]\gt ; .
On some Definite Integrals , and a New AVIethod of a Function of Spherical -ordinates to a of Spherical By ARTHUR cHUSTE R , F. , \mdash ; Read Jnne S S1 . .
THE following investigation deals with some inteo r are useful when it is desired to express : function angular riables means of a series of spherical surface harmonics .
An concerning rals leads to a method which considerably reduces the arithmetical laboul in roductions , and secures in practice the advantage of obtaining the numerical values of the coefficients of lower degrees independently of those of higher degrees .
The zonal harmonic of degree is denoted by and as usual by S : S. The tesseral harmonic of and type ] loted by , where In these equations represents an angle measured on a sphere from a point as pole : and .
The longitude measured from standard meridian is denoted name given by HEINE , and translated by TODHUNTER as " " ctions oftheFirstKind , \ldquo ; umbersome factor erhaps appear tppropriatethese functious " " Tesseral Functions The tesseral function converted into atesseral harmonic by because it is only the factor which gives , function its tesseral\ldquo ; character , but it is short and suggests at one the function it denotes .
: The present investigation deals in grsat part with the definite rals , taken over ths surface of a sphere of nnit radius , ) the of two tesseral functions ( 824 .
) SSOR A. ON SOME INTEGRALS , the product tesseral nction and or , the tesseral functions being all referred to the same axis .
I shall denote , as usual , the factorial product of the numbers up to by !
, but have found it necessary to introduce a separate notation for the products of successive even or successive odd numbers .
I consequently define ! .
Starting with 1 , 2 ! !
it follows that , positive values of where the last factor is either 1 or 2 , according as is odd or even .
We may extend the definition to negative of the argument , for the successive substitution of , into the first equation , gives and generally if is negative and odd , , the original equation gives ! !
for ? ! !
and similarly for all negative and even values of ! !
is infinite .
The ratio of two of these factorials of negative numbers is , however , finite , for it is easily shown that if and be two negative numbers , whether even or odd , One of the advantages of a separate notation for what may be called the " " alternate\ldquo ; or " " double\ldquo ; factorial , is due to the fact that it often inconvenlence of for odd and even numbers .
S2 .
of It is convenient to collect together some equations which will be required .
Most of these equations will be found already in previous writings , such as HEINE'S eatise or ADAMS ' " " Researches in erresl i\ldquo ; As onal , it is only necessary to quote the well-known relations : ( A ) , ( B ) , . . . . . . . . .
( D ) , . . . . . .
( ) , ( F ) , ( G ) , ( H ) .
As special cases of ( G ) and ( H ) we may put equal successively to , and .
The following equations are thus obtained : , , , , , .
The formula ( A ) is well known and may be obtained by differentiations of ( 1 ) , substituting in the result an equation derived differentiations of ( 2 ) .
184 A. SObiE DEFINITE Equation ( B ) is the result of differentiations of ( 2 ) , and multiplication by .
( C ) may be proved by combining with the fundamental equation for zonal harmonics .
The latter leads directly to .
which we derive ( by equating the two expressions on the right-hand sides ) : or , after multiplication by be now ) stituted f and , the equation ( C ) is obtained .
If written for and for in ( C ) we may combine the two equations , so as to give ( D ) and ( E ) .
( F ) is an immportant relation obtained from ( B ) and ( C ) .
The Inulae ( and ( H ) are easily derived by direct differeIltiation and a few simple transformations .
S 3 .
If the equation is integrated with respect to between the and , the left-hand side vanishes at both limits : hence , after changing from and , by applying the same process to the right-hand side , Repetition of the same proceeding will ultimately lead for even values of to WITH fION TO But where the special case , for which the integral vanishes , is excluded .
Hence , if be even .
; , if be odd ; and finally , if and be even ; , if be even and odd .
For odd values of , it is more convenient to use formula which leads to : : By repetition of the same process , we get for even values of ultimately S : S. As the expression under the integral of the side vanishes , the value of the integral on the left-hand side is zero .
When / is odd we are ultimately led to But ' .
Collecting the results , we generally , if be even , where is equal to 2 or as be even or odd .
