]\gt ; On the Cocfficients Fourier Series .
On the theoretical side I have shown that J. J. Thomson 's formula is confirmed by the present method of analysis , and that the Lorentz-Einstein formula for a " " contracted\ldquo ; electroll requires to be modified in the manner stated .
If it be argued , as may fairly be done , that the experiments do not conflict with the instein formula , we require to up the view that the electron is contracted in the proportion they assume .
But it is fairly certain that there is an electron of form intermediate between Abraham 's sphere and Lorentz ' spheroid which would have a transverse inertia correctly given by .
This , however , implies .
up the correspondence assumed in the relativity theory , and we might choose between this and the alternative view ( by no means unreasonable ) that electrons as obtained in experiments possess ordinary as well as electric inertia .
It appears that still further experiments are required , and it would be helpful if results were represented in a form that could be used to test any formula and not any particular one .
The subject is still too uncertain to warrant the view that any special formula represents the end of the matter .
On the PIode of to Zero of the Coefficients of a Fourier Series .
By W. H. YOUNG , F.R.S. ( Received February 2 , 1917 .
) S1 .
In the present communication I give a number of results on the mode of approach to zero of the coefficients of a Fourier series , to which I have already made allusion in my paper on " " The Order of nitude of the Coefficients of a Fourier Series These results are not merely curious , they have a real importance , and give one an into the nature of these series , which cannot easily be gained without them .
Indeed , while the earlier paper leads , as I showed in a subsequent communication , to the discovery of classes of derived series of Fourier series , which , although not themselves Fourier series , none the less conver , and are utilisable in a similar manner , and is therefore in a certain sense of practical interest , the present paper does something towards the elucidation of the eneral theory of the convergence of Fourier series themselves , as well as of their derived series .
It will be sufficient to give a single instance .
I have recently shown that VOL. XCIII .
Prof. W. H. Young .
On the Mode of the well-known test of Dirichlet for the convergence of a Fourier admits of a remarkable generalisation .
It follows from Theorem .
1 , given below ( S7 ) , that gence secured by that test and by its generalisa .
tion alike possess what may be called greater strength than the rival tests Dini and de la oussin .
S 2 .
In the course of the proofs of the auxiliary theorems Al , , we shal ' .
require the following Lemmas : LEMMA cxpression according as , and monotone djscending or ascending with respect to according as LEMMA 2.\mdash ; For of , the ( jxpression positive , it decreases or with respect to according as LEMMA \mdash ; If , then mcux .
value of , for Also if , then value of , for LEMMA \mdash ; If , or if , then ; and if , then S 3 .
THEOREM Al.\mdash ; If , and , where , then either oj the is a bounded function of The argument being the same in both cases , we only the sine expres .
sion , which we shall denote by S. Let us consider separately the values cf and those , where any fixed positive quantity .
In the former case , so that is bounded .
In the latter case , , and we have rv Approach to Zero of the Coefficients of Fourier Series .
457 Hence ( 1 ) where rv , ( 2 ) and only differs from in having sines in place of the cosines .
Let us divide the interval into equal parts ; each of these is , by hypothesis , less than , and the point will lie in one of them , or coincide with one of the points of division .
this part as the , we have therefore , .
( 3 ) Diyide the summation into two parts at the s-th term .
and for the two parts , we have .
( 4 ) Also , using our Lemmas 1 , 3 and 4 , .
value , , ( 5 ) since is a monotone increasing function of from to Thus , and being is bounded , and similarly is bounded , therefore , by ( 1 ) , is bounded , which proves the theorem .
S4 .
THEOREM and , where , then either of the expressions is a bounded function of .
This theorem may be most easily proved by the same method as that used in the preceding article , bearing in mind that , since is now greater than unity , is positive and monotone increasing .
We prefer , however , to give the method , as it involves certain expressions required in the sequel .
The equations ( 1 ) of the preceding article still hold .
We then proceed as follows:\mdash ; , ( 6 ) where rv .
Prof. W. H. Young .
