]\gt ; Series the ojFunctio .
217 of the directions in which valency acts .
It is well known that this is a problem which has attracted much at , tention of late in the case of nitrogen .
We are studying the salts of tervalent metals generally , in order to ascertain if the method be applicable to all such elements : it is proposed also to apply the method to the elements of the nitrogen group .
In conclusion , we cannot refrain from calling attention explicitly to the proof that is again given of the service raphy can render to chemistry and of the value of the Barlow-Pope theory as a means of correlating external form with internal molecular structure .
On the Convergence of Certain Series the of a Function .
By Prof. W. H. YOUNG , Sc. D. , F.E.S. ( Received May 16 , \mdash ; Read June 13 , 1912 .
) S1 .
In a paper " " On the Fourier Constants of a Function published in the 'Proceedings ' of this Society , * I showed how certain theorems , previously given by myself , might be employed to obtain formulae the sums of certain series involving the Fourier constants of a function .
In the case in which the function is bounded , it was only proved that the formulae hold wheuever the series converge , or , more generally , when the summation is performed in the Cesaro nlanner ( index unity ) .
In a more recent paper , also published in the 'Proceedings ' of this Society , " " On a Mode of Generating Fourier Series I showed incidentally that the formulae are still applicable when the function has every power of index less than su1llmable , provided only the index , which occurs in the formulae in question , is greater than .
Here again the theorem , as stated , contained the restriction that the summation was to be made in the way .
The main object of the papers in question was , in fact , to explain how certain methods might be employed , and these methods in themselves inadequate for the purpo removing the restriction .
The theorems used involve the ) eneral theory of the integration of the Fourier series of vol. , S4 .
1911 , , vol. 85 , p. 422 .
218 Prof. W. H. Young .
The Convergence of Certain [ May16 , function term-by-term , when multiplied by another function .
The coefficient in the series and considered , is , however , itself the typical Fourier constant both of an odd and an even function , which may be expected to possess special properties bearing on the matter in hand .
A careful scrutiny of these properties has accordingly enabled me to take the step of removing the restriction above explained .
The former of the main results , above stated , may be otherwise expressed by saying that , though the Fourier series of a mded function need not converge , even if the function be continuous , it , and its allied series , will be made to converge , by dividing its coefficients by any power , however small , of the index denoting their respective places in the series .
This affords a convenient necessary test that a given series is the Fourier series of a bounded function .
In the same have a corresponding necessary condition that a Fourier series should have a function whose power is sulnmable for associated function .
It is worth hat in all these statements the function whose Fourier series is considered need not be homogeneous in character .
They still hold if the function be only summable , except in a fixed interval con the origin .
The paper terminates with a comparison of the result of the present paper a function whose power is summable , and one deducible from a theorem given in my paper , " " On the Mode of Generating Fourier Series From the theorem in question it , in fact , may be shown that the series whose general terms are and converge absolutely , when It should finally be remarked that the interesting question arises as to how the coefficient may be replaced in the theorems of the present paper and its predecessors by a coefficient of a more general nature .
This question is too large a one to be dealt with in the present communication .
S 2 .
We with the following general theorem:\mdash ; 1.heorem.\mdash ; The succession of the partial summations of the series nx , , 'is bounded below , whatever fixed positive value be ascribed to .
The is true of the , where , is any monotone descending with zero as limit .
Let denote the partial summation of the latter series , which includes the former as a special case .
Then 1912 .
] Series Involving the Fourier of Function .
219 where Hence Here every term except the first is positive , since , and the q 's form a positive monotone decreasing sequence .
Thus and , therefore , in the interval . .
( 1 ) Again where Thus 2 .
( 2 ) Since each term on the after the first is positive and therefore , in the interval , .
( 3 ) From ( 1 ) and ( 3 ) it follows that , in the interval , since , which proves the theorem .
Cor.\mdash ; The partial summations of the series all positive , provided form a monotone sequence with zero as limit .
The proof then stops at ( 1 ) , which holds for It is easily seen from ( 2 ) that the partial summations are also bounded above in the right-hand half of the interval .
This is , however , obvious from the well-known fact that series of the form as well as those of the form , when the 's form a monotone descending sequence with zero as limit , converge.uniformly in every closed interval not containing the origin .
This follows immediately by the use of Abel 's Lemma .
Since the series is a sine series , it immediately follows that , in the interval , the partial summations of the series are bounded above , and are negative if , , are S3 .
The preceding result permits us to prove the following theorem:\mdash ; Theorem.\mdash ; If is any bounded function , whose Fourier sine constants are then the series 220 Prof W. H. Young .
Convergence of [ May 16 , to rrhpre the of the Moreover , the , afler ?
liach ternb by be integrated ter/ 7b-by-ter7/ b any interval or of points .
It is evidently only necessary to prove the latter statement , from which the former statement imnlediately follows , choosing for the interval of ration the whole interval .
Since the series is an ordinary Fourier series , the integrated series to a Lebesgue integral , *whose integrand is the metion to the Fourier series , that is , since the Fourier series everywhere .
In other words , the Fourier series in question is capable of term-by-term integration .
Again , by the preceding theorem , the partial summations of this Fourier series are bounded below in the closed interval , and bounded above in the closed inlerval Now has proved that , if the partial summations of a series of functions are mded either below or above in any interval , the necessary and sufficient conditions that the series should be capable of term-by-term ation are ( a ) that the series should converge , that the double limit Therefore , denoting the partial summation of the Fourier series , the preceding double limit is zero , and therefore , being a bounded function , The series is therefore , by theorem of Vitali already quoted , integrable term by term , since converges to .
Moleover , by a further theorem of Vitali 's , S this series is absolutely See W. H. Young , " " On the Conditions that a Trigonometrical Series should have the Fourier Form ' Lond. Math. Soc. Proc 1910 , Series 2 , vol. 9 , p. 423 .