When is odd , the integral vanishes .
The results of this paragraph allow us to represent a quantity which is constant over a sphere , in terms of a series of tesseral functions which are all of the same type , the type being arbitra1y .
It follows that we may express and in terms of a series of tesseral harmonics of type V0L .
PROFESSOR A. SCHUSTER ON LQOMI DEFINITE INTEGRALS , To find the values of the coefficients of such a series we put integrating over unit sphere , aftel multiplying as usual by , we find Substituting on the hand side the integral found aboye we obtain where is 2 or according as and are botb even or both odd .
We obtain , for instance , in this way for S4 .
If we put in equation ( G ) , and after changing from to integrate both sides , we obtain : If be odd , a continuation of this process reduces the degree of the tesseral function on the right-hand side until it becomes smaller than , and this will happen before any of the factors in the denominator become zero , so that in that case the on the left-hand side is zero .
If be even , the integral on the right-hand side ultimately becomes where is 2 or as is even or odd .
The integral to be determined now becomes ( if be ev en , if WITH TO RICAL HARMONIC 18.7 A little care is necessary in the interpretation of the square bracket on the righthand side .
All the factors of the product contained in it may be positive or negative , but when is intermediate between and , some may be positive and some negative .
In the latter case , one of the factors will be zero when is even , and in that case the integral on the right-hand side is zero .
The above expression does not include the special case .
By extending the notation of double factorials : to negative numbers as defined in S 1 , we may also write , including the cases when , or both and , are negative , or when if be even , : S. if be odd .
The integral reduces to the one just determined with the help of ; equation ( A ) .
We thus find : .
if be odd , S4 the last equation becomes if be even .
The factor takes , as before , the value 2 or according as is even or odd .
For the special case , the tesseral harmonics reduce to zonal harmonics , and if be odd .
If be even and , the fraction on the right-hand side is zero , because in that case ! !
, and the numerator remains finite , but if , the value of remains finite whether be even or odd , and , in that- case , the right-handside m be written ( avoiding negative arguments of the factorials ) : : If be odd and , we may transform the negative factorial and write the value of the integral \mdash ; .
Similarly we obtain ' if be even , otherwise the integral is zero .
This includes , as particular cases , if be if be odd and if ?
Dr. W. D. NIVEN ( ' Phil. Trans vol. ( 1879 I p. 879 ) has already obtained an expression for the integral .
His results are identical with the above , when allowing for the difference in the notation and after correcting an ious misprint in the equation marked ( 16 ) on p. 388 .
S 5 .
If we write don'n the differential coefficients of by means of the equation .
, and repeat the same process times , we obtain a series beginning with There being no terni containing a zonal harmonic of higher degree than , we conclude that the above integral vanishes when , and that when If we may transform the inteo.rale , as follows : After partial integrations , in which the first term always vanishes at both limits , the integral becomes WITH APPLICATION TO SPHERICAL IONIC ANALYSIS .
189 The on the right-hand side has been found in the last paragraph .
Writing and , we note that is necessarily as , and that positive and even .
If , further , be even must be even also , because the difference between these quantities is .
We may now write the value of the integral if and is even , , in all other cases .
If , the aboye reduces to $ * S C ) .
: Before discussing the general integral of the product of two tesseral functions , it is convenient to obtain the solution in a few special cases .
When the type of the two .
tesserals differs by 2 , we may transform the integrals as follows : By the application of RODRIGUEZ'S theorem the right-hand side reduces to Integrating each term partially times we arriye at the equation If be odd , or if , the integrals on the right-hand side vanish .
If and even , we must substitute , , as founl above , More particularly when we have the integrals , , if is even and in all other cases , : The result that has zero value whenever can be extended to the more general integral provided then .
To show this we need only consider the series , , where the coefficients may be determined from the integra]s found above .
Multiplying both sides with and integrating , all the terms vanish when and hence in that case .
From we may proceed to and so on .
To obtain the integral , we use formula ( F ) , multiplying both sides by and integrating .