On the Mode of By Lemma 2 , the expression is positive and monotone decreasing , if .
Hence , by Lemnlas 3 and 4 , the summation which occurs in the expression for is numerically less than This has to be multiplied by , and is therefore evidently bounded , since , by ( 3 ) of the preceding article , is bounded .
Also , as in the discussion of ( 5 ) of the preceding article , , so that this is also bounded .
Thus , ( 7 ) which is the expression above referred to as required in the sequel and leads to our present theorem , since by Lemma 4 , is bounded , so that A similar result holding for , it follows , as in the preceding article , that , which is the ression given in the enunciation of the present theorem , is bounded .
This proves the theorem , when If , we still have ( 6 ) ; but now is monotone increasing ; the term with the summation on the right of ( 6 ) is therefore numerically less than since , in the place of , we have by indeterminate forms , Lt Instead of ( 7 ) , we have now ( ) ( 7 ) whence our result follows as before .
S5 .
THEoltEM Bl.\mdash ; If , and , where either the pressions is a boundedfu nction of .
It is only ecessary to discuss the sine expression .
Approach to Zero of the fficients of a Fourier Series .
459 If , the expression is clearly bounded .
For , denoting it by , we have , since is monotone descending , value o If , we have , ( 1 ) where , ( 2 ) where has the same meaning as in the preceding articles , being the same , with in place of , and .
( 3 ) If is bounded by Theorem Al , and by Theorem A2 is of the order .
Also since is positive and monotone increasing , Thus is bounded .
Similarly is bounded , and therefore , by ( 1 ) , the required lesult follows , when If we have , using the expressions ( 7 ) and , found in S4 , Also Since is positive and monotone ascending , the summation here is numerically less than where the last factor on the right , when ? .
Since then Prof. W. H. Young .
On the Mode of Hence Thus is bounded , whence as before the required result follows , when .
This completes the proof of the theorem .
S6 .
The same method , using only the formula ( 7 ) and not ( 7 ) , gives us the theorem :\mdash ; B2.\mdash ; If , and , where , then either of the expressions cossin ' is bounded .
We foresee now the general theorem:\mdash ; THEOlElI .
, and , where , then eith of the expressions is bounded .
if , the same is true when we change the index in into The proof of this theorem may be left to the student .
S7 .
We now pass to the main theorems of the paper .
THEOREM is the typical coefficient of the Fourier cosine seri.es of an which ?
such that ( i ) a bounded function of in the whole interval of periodicity , and ( ii ) has unique finite nit as then ( Ck ) , when .
We have ( Ck ) where Thus , by Theorem Al , is bounded for all positive values of Let us consider first the part of the integral due to the interval , where is a large positive quantity , independent of , and let us only consider Approach to Zero of the Coefficients of a Fourier Series .
461 values of greater than .
Let the in this interval be denoted , and in the remaining interval by .
We have then Now in the interval the expression is a function of bounded variation , for its total variation is 2 , which by ( i ) is finite .
Let us denote this function of bounded variation by .
Then by ( i ) , , where is independent of and , we may therefore write , where It follows at once that is a bounded function , say numerically less than , where A is independent of and We have , therefore , denoting by , .
Using integration by parts for the evaluation the last integral , this giyes Hence which is as small as we please , choseu sufficiently Next , changing the variable to , and writing , we have Prof. W. H. Young .
On the Mode of Now is easily seen to be a bounded function of , and , as , it approaches , where tu Again , , as , by the hypothesis ( i ) ; and it is a bounded of .
Thus , both factors of the integrand of the in the expression last given for I2 being bounded , we may , when n , integrate term by term , and get which is of the order , and is therefore as small as we please , being chosen sufficiently large .
This proves the theorem .
S8 .
The companion theorem is as follows:\mdash ; THXOREM \mdash ; If the coefficient of the sine ser.ies of an odd function , which such that ( i ) bounded function of the ?
ohole interval of iodicity , and ( ii ) has a limit as th ( Ck ) The proof is the sanie as that of the preceding article , sines taking the In fact W. H. Young , " " On Infinite Integrals involving a Generalisation of the Sine and Cosine Functions , ' ' ' Quart .
