]\gt ; [ 489 ] XVII .
On a Class of Identical Relations in the Theory of Eltiptic Functions .
J. W. L. GLAISHER , M.A. , Fellow of Trinity College , Cambridge .
Communicated by JAMES GLAISHER , F.R. Received November 23 , \mdash ; Read January 14 , 1875 . .
THE object of the present paper is to notice certain forms into which the series for the primary elliptic functions admit of being thrown , and to discuss the identical relations to which they give rise .
These latter , it will be shown , .
may be obtained directly by the aid of FOURIER'S theorem , or in a less straightforward manner by ordinary algebra .
henever we have a periodic function of , say , such that , it is well known that we may assume , for all values of .
; and if be even , so that , then , \amp ; c. all vanish ; while if is uneven , so that !
, , \amp ; c. vanish .
If is such that , then we have or according as is even or uneven .
But there is another totally different form in which may generally be exhibited , viz. or ding as or The sine and cosine cannot be so expressed , but the other primary circular functions do admit of this form , as , ex. , in the formulae X[DCCCLXXV .
3 ( in which , after the first term , the series proceed by pairs of terms , so that for every term there is a term .
Thus in general ( although the sine and cosine are , as just mentioned , exceptions ) we shall have , by equating the different forms of , identities such as ex. .
( if is even ) Also , it will be seen in that in certain cases even when is not periodic it may be exhibited in the form , and we shall obtain identities in which the two sides of the equation are non-periodic . .
Before applying these to the elliptic functions , it is convenient to write down at once the following eight formulae , which are to be found in the ' Fundamenta Nova ' ( pp. 101 , \amp ; c and which are all placed together in ' Theory der elliptischen Functionen ' ( Leipzig , 1861 ) , pp. 226 , 227:\mdash ; am ( 1 ) am ( 2 ) am .
( 3 ) am -\amp ; c ( 4 ) am ( 5 ) am -\amp ; c .
( 6 ) , ( 7 ) am -\amp ; c ( 8 ) wherein , of course , In what follows , let , and take so that , and MR. J. W. L. GLAISHER ON THE THEORY OF ELLIPTIC FIJNCTIONS .
491 Also let and , so that .
The process of transformation into the form may be conveniently exhibited on ( 2 ) ; we have am am am which , from . .
( 9 ) The process requires that should be , that is , that should be ; but as both sides of the equation are such that they change sign without being altered in value when is written for , we see that the result obtained is true for all values of Thus we have am for all values of If in ( 10 ) we take , we have .
; or , writing and for and , and therefore for which is at once seen to follow from ( 7 ) , and is given by JACOBI , Fundamenta Nova , ' p. 103 .
It is , of course , easy to deduce ( 9 ) directly from the infinite product for consider , which Taking the logarithm and differentiating , we obtain , after a little reduction , Similarly , from the uneven factor we get ; thus Replace by , that is to say by , and remembering that , we find am Herein write for and for , and we obtain the value on the right-hand of ( 9 ) for am ( ui , that is , for am If the other formulae in the group ( 1 ) to ( 8 ) be transformed in the same way , viz. by use of the identical equations am , am am we obtain the following seven formulae:\mdash ; sin am .
( 11 ) am , .
( 12 ) am , ( 13 ) , ( 14 ) , ( 15 ) .
( 16 ) am ( 17 ) .
J. W. L. GLAISHER ON THE TlfEORY OF ELLIPTIC FUNCTIONS .
It must be remarked that in ( 11 ) and ( 14 ) the number of terms must always be uneven ; this point will be noticed at greater length further on . .
Writing the hyperbolic sine , cosine , \amp ; c. as sinh , cosh , \amp ; c. , these formulae may also be written in a somewhat different form : thus am sech sech sech am and similarly the others .
I do not think it likely that the formulae ( 10 ) to ( 17 ) are new , but I have not succeeded in finding them anywhere .
