IX .
Al new Method of computing the Value of a slowly con .
verging Series , of which all the Terms are affirmative .
Bv the Rev. John Hellins , F. R. S. and Vicar of Potter's-Pury , in Northamptonshire .
In a Letter to the Rev. Dr. Maskelyne , F. R. S. and Astronomer Royal .
Read March 8 , 1798 .
REV .
SIR .
, Potter's-Piiry , February 8 , x798 .
THAT several of the most curious and difficult problems in physical astronomy have hitherto been solved only by means of slowly converging series , is a truth which you are , well ac_ quainted with , and which may be seen in the works of the late learned EULER , and others , on that subject .
Of this kind of series is the following , viz , ax + bx'+ cx ' +d x4 &amp;c. ad infinitum , when all the terms are affirmative , and a , b , c , &amp;c. differ but little from each other , and x is but little less than i ; to obtain the value of which , to seven places of figures , by computing the terms as they stand , and adding them together , is a very laborious and tiresome operation ; and therefore some easier method of obtaining it is very desirable .
About five years ago , the consideration of this matter was recommended to me by Mr. Baron MASERES , ( who not only employs his own leisure in explaining and improving the higher parts of the mathematics , but also encourages others who make the same laudable use of their leisure , ) to whom I then communicated the method of computation explained in the paper enclosed in this letter .
As this method is general , for all slowly converging series of the form abovementioned , ( which is generally allowed to be the most difficult , ) I am induced to present it to you , requesting that , if it meets with your approbation , you will communicate it to the Royal Society .
I am , Rev. Sir , &amp;c. JOHN HELLINS .
P. S. I need not observe to you , that it is not requisite to the summation of the series mentioned in this letter , that b should be less than a , c less than b , d less than c , &amp;c. but only that the first , second , third , &amp;c. differences of these coefficients should be a series of decreasing quantities : for , you well know , there are series of that form , which arise in physical astronomy , of which the coefficients are actually a diverging series , and yet the sum of the whole is a finite quantity .
And the same thing is evident , from the bare inspection of the theorem which ; I shall presently use .
i. The computing of the value of the series ax + bx'+ cx ' + dx + &amp;c. ad infinitum , in which all the terms are affirmative , and the differences of the coefficients a , b , c , &amp;c. are but small , though decreasing , quantities , and x is but little less than i , is ( as has been before observed ) a laborious operation , and has engaged the attention of some eminent mathematicians , both at home and abroad , whose ingenious devices on the occasion entitle them to esteem .
Of the several methods of obtaining the value of this series , which have occurred to me , the easiest is that which I am now to describe , by which the business is reduced to the summation of two , three , or more series of this form , viz. ax -b x+ c3dx , &amp;c. and one series of this form , viz. px " + qx'n+ rx+ Ti &amp;c. where n is 4 , 8 , i6 , 32 , or some higher power of 2 .
The investigation of this method is as follows .
2 .
The series ax +b x+ caxd + dx4 + ex'+fx6+ c. Is evidently equal to the sum of these two series , viz. ax bx1+ cxi.dx + exs-f ] , &amp;c. *+ 2bXs *+ 2dX ' *+ 2fXI , &amp; C. of which , the value of the former is easily attainable , by the method so clearly explained , and fully illustrated , by Mr. Baron MASERES , in the Philosophical Transactions for the year i 777 ; and the latter , although it be of the same form with the series first proposed , yet has a great advantage over it , since it converges twice as fast .
Upon this principle , then , we may pro .
ceed to resolve the series 2bx1+ 2dx4+ 2fx6+ 2bx'+ 2kx ' + 2mx'1-r &amp;c. into the two following : 2bx2dx4 r+ 2fx6.-shx'+ 9kx ` .
2mx ' , &amp;c. *+ 4dx4 *+ 4hx ' 3 .
Let it be proposed to find the value of the series x+7 x3 x6+ &amp; .
di 9+ t+ 4+P 6+ iflinitum , when x 4 .
In order to obtain the sum of this series , with the less work , it will be requisite to compute a few of the initial terms , as they stand. .
For , if we begin the .
operation with computing the value of x_ IZ + -3_ , &amp;c. .by the differential series be- , 234 fore mentioned * , the values of D ' , D " , D " ' , &amp;c. will be -i , 3.23 &amp;c. respectively , i. e. I ?