We may without loss of generality take to be smaller than Utilising the result which has just been obtained , one integral on each side drops out and the equation becomes for even values of if and even if odd .
WITH APPLICATION TO .
191 Reverting now to equation and multiplyin both sides with , we obtain @ ; : The integral on the right-hand side may again be transformed in the same : manner , changing from to .
This leads to If the same process be continued until the integral on the right-hand side becomes , the remain , terms on that side will consist of a series in arithmetical progression .
Adding this , we find , if bc even and If is odd or if the integral is zero .
S 7 .
The integral of the product of a tesseral function and of a power of having been obtained in S 4 , the above integrals are found by expanding the tesseral function or the zonal harmonic in a series by powers of From the expression of a tesseral function of degree and type , as it is generally given , viz. : it is seen by writing and expanding the binomials , that the term of lowest power will be , and that if be even , the term of highest power is The coefficients of the series , which are given by , are most easily obtained by going back to the differential equation : Substituting , and changing variables , this becomes 1 ? .
If the series . .
satisfies this differential equation , the coefficients and must connected by the relation is seen by substitution .
The first coefficient is determined by the fact that it is equal to the value of when .
This quantity is known to be equal to 1 The other coefficients may now be determined in terms of this , and we find in this way for the series The series breaks off with the term , when is an even number , but also holds if is odd .
The factor of in the series reduces to WITH APPLICATION TO RICAL H When is an odd nmnber it will be more convenient to use a different series .
Writing , and changing the variable to as before , the differeIltial equation becomes : and from this we find for the series .
The factor of in the series is found to be In considering the integral , without loss of generality , take to be smaller than .
If of the two quantities and , one is odd and the ?
other even , the integral vanishes .
If and are both even numbers , we may express in terms of a series of powers of and obtain in this way : I The symbol is intended to express that takes successively the values , . .
, leaving out the alternate .
The constant is equal to 2 or according as is even or odd .
If and are both odd , we find in the same way : .
' The value of is the same as before .
Our result shows that when is even , the integral vanishes when For when is even , will also be eyen , for all values of , and V0L .
OR .
SCHUSTER ON SOME DEFINITE TEGRALS , it will be negative because is the smallest value of and is larger than ! !
must therefore be zero , and all the terms of the sum vanish unless there is a negative even factorial in the numerator .
The only factorial which can become negative is ! !
, but being the highest value of , it will not be negative when .
For the special case , , all the terms of the sum vanish except the last , for which .
We have in that case and the integral reduces to ' as already found in S 6 .
The same expression is found when is odd .
Another and generally more convenient expression for is found by expanding instead of in terms of powers of .
We thus obtain , when and are both even , and when and are both odd , Writing we may put the into the following form : , if even , even , and excluding where where , if is even , and , if is odd , WITH APPLICATION TO SPHERICAL HARMONIC ANALYSIS .
the last term of the series is but the series breaks off before the end , if is even and owing to the actor A vanishing .
The number of terms in that case is Similarly when and are both odd 2 , where where .
and the coefficients A have the same value as before .
The last term of the series 1s givesTo.only p in the previous investigation .
This ' 2.4.2.4 . . .
if and is even , and @ , odd .
Thus , if be even .
and if be odd : \amp ; .
SCIIUSTER ON SObIE DEFINITE INTEGRALS , The equations do not hold for as has already been explained .
S 8 .
These important integrals are obtained in two different forms according as or the trigonometrical functions are expressed in terms of the powers of .
In the former case the problem reduces itself to the evaluation of ralsa of the form and It may easily be proved that O. when p is odd , when is even .
when f ) is even , when is odd .
is when 1 ' is odd .
when is odd , when is even .
In these equations is equal to 2 or according as is even or odd .
From the results of S 7 we may now write , if be even and odd , WITH APPLICATION TO ICAL HIONIC ANALYSIS . .
When and are both even the integral vanishes .
If be odd and even , Si It will be noticed that the integral is zero henever ( is even .
Hence if is expressed in terms of tesseral functions , 1 ' being cycn , odd values of occur when is even , and only evcn values is an odd number .
The reverse .
will be the case when is odd .