( ' Die Lehre von den elliptischen Integralen .
Berlin , 1864 , p. 33 ) gives the corresponding forms for \amp ; c. , but he does not allude to the similar expressions for the elliptic functions .
It would , however , in any case have been necessary for the explanation of the rest of this paper to have written down the latter and demonstrated one of them . .
By equating the values of am am , \amp ; c. , as given by ( 1 ) to ( 8 ) and by ( 10 ) to ( 17 ) , we obtain a series of identities of an algebraical character .
which are independent of the notation of elliptic functions ) .
Thus from ( 2 ) and ( 10 ) we have remembering definitions sech sech ( sech ( viz. sech sech This may be written interchanging in the ather more convenient form sech sech ( sech ( sech ( sech ( .
( 18 ) In the same way , by comparing ( 1 ) and ( 11 ) , we find ; ( 19 ) and by comparing ( 3 ) and ( 12 ) , sech secb ( sech ( .
( 20 ) The comparison of ( 4 ) and ( 13 ) gives cosech cosech ( cosech ( cosech ( which , on replacing by , becomes cosech cosech ( cosech ( - .
( 21 ) From ( 5 ) and ( 14 ) we deduce .
( 22 ) The comparison of ( 6 ) and ( 15 ) gives cosech ( cosech ( cosech ( which , on replacing by , becomes cosech cosech ( cosech ( ( 23 ) The comparison of the forms for , ( 7 ) and ( 16 ) , merely gives an equation which , on replacement of by , is identical with that resulting from am , viz. ( 20 ) , while the forms of am , ( 8 ) and ( 17 ) , lead at once to ( 21 ) .
In the expressions on the left-hand side of ( 19 ) and ( 22 ) the number of terms included must be uneven .
It is proper to remark that the formulae for .
can be readily deduced from those for .
; thus ( 18 ) is a consequence of ( 20 ) and ( 23 ) of ( 21 ) .
For ex. .
in ( 20 ) write for , and we have sech sech ( sech ( MR. J. W. L. GLAISHER ON THE THEOBY OF ELLIPTIC FUNCTIONS .
Double this result and subtract ( 20 ) from it , and we have ( 18 ) .
In a similar way ( 23 ) follows from ( 21 ) .
The converse proposition is not .
given thevalue of we cannot deduce the value of .
The results admit of being connected directly with FOURIER'S theorem in the following manner : it is of course well known that every integral of the form or , let us write , gives rise to a series ; and that similal .
lie from thele follows ; and it will now be shown that if is an even function of , and if then and ; ( 25 ) also , that if is an uneven function of , and if then , ( 26 ) and sin .
( 27 ) It is sufficient to prove one of these formulae ; take ( 24 ) .
Since is an even function , .
( which call ) is a periodic function with period , and the right-hand side of ( 24 ) must be of the form Now , being even , But , on taking and , on taking ; thus unless , in which case so that ( 24 ) is proved .
Formula ( 25 ) may be either obtained independently by a similar treatment of the integral or it may be deduced from ( 24 ) by writing therein for rking that by this substitution becomes ) and subtracting ( 24 ) from the double of the equation so formed .
Similar processes apply to ( 26 ) and ( 27 ) .
The method by which the formulae ( 24 ) to ( 27 ) have been just obtained is the same as that by which Sir W. THOMSON Quarterly Journal Mathematics , deduced the theorem ( 28 ) from the integral nx It was after reading Sir W. THOMSON'S paper three or four years ago , that I made a list of all the suitable integrals of the form that were given in Professor DE HAAN'S 'Nouvelles Tables d'Integrales definies ' MR. J. W. L. GLAISHER THE THEORY OF ELLIPTIC FUNCTIONS .
( Leyden , 1867 ) , and deduced therefrom the resulting identities .
The only formulae so obtained which appeared of interest were , in fact , those which are given in the present paper , viz. ( 18 ) to ( 23 ) ; but at the time I was not aware of their connexion with the theory of Elliptic Functions .