I c. &amp; cwhich is a series decreasing so very slowly , that the only advantage obtained by this transformation of the series is in the convergency of the powers of + instead of the powers of x , which indeed is very great ; for , x being=-90,.-Lis= 2 .
; so that the new series -94 + 2-i ?
3 + I. f+ &amp;c. although x.@ I+ 3i ... -I - .
I. 0o &amp;c. yet converges more than seven times as swiftly-f .
But , if we begin the work by computing the first eight terms of the series , as they stand , and then compute 134 the value of -x+ , x_ h + , &amp;c. ad infinitun , by the same 9 101II theorem , the values of D ' , D " , D " ' , &amp;c. will , be 12 9.10 I 9.10.11S 9.2 . !
3 &amp;c. which is a series decreasing , for a great number l The theorem best adapted to this business is the following ; viz. ax bxt+ ax D ' 1 DWX3 D " ' ' x4 CX3dxi , &amp; C. =+ ( '+ )+ ( '+X)3 + ( D+X)4 + Mc .
D ' being =a I+x 2 _b , D)= a2b+c , DI"= a3b+ 3c -d , Mc .
See Scriptores Logaritbmici , Vol. 111 .
p. 290 , where b , c , d , &amp;c. detote the same quantities that a , b , c , &amp;c. do here .
t. is =0.4736842 , and\ is = o.478zg6g .
i9 10 of its terms , much more swiftly than the powers of -- , and , in the first seven terms , much more swiftly than the powers of .
The value of the series -I .
ToI + T-I *2 &amp;c. is therefore the series .
L 9+ '. .
+ .
2 10 9 I9 9.10 19 9.10.111 9+ 993 + &amp;c. the first seven terms of which converge above fourteen times as swiftly as the other ; or , in other words , the first seven terms of it will give a result much nearer the truth than a hundred terms of the other .
And if , instead of the ' first eight terms of the proposed series , the first twentyfour terms were computed as they stand , and then the value of the series + &amp;c. by its equivalent , jl$j+ 25 26 27 28 X x ) +2 x3--3-X + &amp;c. the 25.26 ( I+X ) 25.z6.27 ( i1x ) ' 25.Z6.Z728 ( I +x)4 rapid decrease of the coefficients .
L Z6 ' 25526.27'o &amp;c. compounded with the decrease of the powers of j+j ( in the present case the powers of hI 9 ) produces such a very swiftly converging series , that eight terms of it will give the result true to eleven places of decimals .
It may be further remarked , for the sake of my less experienced readers , ( for whose information this article is chiefly intended , ) that the second term of the last series is produced by multiplying the first by 25 1+X ' P 61 +X , 2Q X R.27 1 +X ) , S.3R X 28 1+.X ; which form is well adapted to arithmetical calculation .
And , that a similar form for -that purpose may always be obtained , when the coefficients of the proposed series are the reciprocals of any arithmetical progression , has been shown by Mr. SIMPsoN , in his Mathematical Dissertations , p. 64 .
Having premised these observations , I now proceed to the arithmetical work ; in which I shall use , and explain , such other devices as have occurred to me for facilitating it .
A. Since the work of computing the value of the proposed series is so much shortened , by first finding the sum of a moderate number of initial terms , and then resolving the remaining terms into several series , in the manner described in Art .
2 , I will begin with computing the first twenty-four terms of it ; and , to facilitate this part of the work , I will separate these terms into three parts ; viz. x+++++++ X9 x"1 wxs ZTo *X3 +4 was .
G + Ale __ +I X +__ ; I_+9 10 II 12 +3 14 ? ?
15 i6 , ,17 is8 r19 %O -wt s x23 X+ x24 17I 8 '9 20 21 22 23 24a which are evidently equal to *+ X+ X3 + 64 + VI+.6.+ X+ Xa |~~ 3X56 +7 2+ , , 5+ , XrJ -~r I07+ + +9 t 20 21 +--+ + 10 12 + ~~13 +14 1 17 ( $x9 20 21Z 2 23 ?
24 ) which three parts , for the sake of reference , may be called A , Bx ' , and Cx'6 .
It is evident also , that the numerical values of the first eight powers of x are all that are required for finding the sum of these twenty-four terms of the series proposed : for Bx8+ Cx"6 is x8 ( B + Cx8 ) ; and this product +A= the sum required* .