We find similarly , if 21 \mdash ; is odd and l ) odd , : : .
1 ' and when and are both even , ; when is odd the rrals vanish for all values of We may now prove the theorem which has been mentioned in the first paragraph .
If and are both even , is zero , unless is odd .
In that case the quantity which occurs in the denominator will be even , as takes up 198 A. SCHUSTER ON SOME DEFINITE INTEGRALS , can successively the values , \amp ; c. The highest value which can take , and this is when .
It that when and is even .
The restriction that is even is not necessary , because when is odd will be even and the factorial in the denominator , according to the above equations , is !
This again will be infinitely large whenever it is negative , or taking the highest value of , which is now , whenever We prove exactly in the same way that when and is odd .
The value of the integrals in the general case may conveniently be expressed by two series and 2 defined by Then if be odd and is even .
if ' be even and is odd .
whenever is even .
if be even and is even .
if be odd and is odd .
whenever is odd .
constant c takes the value 2 or , according as is even or odd .
The series and break off in all cases ultimately when one of the factors of the products .
or .
become zero .
When is odd , the series may break off before it has reached its full number of WITH APPLICATION TO HARMONIC ANALYSIS .
199 terms , owing to one of the factors the product becoming zexo .
In the special case the value of reduces to The above equations are inconvenient for numerical calculations , unless be small , or be even and small .
We obtain quite different and generally more .
Let convenient expressions , if we begin by expressing or in terms of a series proceeding by powers of and generally : S : : b@ Then the well-known expressions for the trigonometrical functions of the multiples of an angle may be written : If be even : \mdash ; . . .
; and if be odd : \mdash ; . . .
\mdash ; . .
We derive this , being even : and l ) being odd : Hence from the results of S4 , if be even and odd : .
The constant is 2 or , according as is odd or even .
If be odd and even , and with the same meaning of : Similarly if and are both even , and if and be both odd , In the two last equations is equal to 2 when is even and equal to when is odd .
In the interpretation of the summation a little care is required when and or ?
are negative , as the alternate factorials are infinite for even values of the argument .
To put the equations into a form useful for numerical calculations , it is convenient to designate by separate letters the following six series:\mdash ; . .
\mdash ; . . .
WITH APPLICATION TO ONIC \mdash ; . . .
\mdash ; . .
, . .
: ' . .
: ' The series are all continued until they break off , which may happen either because the factors , or one of the other factors takes zero value .
With the help of the functions just defined , the integrals now take the form if be even , odd , even .
if be even , even , odd .
if be odd , odd , odd .
if be odd , even , even .
if be even .
if be even , if be even , odd , ?
odd .
, if be odd , even , odd .
if be odd , odd , } even .
if be odd .
By these last equations with the results obtained when the same rals vere expressed in terms of the series and , we may derive the following special values : if be even , odd , even and , , , , even , odd , , , , odd , , , , , , , , , odd , even , , The two scrics marked and had to be introduced because the series for break off as soon as one of the factors becomes zero .
But in the summation which gave rise to , .
: ! !
begins to be when , and hence the terms of the series will drop out until is negative , i. e. , until .
For higher values of \mdash ; will be finite .
There is , therefore , a second portion of the series not included under ] , and it is this second portion which appears as S 9 .
ctween M ions exist between the series which serve to express the integrals and , and these relations are important for the calculation of numcrical tables .
Starting from the equation may multiply or ) cTlld i on both sides .
Expressing the result means of the series , we obtain the following set of equations : ' .
WITH APPLICATION TO ANAL Y Other relations are obtained as follows : and making use of , S2 , If be even and odd , the integral on the left-hand side depends on , while that on the right-hand side depends on and .
For even and even , the Treating the integral in the same way , and collecting the results , we left-hand side may be expressed by , and the -hand side and find : We may also connect together the different values which the same function assumes for different values of Starting from the identity we find , transforming the right-hand side with the help of equation ( A ) , S 2 , the integral to be Three other equations may be obtained corresponding to this ; by substituting the for the sign in the left-hand sign , or by changing the sine function into the cosine function .