It was only recently , after obtaining the values of am \amp ; c. in ( 10 ) to ( 17 ) , that I remarked that the resulting identities were the same as those which I had previously deduced by the aid of Sir W. THOMSON'S principle .
It was shown by CAYLEY at the end of Sir W. THOMSON'S paper that the identity ( 28 ) corresponds to ; . . . . .
( 29 ) and it is singular that all the identities that follow from the method of this section thus appear to correspond either to elliptic or theta-function transformations .
Speaking generally , the only evaluable integrals of the requisite form are derived from cos2 and bx ( including as derivations the corresponding sine formulae ) , of which the former give rise to theta-function relations , and the latter to elliptic-function relations . .
The integrals that produce the formulae ( 18 ) to ( 23 ) , and the manner in which the latter are obtained from them , deserve some attention .
Thus sech whereby ( 18 ) and ( 20 ) follow at once from ( 25 ) and ( 24 ) .
In a similar way we can show that ; but the series obtained from the direct application of this integral would not elgc : and in order to deduce ( 21 ) and ( 23 ) from ( 26 ) and ( 27 ) , it is necessary to express the integral in the form and to make use of the formulae This lenders the process not so satisfactory from a logical point of view ; but practiNDCCCLXXV .
cally our knowledge that .
and .
are the FOURIER'Stheolem equivalents of and would be sufficient to leave no doubt of the accuracy of the formulae so obtained .
In regard to the other two integrals required for ( 19 ) and ( 22 ) , viz. and it is to be observed that , stated in this form , their values are indeterminate ; for the former sin and the latter both of which involve .
But in point of fact for our purpose the of the limit of the integral is not arbitrary , but is to be of the form , the lower limit being replace limits and ) .
Taking then infinite , Similarly cosech and therefore cosech cosech , ( 30 ) whether be even or uneven , if is uneven ; whence the result in ( 19 ) follows directly .
MR. J. W. L. GLAISHER ON THE THEORY OF ELLIPTIC FUNCTIONS .
A similar course of procedure shows that iiom which ( 22 ) may be derived .
In his ' Nouvelles Tables , ' T. 265 , Prof , DE HAAN assigns definite values to the indeterminate integrals and ; and it is noticeable that , if these values be used , they lead to the same results as those just investigated .
The reason is that the integrals in DE HAAN are in effect evaluated on the assumption that ; and if in ( 30 ) we had , in place of the first two terms , viz. written it is clear that the final result would have been the same .
It may be remarked that the identities ( 19 ) and ( 22 ) may be somewhat genelalized by means of the integrals other identities may be derived from ; in which , of course , is to be supposed less than .
The well-known reciprocity of and in the nx dx , nx dx leads to a corresponding reciprocity in the formulae ( 24 ) to ( 27 ) .
Thus from the first of the integrals we deduce that , and .
being both even functions , if then .
; and if then Also , from the second integral , and being uneven , if then ; and if then Applying these formula to the identities ( 18 ) to ( 23 ) , we see that ( 20 ) is its own reciprocal , as also is the case with ( 18 ) , ( 22 ) , and ( 28 ) ; while ( 19 ) and ( 23 ) are reciprocal to one another .
Alth.ough CAUCHY , in his memoir " " Sir les Fonctions reciproques\ldquo ; ( Exercices de Mathematiques , second annee , 1827 ) , has deduced , by means of his calculus of residues , a theorem which is in fact ( 24 ) , he does not appear to have specially remarked the reciprocal chaxacter of the equations .
The application of the formulae presents no difficulty .
For example , comparing ( 18 ) with the first of the second pair , we have sech x sech ; whence the reciprocal formula is sech sech sech which , on replacing and by and respectively , coincides with the original formula ( 18 ) .
