]\gt ; On the ofSuccessions of Fourier Constants .
By Prof. W. H. YOUNG , Sc. D. , F.R.S. ( Received June 25 , \mdash ; Read June 27 , 1912 .
) S1 .
In a short note to appeal in the ' Comptes Rendus ' I have shown , by means of reasoning of a somewhat delicate nature , that , if we multiply together corresponding constants of two successions of Fourier constants , the succession of constants so obtained is the succession of the Fourier constants of a function whose summability depends on those of the two functions to which the two given successions belong in a certain definite way .
If and represent the indices of the summability of the given functions , that of the new function so obtained is denoted by .
I have there shown how this theorem may be applied to obtain a notable extension of Parseval 's theorem , and I have briefly indicated that we are thus able to give to that generalisation a still more complete form , namely , that the series , where .
is any positive integer , converges if the power of the function of which and are the Fourier constants is summable .
In the present communication I propose to give the necessary additional reasoning by which this result is obtained , and to deduce other important consequences .
For this purpose we quire the following generalisation of the first result noted above : the denominator of the expression where denotes the sum of the products of the positive quantities , taken at a time , is positive , then the succession of constants obtained by multiplying successions of Fourier constants together is associated with a function whose index of summability is , when those of the separate functions are , for all values of .
from 1 to ?
In this statement the sine and cosine constants of the original functions may be interchanged an even number of times without affecting the truth of the statement .
When the 's are all equal , our formula takes the simple form Prof. W. H. Young .
[ Jime 25 , The lilnitation that the denominator has to be positive imposes a certain restriction in the enunciation of the theorems obtained .
Thus it appears that , if , and and denote the typical Fourier constants of a function whose power is summable , the series and converge this is an extension for the case in which of a result I had already obtained by a different method in a communication already made to the Society , namely , that these series converge , whatever positive value be imputed to .
The absolute convergence in the case when follows from the convergence , proved below , of the series As has been stated , our results have only been obtained on the supposition that is less than unity .
That the convergence of this latter series does not continue to subsist when is greater than unity , follows from the examination of the Fourier series which defines Weierstrass 's non-differentiable function .
It appears , in fact , that continuous functions can be constructed whose Fourier constants , when raised to any positive power , however , give a divergent series when multiplied by any power of , however small .
Thus we have no new theorems about the squares of the coefficients , if the function has more than its square summable ; none about the fourth powers of the coefficients , if it has a power higher than 4/ 3 summable , and so on .
Moreover , the remaining theorems of the type considered are certainly the most that can be obtained by the method here explained .
In particular , we can only assert that the series whose general term is , converges , if , and can only assert that , when is odd , the convergence is absolute , if .
It shouJd also be remarked that we are only able to prove the former of these statements by applying the result of a paper presented some weeks ago to the Society .
When the denominator of the expression vanishes , it appears that the new function is a continuous function , so that its Fourier series converges when summed in the Cesaro manner .
This is a remarkable extension of a result which I have already given for a pair of series .
This extension leads to the conclusion that the series converges when the 's are positive , and , more generally , converges , whatever be the signs of the , when summed in the Cesaro manner , provided the power of the associated function is summable ; and if , as appears probable , the power of the function associated ith the allied 1912 : ] Multiplication of Successions of Fourier .
333 series of a Fourier series is summable when a greater , or perhaps even an equal , power of the function associated with the Fourier series is summable , it leads to the result that the series converges in the same manner , if the summability of the function is more than , or perhaps even equal to , .
It appears improbable that more than this can be stated .
Indeed , when is zero , it is not true that eVen in the ordinary manner , in the case in which the function has every power summable , or is even bounded .
Still less is it true that converges .
There is an obvious direction in which those of our results which contain power of may be generalised .
A modification may be shown to be allowable .
We may , for example , for such an expression as , where , write , where is a suitable monotone function whose limit is zero .
It would be foreign , however , to the purpose of the present paper to extend its results in this direction .
S2 .
In the note cited from the ' Comptes Rendus , ' I show first that if and summable , exists as convergent , jxcept for values of forming set of content zero , summable function of .
I then prove that the series , ( C ) where and are the constants of the ) functions and , is the Fourier series of I then obtain the following inequality , where Putting , and the power of the former and the power of the latter to be summable , it follows from the first result quoted that the first integral on the right of the preceding inequality exists as an absolutely convergent integral , except for a set of values of of content zero , and represents a summable metion of Since the second and third integrals exist , the second being a bounded function of , and the third independent of , it follows that the left-hand side represents a summable function of .
In other words the series ( C ) is ' Fourier series of a function whose power is provided has its its and VOL. LXXXVII.\mdash ; A. 2 Prof. W. H. Young .
June 2 S3 .
Hence by Parseval 's theorem in its completed form , converges if and are summable , provided Choosing , we may take , ( 2 ) since the function corresponding to the Fourier series differs only by a continuous function from a constant multiple of Fronl ( 1 ) and ( 2 ) and therefore Thus we have the heoreln : \mdash ; If is the typical constant of a function , power is converges .
th converges , being typical of , since the reasoning of the article holds when we into This is an extension of the result given in S 15 of my paper " " On a Mode of Generating Fourier Series where the lower bound given for was .
S4 .
Hence we have also the following theorem:\mdash ; Theorem.\mdash ; If constants of a function whose poeoer le where , the series converge absoh , provided .
For by the theorem , provided .
Hence , for such values of , which proves the first statement .
Similarly , the second statement follows , changing into I have ah.eady proved that these series are convergent , though not that they are absolutely convergent , also for values of , in a paper presented to the Society on May 11 of the present year , entitled " " On the ence of Certain Series the Fourier Constants of a Function 1912 .
] of Successions of S .