Journ vol. 43 , p. 163 See , for instance , my paper on " " Successions of Integrals and Fourier Series ' Proc. ' Ser. 2 , vol. 11 , p. 78 , Cor. 1 ( 1912 ) .
Approach to Zero of the fficients of Fourier Series .
463 place of cosines .
The only essential difference is in the expression for the .
now have tn Wbence which , since is as large as we please , so that the part of this expression involving may be ected , gives Lt This proves the theorem .
S9 .
THEotlElt \mdash ; If is the pical of the Fourier series of an nction which is ( i ) [ xf ] is a function of in the whole interval of periodicity , and ( ii ) has a unique limit as th , as .
We have where By Theorem Bl , is bounded for all positive valnes of as Writing , as before , and for the two parts of the integral due to the intervals and ) , we find , precisely as in the preceding article , that , as approaches a limit or limits as small as we please , being conveniently large .
The treatment of is the same as in the preceding article ; we only have to notice that is now bounded , and that now stands for Prof W. H. Young .
On the Mode of , the meaning attributed to it in the present article .
The itin g is , therefore , where .
Hence by the same process as in the preceding article , Since l ) is as as we please , the part due to the upper limit of tion may be neglected , and we get Lt This proves the theorem .
S 10 .
The companion theorem is the following:\mdash ; THEOREM \mdash ; If the typical coefficient of Fouriof an odd nction which is such that ( i ) [ xf ] is : bounded function of in whole interval ( ii ) limit as ; then The proof is the same as that of the article , cosines being replaced by sines .
The part of the is , as before , ible .
In the calculation of , the part due to the interval , the argument proceeds as before till we come to the calculation of the function .
We then have , whence , by the same reasoning as before , our present result follows , since , when to Zero of the Coefficients of a Series .
465 S11 .
The following theorem might be proved precisely in the same way as those which precede it .
The proof here iven is slightly different , and might equally well have been employed in the other cases .
\mdash ; If is the typical fficijnt of the Fourier series of the even , oohich ?
th ( i ) [ xf ] is ( ii ) 3 !
has q ) then We have , as before , , where ) Thus is bounded for all positive values of 2 ' .
divide the , as before , .
into two parts , and , where , integrating by parts , .
Here the quantity in square brackets may be neglected , for by ( i ) is bounded , and , by what has already been remarked , is bounded , so that the last factor , , determines the value , and this is as small as we please when and taken sufficiently large .
Also , , as therefore , for the function of bounded variation , where the latter integral , , is Prof W. H. Young .
On the Mode of nullerically less than a certain stant multiple of , which is again small as we please .
On the other hand , writing for .
a constant .
This is again as small we please .
Thus , finally , is numerically as small as we please , and being chosen sufficiently large .
changing variable calculating before , using for , we see that is bounded , and has for limiting function , , where tu tu , say .
We thus get Proceeding to the limit , and integrating on the right term by term , we get whence The value of at the upper limit of ration is as small as we please , provided be sufficiently large .
We are therefore only concerned with its value at the origin , which is easily seen to be zero .
This proves the theorem .
Approach to Zero of the Coefficients of Fourier .
467 S12 .
The companion theorem is the THEOREM \mdash ; If is th ( pical coefficient of the Fourier of the function , which is such that ( i ) [ xf ] is ( ii ) 3 !
has a limit As in the earlier theorems , we now only have to change the cosines into sines .
There no alteration of interest till we come to the expression for the function .
We then have tu tu , say .
We thus get As before we need only retain the part of this expression due to the which is This proves the theorem .
S 13 .
The general theorem and its companion theorem are now evident .
We may add that these theorems may be deduced from one another by an inductive process .
have purposely adopted the procedure in the text , at , the processes employed seem to be sufficiently instructive in themselves to be worthy of publication .