On looking at the formulae ( 18 ) to ( 23 ) it appears that although we have transformations for sech sech ( sech ( , cosech cosech ( MR. J. W. L. GLAISHER ON THE THEORY OF ELLIPTIC FUNCTIONS .
cosech , and , there is none for either or th .
; it is therefore interesting to inquire what are the formulae in these cases .
If we write ( 21 ) in the form cosech cosech ( cosech ( and reciprocate it by the third pair of formulae of , we obtain the following result , cosech which apparently ought to be the first of the two formulae sought ; but in point of fact this equation shown actual calculation , is not true .
It seems natural to recur to the integral ( 30 ) , viz. from which , since the first two terms of the right-hand member vanish when is even , we have cosech ; whence ultimately , since cosech ; ( 32 ) but this result is not true either , and for the following reason:\mdash ; Let , and ( infinite ) , and suppose an uneven function of which , when Then so that is periodic ; but so that is not periodic .
Therefore we have no right to assume that between the limits and of can be expressed in the form the true form being We may , however , assume that between the limits and of ( x)\amp ; c. .
; and then cosech We thus find that between the limits and of ( and therefore also between the limits - and of ) cosech cosech , .
( 33 ) the terms on the left-hand side being uneven in number , and such that for every term there is also a term .
If we write for in this formula ( 33 ) we increase the left-hand side by , that is by 2 , while the right-hand side is increased by , that is MR. J. W. L. GLAISHER ON THE THEORY OF ELLIPTIC FUNCTIONS .
by 2 also ; while if we replace by both sides are diminished by 2 ; so that ( 33 ) is true universally for all values of , on the understanding that the left-hand side is viz. that after the first term the series is to proceed by pairs of terms ; so that for every term there is also a term , and the whole number of terms included is uneven .
for the series is the value of which is unity ; and not which is equal to zero .
If we write for , and suppose the terms arranged in pairs from the beginning , we find cosech cosech ( 34 ) as the unity which is introduced on the right-hand side by the is cancelled by the unity on the left-hand side , which results from the supposition that the number of terms is even .
The last equation is , in fact , the relation ( 35 ) Fundamenta Nova , ; for ; so that ( 35 ) becomes of which the left-hand side ; and the identity becomes being and being .
We see from this investigation also that the left-hand side must consist of an even number of pairs of terms .
As ( 35 ) is obtained by diffelentiating logarithmically the formula it follows that ( 34 ) is a form of the identity that results from differentiating logarithmically The formula corresponding to ( 33 ) for the hyperbolic cotangent can be shown , by a process similar to that by which ( 33 ) was itself established , to be ( 36 ) which holds good universally , on the same understanding , with regard to the number and order of the terms , as that which was found requisite for the truth of ( 33 ) . .
I now proceed to show how the identities which have been obtained in the preceding sections by elliptic functions , or by FOURIER'S theorem , can be deduced from the ordinary formulae for the cotangent and cosecant , viz. , ( 37 ) , ( 38 ) by elementary algebra and trigonometry .
Thus to prove ( 18 ) we have ; whence , by subtraction , MR. J. W. L. GLAISflER ON THE THEORY OF ELLIPTIC Now ; whence - ; on replacing and by and , we obtain the formula .
Now from ( 38 ) \mdash ; \amp ; c. ; whence , writing xi for sech and sech sech ( sech ( \mdash ; \amp ; c Adding these expressions together in columns , and transforming each column ; ( 39 ) , we find sech sech ( sech ( sech ( sech ( . . . . . . . .
, which , after summation of the columns , sech sech sech , which is the identity ( 18 ) , that was in deduced from the formula am am ( ui , and in from the integral sech x sech .
The other identities , ( 19 ) to ( 23 ) , admit of being demonstrated in exactly the same way .
The formulae of transformation , similar to ( 39 ) , that are required are , , , the first resulting from , and the other two from .