Each of these three equations gives two relations according as is even odd .
The following eight relations are thus obtained : A. CHUSTElt O RALS , ' , , , , , , , .
S 10 .
of the Four Series .
For calculation of the numerical values of the series , the special case was taken as .-point .
In this particular case the well-known which lolds when is odd and , gives when is eyen and ' odd .
The values of were calculated by this equation for all values of 1 ) np to and including , and for all values of up to and .
The first of the equations , S9 , was then applied , and substituting , and , the value of was calculated .
Similarly putting \amp ; c. , the values of for even values of were all tabulated .
By successive steps I similarly fOund , \amp ; c. , up TO tICAL IONIC A to The last value so obtained , and some of the intermediate ones , were calculated independently from the series defining , numerical errors being easily detected in this way .
The values of were dealt in a similar manner .
As a final check , the values of , and were obtained and by means of the second and fourth of the eqnations ( L ) .
S11 .
the Previons Results to the Exprcssion of by of rmonics .
The results obtained in this investigation lead to a new culating the coefficients of the series of spherical rmonics which epresent a function of spherical co-ordinates .
If the of this function yiven for points of sphere , coefficients of the term involving is known to ) solely on the integral taken ovel the of the sphere .
This theoretically perfect proceeding .
has to be modified in practice the values of are known only at definite pomts , so that the mediate p have to be evaluated a process .
of inerpolation on the sition that is continuous everywhere .
All methods lyhich have been applied so , suffer from the serious enience .
that the method of least squares is applied in such a way as to make the value of the coefficients of lower depend on the nulnbel .
of terms which are taken into account .
That is to say , the not independent of each other as they ought to be .
A. method suggested by F. E. , which is free this defect , introduces .
other complications , and has never , as .
as I know , been applied in practice .
The theorem of S8 offers a simple solution of the practical difficulties , and reduces the whole problem to an expansion by means of FOURIER'S analysis , which can be carried out either by the known process of calculation , or by mechanical means .
Let be expressed , in the first place , for different circles of latitude as a series proceeding by cosines and sines of the longitude and of its first step is common to methods .
The result may be expressed lnbolicall 1 here are functions of the colatitude .
If their values are known at equidistant circles of latitude , and at the poles , we may determine the coefficients and , which satisfy the equations If we giye to the successive , we shall have * equations to determine ients .
The solution is conducted according to well-known rules , making use of the proposition that , if and are lnteger numbers smaller than If and are equal to each other the sum on the right-hand side is equal to The factor of the first and last term should be noted .
If we designate by , the value of for , and at the successive circles of latitude , it follows that The condition under which the coefficients are determined is that the coefficients of the Fourier series than are zero .
It will be noted that , even if we suppose some of the lower coefficients to vanish , the equations will still give those values for the remaining coefficients which , according to the method of least squares , fit in best with the assumed values of , but that in the calculations half weight only is given to the values at the two poles .
[ The above method of ining the coefficients , which is identical with that in common use , when the range of is , is convenient whenever the function to be analysed is known at every point of the sphere , so that the coefficients and may be determined for a sufficient number of equidistant circles of latitude .
But other methods are available , and hence the process of obtaining the coefficients of the series of spherical harmonics which I am endeavouring to explain , is not restricted to cases where the original function is known everywhere ; provided it is continuous , as well as its derivatives over the surface of the sphere .
Graphical interpolation may be employed to determine the function at unknown points with sufficient accuracy , and there are several good mechanical devices in existence , by means of which at any rate the lowest and most important coefficients of the Fourier series may be found .
It will , in some cases , materially help this process of interpolation if it is remembered that continuity at the pole involves the vanishing of all the values of and except .\mdash ; August 2 , 1902 .
] Having calculated the coefficients , the reduction to spherical harmonics is made , by substitution of where the values of may be calculated and tabulated once for By the theorem of S8 , all coefficients vanish up to and inclusive of if be odd , so that WITH APPLICATION TO ICAL HIONIC ANALYSIS .
the spherical harmonic of degree will only depend on the Fourier coefficients which is equal to or smaller than .
