicated process we arrive at the following result : Theorem.\mdash ; If be a summable function of , and log log , the integra and $ are , for valuof belonging to a certain set of zero , bounded functions of 1912 .
] Formation of Usualty Convergent Fourier Series .
We have For convenience of printing the quantities which now occur are to be supposed all taken in absolute magnitude .
Moreover , as in S6 , we may confine our attention to a conveniently small interval containing the origin , and not containing either the point unity , nor , and we need only consider the positive part of this interval .
We have then logJog log log Multiplying out , we get the sum of four terms , each of which is bounded , namely mu and mu , each multiplied by Hence , since is summable , the required result follows as before , for the first integral .
In a precisely similar way it follows for the second integral .
S8 .
The general theorem that I have been establishing step by step with regard to the order of infinity of the successions of our summations is now evident .
It is as follows:\mdash ; Theorem.\mdash ; If a mmable function of , the -th partial snmmation of the Fourier series of , and the corresponding summation connected with its allied series have an order of infinity which , for every and , is for every not belonging to a certain set of content zero , less them that of The reader may be left to complete the proof by induction , or otherwise .
S9 .
We are now able to prove the main theorems which form the subject of the paper .
We have only to take the fundamental theorems of SS2 and 3 instead of that of S2 .
Theorem 2.\mdash ; Under the same circumstances as in preceding theorem , the series , and converge usualty .
For since the condition ( i ) of SS2 and 3 is satisfied .
The condition ( ii ) is satisfied in virtue of the Lemma 1 of S4 .
Finally the condition ( iii ) is satisfied by S6 .
..Hence by SS2 and 3 and Abel 's Lemma the theorem is true .
Theorem 3.\mdash ; Under the same circumstances , the series converge usuatly .
1912 .
] Ynation of Converge.nt Fourier Series .
Hero we take and We then have showing that the condition ( i ) of SS2 and 3 is satisfied .
The condition ( u ) is satisfied in virtue of the Lemma 2 of S4 .
Finally the condition ( iii ) is satisfied by S7 .
Hence by SS2 and 3 the theorem is true , using once more Abel 's Lemma .
Theorem 4.\mdash ; Under the same circumstances the series 2 and 2 sin , where , usually .
Here we take We then have showing that the condition ( i ) of SS2 and 3 is satisfied .
The condition is satisfied in virtue of Lemma 3 of S4 .
Finally the condition ( iii ) is satisfied by S8 .
Hence by SS2 and 3 the theorem is true , using once more Abel 's Lemma .
S10.*If , instead of the theorem that exists as a function of almost everywhere , when and are summable functions , we employ the connected result that this function exists everywhere and is continuous , when the summabilities of and ars suitably connected , we obtain in a similar manner information as to the order of infinity everywhere of the partial summations of the Fourier series of the various types of functions .
In this way not only are results such as those exposed in my papers " " On the Convergence of Certain Series , etc. , \ldquo ; and\ldquo ; On the Fourier Series of Bounded Functions , \ldquo ; confirmed , but further new ones are obtained .
S11 .
On the other hand , Mr. G. H. Hardy points out that it is possible to slightly extend the results of this paper as regards Fourier series by utilising a theorem of Lebesgue 's instead of the above-mentioned theorem .
Indeed , we are thus enabled , in the case both of the Fourier series and its allied series , to replace the two 's by unities in the general convergence factor .
Moreover , in the case of the Fourier series\mdash ; as distinct , be it said , from the allied series \mdash ; we can , as he remarks , go still farther and obtain a further slight , but , as it *SS10-12 have been added during the passage through press , Feb. 8 , 1913 . .
supra .
'Lond .
Math. Soc. Proc 1912 , Ser. 2 , vol. 12 .
between the great mirror and its principal focus , is concave , and therefore shortens the effective focal length , in place of increasing it .
The deformations from spherical figures are also so great , especially for the great mirror , as to leave it doubtful whether the construction discussed could ever be the model for practicable instruments .
If we keep to the Cassegrain form , spherical aberration and coma may equally be corrected by deformations of the mirrors which , though large , are less extreme , but there remains a onounced curvature of the field .
For this reason I am led , in the present memoir , to consider more complicated systems produced by the interposition of systems of lenses .
Achromatism be preserved completely for a single focus if there are three lenses of focal length determined when their position are given , and if all are made of the same glass .
One of these lenses , which I 'lloy .
Soc. Proc 1912 .
' K. Gesell . .
Wissenschaften zu Gottingen , Math.-Phys.-Classe , ' Neus Folge , 1906 , vol. 4 .