The following expressions , which are analogous to that used for sech in the last section , are also needed:\mdash ; cosech all of which follow from ( 37 ) and ( 38 ) at once in the same way as that by which the formula for sech was obtained .
Only one point calls for notice in these demonstrations , viz. in the proof of ( 20 ) we find sech seoh ( sech ( .
J. W. L. GLAISHER ON THE THEORY OF ELLIPTIC FUNCTIONS .
and in order to obtain the correct result we must replace the indeterminate series , , by .
Cases in which the method gives results absolutely erroneous will be noticed in It will have been seen that the process of consists in ) each term of the original series by terms ( infinite ) , and therefore the original expression itself by terms .
Each series of terms formed by adding the vertical columns is transformed into another series of terms , so that we thus replace the first scheme of terms by a second scheme of terms , which latter system , being such that the columns admit of being summed as ordinary geometrical progressions , gives the second side of the identity to be proved . .
A question that naturally arises is to inquire what are the results which should obtain if , instead of using ( 39 ) and the similar formulae for the conversion of one series into another , we were to replace at once these series by their finite summations- , .
instead of ( 39 ) to take .
We thus find sech sech ( sech ( , ( 40 ) while the left-hand side also sech ( 41 ) from ( 18 ) .
Although ( 40 ) is the identity which we have absolutely proved , we may regard the fresh identity as being that which follows from ( 40 ) and ( 41 ) , viz. ( writing for the moment in place of , and in place of ) .
( 42 ) This result follows immediately from another form of the series for the cosine amplitude ; for on p. 113 of his ' Lehre von den elliptischen Integralen und den ThetaFunctionen ' ( Berlin , 1864 ) , SCHELLBACH finds .
( 43 ) We easily see that am ( 44 ) and the comparison of ( 43 ) and ( 44 ) at once gives ( 42 ) , since .
The result ( 43 ) is also given in the ' Fundamenta Nova , ' p. 102 .
It thus appears that by absolutely summing , instead of transforming , in the process of we obtain the series of formulae which SCHELLBACH has given on pp. 113 , 114 of his treatise , so that all the formulae and identities which arise from the transformations of the elliptic functions are algebraically exhibited by the method of .
It is unnecessary to write down the series of identities analogous to ( 42 ) for the other functions , as they can be easily derived as above from the values in SCHELLBACH .
It may be remarked that ( 40 ) is a transformation of am am , but ( 42 ) is merely a transformation of am am .
If , therefore , we perform the process of in reverse order .
starting with the trigonometrical side of the identity to be proved , sum the rows instead of transforming them ) we obtain ( 42 ) at once .
It appears at first sight as if SCHELLBACH'S formula am ( 45 ) gave rise to another formula for the cosine amplitude , by writing xi for and changing the modulus from to ; but this , in fact , merely gives an expression already obtained ; for the right-hand side of ( 45 ) , on writing xi for and for , becomes sech which sech x sech x sech sech ( } .
Formulae such as ( 45 ) are the nearest approach I have met with to those numbered ( 10 ) to ( 17 ) and the other expressions at the end of ; but ( besides that an imaginary transformation is required to reduce them to these forms ) they do not put in evidence the periodicity of the functions . .
It is perhaps desirable to place side by side , for convenience of comparison , all the different forms into which one of the functions , the cosine amplitude , has now been thrown .
Writing , as before , MR. J. W. L. GLAISHER ON THE THEORY OF ELLIPTIC FUNCTIONS .
am sech zsech ( sech ( sech z being any four quantities subject to the relations whence x , the identities sech sech ( sech ( sech ( sech ( sech x sech x\amp ; c. Another form will also be given in the next section .
It is scarcely necessary to that corresponding formulae and identities exist for sin am am am am \amp ; c. .
The identities ( 18 ) to ( 23 ) can also be proved by trigonometry in another distinct manner , by starting from the trigonometrical sides of the equations .
Thus .
for ( 18 ) , from the formula sech we have ( writing for for brevity ) sech sech sech .