When is even we secure the same advantage by developing and in terms of the sine functions , and we write in that case : sin and @. .
: To calculate finally a coefficient such as we proceed in the usual way , thus : K- , Therefore when is even and is even S , when is odd and is even .
Similarly when is even and is odd when is odd and is odd .
Combining equations , we find that if denote the coefficient of in the development of , and the coeffi-cient of , when i , s evcn odd is odd even is even and even and odd : when is odd and even and even .
To obtain , substitute and for and The coefficient is introduced because the quantity here designated by is not uniformly accepted as the standard form for a spherical harmonic .
For some purposes ' 208 PROFESSOlt A. SCHUSTER ON SOME DEFINITE INTEGRALS , it is more convenient to take the form of GAUSS , for which THOMSON and TAIT have adopted the symbol , and which is connected with by the relation Another constant , which has beeIl applied by ADOLF SCHMIDT , is based on the consideration that for numerical work it is inconvenient to deal with functions the average values of which differ considerably from each other .
This author therefore takes a function as basis of calculation , and determines , so that the average value of the square of over a of unit radius is equal to unity .
The function defined in way is connected with and by .
hcre is equal to 1 for and equal to 2 for all other values of If a function is to be expanded in terms of , we must therefore write For numerical work the intJoduction of in place of possesses undoubted advantages .
Although at first I was eluctant to adopt the additional complication due to the introduction of a square root and the addition of yet another function to those given by previous writers , I found that the inconvenience of tabulating values differing considerably in magnitude from each other was very great , and I therefore felt myself almost compelled during the course of the investigation to adopt function as above defined .
now ) stitute t the equations for , and , and write : where and if By substitution of the values of , \amp ; c. , we may also more simply , , , , , 7 ' even , , , , , even , eyen , , , odd , , , , , odd , , , , , even , with similar equations for ) and being replaced by and The values of and are given in Tables .\mdash ; VIII .
and their logarithms in Tables IX.\mdash ; XII .
, for values of and up to 12 inclusive .
By means of these tables we may , for stance , write at once the coefficients as far as the third degree as follows : VOL. CC.\mdash ; A. 2 : The calculation of the ten coefficients up to the third degree involves therefore the determination of sixteen Fourier coefficients .
There is no coefficient of zero type , and there are therefore only six coefficients up to the third degree which depend upon twelve Fourier coefficients .
In general there will be coefficients , and coefficients up to and including the degree , giving a total of coefficients in the series of spherical harmonics .
To determine these completely , it would be necessary to calculate Fourier coefficients for the quantities .
and for the quantities , giving a total of Fourier coefficients .
The tables accompanying this paper being calculated for values of up to 12 are sufficient to determine 169 coefficients in the series of spherical harmonics , by means of a simple computation after the determination of 182 coefficients of the Fourier series .
These latter coefficients may be evaluated by mechanical devices .
S12 .
Special Application to the of Magnetism .
The advantages of the method of obtaining the coefficients of a series of spherical harmonics are considerably increased when the quantities to be represented by such a series are not given directly , but by means- of their differential coefficients .
This is the case when the magnetic potential has to be calculated by means of the observed magnetic forces .
If X and represent the components of magnetic force resolved towards the geographical north and east respectively , the magnetic potential is determined by means of the equations If no electric currents of sufficient intensity traverse the earth 's surface , function V can be found which satisfies both equations .
If ysin be obtained in a series proceeding by spherical harmonics , then all the terms which depend on the longitude are at once expressed in a similar series , as the integration according to leaves each term in the standard form .
The treatment of the other component acting along the meridian involves , however , serious difficulties , and it is not necessary here to enter into the question as to the more or less complicated methods by means of which V has been hitherto derived in the standard form from X. The method based on the results of the preceding investigation avoids these difficulties .
For odd values of is expressed in a series , each term of which has the form or , while for even values of , X is expressed in terms of the form and .
After integration with respect to , the formula of S 11 will determine the required coefficients .
We ay treat the eastern force similarly , obtaining a series proceeding by or according as is odd or even .
is then derived in a series proceeding