The numerical values of the first eight powers of x are these : X= 09 , xt= o 81 , 3x 0 729 .
, x4_ o656i , XS= 0'59049 , .
X6= 05314,41 , 7_ X= 04782969 , xs= 043046721 ; from which values of the powers of x , the values .
of A , .
B. , and C , continued to twelve places of decimals , are easily obtained , and are as follow .
The same thing might have been obtained from the first four powers of x , or front the first six powers of x ; as might likewise the sum of 32 , 36 , 40 , or any other number of terms that is a multiple of 4 .
or 6 .
But the number 8 is here chosen , for the sake of a subsequent use that will be made of the value of x$ .
X = 0.9 x2 7o0405 x3 0243 o= 16402,5 o 118o9,8 56= o0o8857,35 7= oo6832,81285,71 Xs S= 0-05380,84012,50 The sum is 2-04083 , So2g98,21 A. x=0'1 9 =0o081 10 X3 o o6627,27272,73 X4 12 =0 ?
05467,50000,00 X'_ 004542,23076,92 1-3 4 =0-03796,00714,29 -X 0o03188,64600 , oo Xs 6=0 02690,42006,25 The sum is 044412,07670,19 B. - ?
05294 , 11764,7.1 x 20 0'03280,50000,00 21 = 0o0281 1,8t5714,29 2 000241 5,6490-,92[ 17 23--0 2079,551738,91 x3= 0793,61337,50 The sum is 0'26012,12291,85 = C. -We may now quickly find the sum of the first twenty-four terms of the series proposed , by two multiplications by *8 , and two additions , as was pointed out in the former part of this article ; but , since the labour of these two separate multiplications may be saved , I shall now proceed to compute the van 21 lue of the remaining terms of the proposed series .
O20x26 x27 x58 x19 6 .
The remaining terms are xz +-2 + 7+8+2 23~~~~~~~2 + X30 + &amp;c. ad infinitum , which , being disposed according to the form mentioned in Art .
2 , are evidently equal to a ?
4x these The sum is + 2712829 C two seriespro o+ed , &amp; c tw P -i2 X olo894,7,3684,2 25 19 P9 _ .
ax 26iX. . ?
00034,51949,7 S =38X-,99 0'00000,061474 27 19 6u XT=Sx =oooooo , oo4ol,6 30 19 V = ... X =0-00000,00002,9 3 ' 19 -3 z Xx='9 The sum of these terms is 00o1930,53338,5 , which , for the'sake of reference , call .
Then we have a4 = xX : -2+.JL 25 z6 27 x4 Xs x6 -ar ~ _+ __ _ P , c 28 29 30 7 .
The part of the proposed series which n6w remains to be computed , is x:X --+ -228.22+ + 236 + &amp; C. 26 28 30 32 34+36+ ad infinitum , or xFx:-+ 1s + is + 16 +I13 141 17 8+ &amp;c. ad infinitum , which may.also be divided into two series , in the manner above shown ; but , to obtain a swifter convergency in the next , as well as the succeeding applications of the differential series , it will be convenient first to compute the value of four terms at the beginning of this series ; which we may quickly do , since all the powers of x requisite in that part of the calculation are ready at hand , being set down in Art .
A. The work will stand thus : xt o o6230,76923,1 = 0o4686,42857,1 x6 = 003542,94000,0 15 X8 i6 0o269,42006,3 ; And the sum is o017150,55786,5 , which call D. Then will Dx " xI x( *+ _+ X+ Xs 8 .
It now remains to find the value of xa ?
x.- .
+ --8 + a4 + 2'6 +I + X22 + &amp;c. ad infinitum , or of its equal , x " x the sum of these two series,.e X4 x6 a8 X a0 tz &amp; : 17 I8 I9 20 2I1 22 2X4 2X8 2X'1 ** + p I Si8 = P-17 i8i 0262,43418,9 QP 8i Q,181X O8 OI O0,44725,9 R=xI= 0-00003,08300,6 s.-.-2-= 00000206.265 20 i1 0 95 , 4S SI T-HI X81 0 00000)01764)1 UT -5X I81 000000,100179,4 22 i 6J 81 [ =-23 X= 0X00000,00020,9 23 i8i 7V SI Wx=-2X = 000000,000002,7 8W 81 X25 Xi 8i00 ? ? ? ? ?