; whence sech sech \mdash ; \amp ; c. , which , as shown in sech xsech ( We thus in the course of the proof obtain another form for sech sech ( sech ( , viz. ; ( 46 ) whence , in addition to the forms for am in , we have am This method of proof is not so interesting as that of , both because the formulae required cannot be obtained in so mentary a manner , and also because the identities ( 18 ) to ( 23 ) are not so directly verified , as their right-hand members are shown to be equal to expressions such as ( 46 ) , which themselves need some transformation before they assume the desired forms .
The formula .
J. W. L. GLAISHER ON TJfE THEORY OF ELLIPTIC A'UNCTIONS .
which was required in the verification , is best obtained by deducing it from the wellknown theorem from which , by writing for and for , dividing the equation so obtained by 4 , and subtracting it from ( 47 ) , we find It is to be noticed is only true if lies between and .
This may be regarded as a consequence of the fact that ( 47 ) only holds good when is positive and less than ; but the necessity for the condition is also evident from the process of verification by ordinary algebra .
Thus the expression in ( 46 ) sech xsech ( sech ( wherein we see that to justify the summations of , and as ordinary geometrical ressions we must suppose to be positive and less than Also since sech sech ( .
is periodic , while the expression in ( 46 ) is not so , we see that the equality will not hold good beyond these limits .
I have worked out the corresponding proofs of the other five identities ( 19 ) to ) ) in the same way , but none of them call for any special remark .
The process is not in all cases exactly similar , as , ex. , in deducing ( 19 ) from , we find ; whence , since the first series on the right hand side , when is positive and less than and \mdash ; \amp ; c. The other transformations to which the method of this section leads are cosech cosech ( cosech ( cosech cosech ( sech sech sech ) which can be readily verified by ordinary algebra in the manner explained above .
In all these identities must be positive and less than .
It only remains to apply the methods of and 15 to the identities ( 33 ) and ( 36 ) , which differ from the others by relating to non-periodic functions .
Employing the method of , we have .
; whence .
( 48 ) MR. J. W. L. GLAISHER ON THE THEORY OF ELLIPTIC FUNCTIONS .
whereas the true equation is .
( 49 ) It is well known that if an infinite system of series be summed by rows and by lumns , the results need not necessarily be the same ; but the above is a striking instance of such a disagreement .
We should be prepared for some ambiguity from the observation that although the value of the left-hand side is liable to a change of a unit according as the number of terms retained is even or uneven , yet in the process of transformation no condition whatever with regard to the number of terms in the columns is , or can be , imposed ; but we should scarcely expect to obtain an absolutely erroneous result by an apparently definite process .
If the same method be applied to the hyperbolic cotangent , we have and finally , ( 50 ) which is also erroneous , the term being omitted on the -hand side .
The method of , however , yields correct results , for cosech cosech .
; whence cosech cosech 2 ( by use of the formula ) -\amp ; c. ; and therefore which is the true formula .
In the same way , since we find that which is correct , and agrees with ( 36 ) .
It is of course easy to assure one 's self that ( 48 ) cannot be true ; for , taking for simplicity , and differentiating with regard to or ; and it is evident that if we take and we should have a positive quantity quated to a negative I thought it of interest to actually verify numerically the truth of the and ( 36 ) in one or two cases .
Workin with seven-figure logarithms , and taking , I found that each side of ( 33 ) was , and for that each side was ; while for each side of ( 36 ) was , and for each side was ; placing beyond doubt the correctness of ( 003 ) and ( 36 ) .
It is a characteristic property of the identities noticed in this paper that in all cases the series on both sides are convergent whatever may be the values of and .