G. Vitali , " " Sull ' Iutegrazione per Serie ' Rend .
di Palermo , ' 190 vol. 23 , pp. 137\mdash ; 155 ; see , in particular , S 9 .
The of the theorem used above is that given to it in my paper " " On Semi-integrals and Oscillating Successions of Functions 'Lond .
Math. Soc. Proc 1910 , Series 2 , vol. 9 , pp. 286\mdash ; 324 ; see SS 23 , 24 , and 26 .
An exceptional set of content zero of course makes no difference .
S .
cit. , SS1 , 6 , and 7 .
This theorem , as well as the otl1er , is included in the articles quoted from " " Semi-integrals.\ldquo ; 1912 .
] Series Invohing the Fourier Constants of Functvon .
221 integrable , that is to say , the statement made in the enunciation of the present theorem is true .
This proves the theorem .
S4 .
To prove the statement for the coefficients of the cosine terms , we have to show that vanishes , denoting the n-th partial summation .
For this purpose we shall require the lemma : Lemma.\mdash ; The function sin , as well as , satisfies inequalities of the form , where and depend on.on is an consequence of the fact that , since is periodic and is monotone , the function , that is , is a bounded continuous function , whose greatest and least values are situated in the interval .
S5 .
In with the sine series we proved that the partial summations are all greater than a certain } ative quantity on the right of the origin , and less than a certain positive quantity on the left .
It was unnecessary to examine whether the partial summations were as a mattel of fact positive , as is known*to be the case when is unity .
In dealing with the cosine series I shall again prove a result hoc , and shall not examine whether , as is the case when is unity , the partial summations are all reater than a certain negative finite quantity .
It is sufficient in fact for our purpose to show that they are all greater than a certain negative summable function .
Theorem.\mdash ; The partial nnmaiions of the sljries all greater tharb , where A is a negative constant only Writing for the n-th partial summation of this series , we have 2 siu S where the partial summation has been shown in S2 to be positive , .
Therefore , using the preceding lemma , * Dunham Jackson , " " Ueber eine trigonometrische Sum 'Rend .
di Palermo , ' 1912 , vol. 32 , pp. 257\mdash ; 263 .
See a forthcoming paper by myself , ' On a Certain Series of Fourier presented to the London Mathematical Society .
222 Prof W. H. Young .
The vergence of [ May 16 , which proves the theorem , since S6 .
We now proceed to apply the preceding result to the series whose general term is , where is the typical Fourier cosine constant of a bounded function .
Theorem.\mdash ; If is bounded nction , whose ourier cosine constants , then the series , converges to where is the sum of the Fourier series Moreover the latter series , after multiplying each term by , may integrated -by-term over any interval or set of points .
For , by the preceding theorem , and the integral of this function converges to .
Therefore , as in S3 , and therefore , since is summable , Hence , as in S3 , the required results follow .
S7 .
We now pass to the case in which has its power summable .
We may , if we please , use the lemma of S4 once more , and write S where is and depends only on .
This partial summation is less than a finite quantity independent of , provided , since is of the order .
Hence for such values of and , therefore , for all values of , that is for all values of Sx has its power summable ; moreover , 1912 .
] Series Involving the Fourier of Function .
223 , But sinpe , by So , whence , since* we have Writing , we have Hence , finally , if denote the partial summation of the series , then for all positive values of such that , we have S8 .
From last result it at once follows that , if any whose power is summable , and has the meaning of the preceding statement , provided that Henc these cireurnstances , th converges , being the of in the Fourier series of By a method similar to that of the preceding article we may establish a corresponding result for the cosine coefficients of the Fourier series of function whose power is summable , but I shall not give the proof Indeed , my chief object in writing out the proof just given is to further illustrate the usefulness of the lemma of S4 .
If we make use of a result * F. Riesz , " " Untersuchungen uber Systeme integrierbarer Funktionen S 2 , 'Math .
Ann 1910 , vol. 69 , p. 456 .
W. H. Young , " " On Successions of Integrals and Fourier Series S6 Cor. and S 4 Cor. , 'Lond .
Math. Soc. Proc 1912 , Series 2 , vol. 11 , pp. 57 and 52 .
We only have to use the Schwarz-Riesz inequality .
224 Series Involving the Fourier Constants of Function .
already obtained in a previous paper , *namely that the Fourier series obtained by the process in question , and its allied series , have continuous functions for their associated functions , we may dispense with further proof .
If , in fact , be any quantity greater than , we may perform the process of dividing by in two .
We may firstdivide by , where ; we thus get Fourier series of continuous , and therefore bounded , functions .
Then , dividing by , it follows by SS 3 and 6 that the series and converge , when and are Fourier constants of a function whose power is sumnbable , provided S9 .
It will be noticed that , in the case of a function whose power mable , what we have proved is that , if , the series whose genero)terms are and necessarily converge , we have not proved that the convergence is absolute .
In this connection it may be well to point out that one of the simple consequences of the last result obtained in my paper " " On a Mode of Generating Fourier Series\ldquo ; is that the series whose general are converge absolutely the sam circumstances .
This follows , utilising simultaneously the convergence of the series and S10 .
Another consequence of the same theorem may also be mentioned .
If and now denote the Fourier constants of the integral of a function whose power is , we can assert that the series whose general terms and converge , if any value .
In particular , however small may be , the corresponding Fourier series and allied series both converge uniformly and absolutely .
To prove this we have merely to employ the inequality making use of the converging series and , where and writing and for the general terms of these series , being so chosen that .
* W. H. Young , " " On a Mode of Generating Fourier Series 'Roy .
Soc. Proc 1911 , , vol. 85 , p. 421 .