From what precedes , it follows that if has its power summable and its power summable , where , the series is the Fourier series of a function whose ( power is summable .
Hence we immediately get the result that , if in addition has its power summable , the series where is the typical Fourier cosine constant of , is the Fourier series of a function whose summable , where denoting the sum of the products of the indices , ( 1 , taken at a time , and the denominator is supposed positive .
We have ingly ving theorenl , which ' oved by induction:\mdash ; Theorem.\mdash ; lf has , for , 2 , , its typical , cosine the scri . .
is the Fourier of afuiction whose } , where denoting the of the products of the tities p , , provided the of is positive .
In fact , since the numerator of the above fraction may be written we have , assuming the theorem proved up to and adding another metion , for the index of the required power a fraction whose numerator evidently has the required form , and the denominator is which , when the products involving are combined with the remaining products according to dimensions , takes the required form .
Since the statement has been proved for three this proves the theorem .
{ 336 Prof W. H. Young .
Cor.\mdash ; If a is , lvhere , is Fourier of a function power is lJnable .
In fact , denoting the denominator of the fraction iven in the enunciation of the theorem , when all the are equal , by , we have whence that Since the numerator is } ) , this giyes the required result .
S6 .
We can now proye the theorem : Theorem.\mdash ; If is a function ?
power is , 'cos'w ; is the series of For , writing .
Hence , by the preceding theorem , the series ( i ) is the Fourier series of a function whose power is summable , where But Thus , by the same argument , the series ( ii ) , got from ( 1 ) by changing into , is the Fourier series of a function whose power is summable .
Hence , by a known theorem , cosine series from ( i ) and ( ii ) by ultiplying together coefficients to form the coefficient of the corresponding term of the new series is the Fourier series of a continuous function .
This proves the theolenl .
As an immediate corollary we have the following theorem:\mdash ; Theorem.\mdash ; If has its power when in ro .
Hcnc , if positive for all of , or the integcis in th *W .
H. Young , " " On a Class of Parametric Integrals and the Theory of Fourier Series ' Bo Soc. Proc 1911 , , vol. 85 , and 1912 .
] Multiplicalion of ofConstants .
337 S7 .
We have seen in S5 that , if there are functions , the fraction giving the summability of the fimction corresponding to the Fourier series of cosines whose coefficients are the products of those in the Fourier series of the functions may be written proyided the denominator of this expression is positive .
If , on the other hand , the denominator is zero , we have , with the notation of S5 , Hence , by the same reasoning as in the preceding article , we have the following theorem , of which the first theorem of the article is a particular case : Theorem.\mdash ; If has its rnmablj , , 2 , its cosine constant the is the of a continuous function , ovided the of the which , if give the of the function corresponding to the series , zero , provided where denotes the of the of th positive uantities p , taken at time .
S8 .
If , as in S3 , we take and , applying the work of the preceding article , put we shall have Hence we have the first result of the following theorem , using the result of S7 .
The final statement follows from a theorem lately proved by 1nyself in the paper already cited in S4:\mdash ; Theorem.\mdash ; If for , 2 , .
, the function has its power summable , and has for its pical .
cosine constant , the oos is the Fourier scries of continuous function , provided , where has signific ) of S5 .
Moreover , the series then Prof. W. H. Young .
[ June 25 , A special case is embodied in the following theorem:\mdash ; Theorem.\mdash ; If has ?
power summable , is its ' cosine constant , is the serics of , provided and is then S9 .
We cnn strife ?
than the convergence of the if ?
fttrther strict p so as to ) , , of course , now an odd integer .
the lestion ?
th absolutely convergent .
When , this from S6 , since the series is convergent , and absolutely for When , the result follows from the final theorem of the preceding article , since the series then convel.ges absolutely , proyided This proves the statement ) at the of the present article .
Putting , we have the special case : If , the is convergent , and it is certainly alt positive of all cases quan tity greatethan Again , have the following:\mdash ; If ' convergent , and it is if in all .
S 10 .
Since has the same degree of summability as , .
the , and has its power summable , the series is the Fouriel selies of a function whose power is summable .
Hence it at once follows from the arguments used that in our 1912 .
] Successions of theorwe replace even number of the successions of cosine constants by sine constants .
S 11 .
If however in the series ( C ) we replace the 's by zeros and the 's by zeros , which , as we saw , is legitimate , we get a sine series , and not , as in the preceding article , a cosine series .
Thus the subsequent arguments used would not apply ; they would of course apply if the 's and 's could be interchanged in the series , that is if it is true that the allied selies of the Fourier series of a function whose power is summable is also the Fourier series of a function whose power is summable .
This proposition has not been proved , although , for various reasons of a more or less convincing character , I am inclined to believe that it is true , or at least that the allied series is the Fourier series of a function whose power is summable , where is any quantity less than S12 .
In the various theorems here obtained there has always been a limitation on the value of .
Thus in the theorem concerning the convergence of had to be less than unity ; if we have the theorem of Parseval ; in the theorem concerning the ence of had to be less than 1/ 3 , while , if , the series is and so on .
It should be noticed that , when has a value greater than the limit imposed in the present paper , the index of the power to which raised changes sign .
Our reasoning evidently fails in these The quest , ion fore remains over as to whether the formulae are none the less true without ths limitation imposed on .
That the answer to this question must be negative appears from a consideration of the series Weierstrass 's non-differentiable function .
This shows that cannot even assert in the case of a continuous function that any positive power of however small , multiplied by any power of , however , need be the general term of a convergent series .
Indeed Weierstrass 's series is where is an odd integer , and .
This is a Fourier series with groups of terms absent owing to the corresponding Fourier constants being zero .
The series under consideration is therefore This will therefore certainly diverge if