For the actual ] ation of the elliptic functions the formulae ( 10 ) to ( 17 ) would be preferable to ( 1 ) to ( 8 ) if the angle of the modulus was very near to , so that was MR. J. W. L. GLAISHER ON THE THEORY OF ELLIPTIC FUNCTIONS .
nearly equal to unity ; but as probably the theta functions ( or their tlansformations as in ( 28 ) ) would always afford the best means of actually calculating the elliptic functions , I have not investigated whether ( 10 ) to ( 17 ) would present any advantages over the formulae which result directly from the change of modulus from to , as ex. .
the formula at the beginning of , viz. am .
There are two well-marked classes of identities that are derived from the theory of elliptic functions , viz. pure algebraical identities , in which only one single letter is involved , as ex. and what may for the sake of distinction be called transcendental identities , viz. in which a function of is equated to a function of .
To this latter class the chief identities discussed in this memoir ; and if special values be to such that the left-hand member of the equation is of the same function of that the right-hand member is of , or , in other words , if the identity is of the form , where such a result is usually very interesting .
The best known identity of this class is ; but there is another elegant formula of the same kind to which ABEL has attention Euvres , , viz. . .
, ( 52 ) the relation between and being of course It seems probable that all the transcendental formulae of this latter class can ) deduced from the trigonometrical identities in and at the beginning of by elementary methods , without the introduction of elliptic-function formulae ; and it is of some interest to verify ( 52 ) in this tarting from the formula ( 23 ) , which may be written we have , on differentiation with regard to MR. J. W. L. GLAISHER ON THE TflEORY OF ELLIPTIO FUNCTIONS .
Put , and while so that ; whence , on integration with regard to const .
viz. . .
, and , as is seen by putting ; so that ( 52 ) is established .
The other identity ( 51 ) , or rather the generalization of it , ( 53 ) ( which is much more difficult to prove by elementary methods than any of the identities discussed in this paper ) , I deduced by algebraical processes from the equation in viz. from ( 54 ) in the Philosophical agazine for June 1874 ( ser. 4 , vol. xlvii .
p. 437 et seq. ) ; but it perhaps is worth while to note here what is the most natural way of obtaining it from MR. J. W. L. GLAISHER ON THE THEORY OF ELLIPTIC FUNCTIONS .
( 54 ) , viz. by help of the theorems , ( 55 ) ; ( 56 ) whence , operating on ( 54 ) with , and making , we have at once wnich is ( 53 ) if we take Of the two lemmas ( 55 ) ) ) the truth of the second is seen at once , for ; and ( 55 ) is easily established , since being put after the performance of the differentiations , But the investigation is not elementary ; and if we assume a of the integral we may as well apply it directly to prove ( 53 ) by FOURIER'S theorem as explained in or employ it as SCHELLBACH has done ( ' Die Lehre von den elliptischen Integralen \amp ; c 1864 , p. 30 ) .
It does not seem to be easy to establish ( 55 ) without the aid of an integral ; for , expanding in ascending powers of , we have to show that when ; and , taking the first term only , although we see at once yet which is divergent , and cannot apparently by any simple method be so transformed that its ralue when may be evident , ithout the intervention of an integral .
rrhus the method depending upon ( 55 ) , though more direct , is not so elementary as that described in the Philosophical Magazine .
MR. J. W. L. GLAISlfER ON THE THEORY OF ELLIPTIC FUNCTIONB .
It is curious that all the formulae of the form .series of sines or cosines which can be obtained by definite integrals , and which possess any interest , should be in reality elliptic-function identities .
Of course every result that can be derived from these identities by differentiation , by multiplication by a factor and integration , \amp ; c. , can as a rule be obtained directly from an integral , which integral itself would arise from a similar treatment of the original integral .
This is true of the identities in the Philosophical Magazine , ser. 4 , vol. xlii .
pp. 422 et seq. ( December 1871 ) ; and , for example , such an integral as erfc erfc ( 57 ) ( where erfc ) would give rise to identities which , however , could be deduced from ( 28 ) and ( 53 ) by a similar process to that by which ( 57 ) can be derived from bx