?4 And the sum of these terms is 0.02701,19108,2 = C. 9 .
To find the value QX8+ 20 24 +C ad infinitum , or its equal , x3 " x+ -+ -+ + &amp; +C .
9 10 11 12 ad infinitum , which is all that now remains of the proposed series , it will be expedient , first , to compute the two initial terms of the series , and then to separate the remaining terms of it into two parts , as has been done in the preceding articles .
These two terms are X4= 0-0729o,00000 , O 8 0]04304,6721oo 10 And their sum is oi 1594,672 10,0E .
1o .
There now remains , of the proposed series , x'x : 4 , + -Ti-+ -+.t,7 + 7F + -l-'+ &amp;c. ad infinitum , =x xthe sum of these two series , sH w8 1x2 *tG Xt 4 Xt I 12 13 14.+ I5 16 0 2x8 2x 16 ' 2*24 +1 *+ I 11 .
The part now remaining , of the proposed series , is x4 2 , *8 2X 1t6 + X24 X " Z4 2 -X4 8x~~~~ ~+.~O ?
~ + &amp;c. ad infinitum , X 12~ + 114 +i6 18 +20 -+ 2Z +euagfnt which may very conveniently be resolved into x48 x these two I+ X8 X:+ 9+ XZl XT X40 V48c seriess+ So serle~i .
2X16 '+ 2X3 ' 2X43 L *~~~81 the value of the first of which will quickly be obtained , by means of the theorem which has been repeatedly used in the preceding operations .
The terms of it are as below , in which the decimal ____ 4304672 !
value of + J8 1 is used , for the reason mentioned in the preceding article .
I -= _ o16666,66667 P=X0 30092,7701.8 0=04298,96717 7 ; X0 30092,77018 00061,70979 R =-_ , X o*30092,77018 = ooooio,8i399 9 10x 0-30092 , ,77oi8 = 000000,97627 T = ; 4x o30092,77o18 = ooooooo683 u= sT x Oo0092,77018 - &gt; 0000001339 14 .
V _O I3Xo %o27701,1 &amp; _ 01 00000,00186 X=0 30092,77018 =o ooooooooo 14 9xY 16 X 0'30092,77018 = Ooooo0ooooI The sum of these terms is o021139,2,5631 12 .
There now remains , of the proposed series , only X48 2x16+ zx3 % 2X 2X6~r'4 X 28 + 2-X + 2xX4 + 21X4 + &amp;c. ad infinitum , the value of X48 X X6 x32 X48 which , or of its equal , xX+ 5- ' + 6+ 7+ &amp; C may easily be obtained without any transformation , since the powers of xi6 ( = the powers of o18530,2o188,9 ) decrease swifter than the powers of 5 .
The numerical values of these terms are as below .
I16 = 004632,55047 4 =o oo686,73676 = 000olo6,04476..= O-oooi6,84312 7 __= 0100002739 8o oo)73093 *T96 -= 0o00000,44982 9x 12 = 0'00000O07502 100 = ooooo0,01264 a:144 -0'000(00,00215 x'"o.-1 0000,00037 '37 147 a:'97 .
15 The sum of these -terms is 0-05445$4461 1= F. 13 .
The values of the several parts , .
into which the proposed series has been resolved , being now so far obtained that we have only to multiply each by its proper factor , viz. the numerical value of x8 , x6 , x3 , &amp;c. and add the products together , to get the sum of it ; this , therefore , is now to be performed .
And , in this part of the calculation , several multiplications may be saved , and no larger factor than x8 be used , by attending to the method described by Sir ISAAC NEWTON , in his Tract De Analysi per .
Equationes infinitas ; p. 1o .
of Mr. JONES'S edition of Sir ISAAC 's Tracts ; or p. 270 .
Vol. I. of Bishop HORSLEY'S edition of all his works .
The manner in which this is to be done will appear , by collecting the several parts from the preceding Articles , and exhibiting them in one view , thus : A+ Bx8+ Cx6+ a+ Dx x'+ C+E xx"+7xy0+ S+ Fx x= the sum of the proposed series .
Now , ist .
Calling A+ F , z ' , and multiplying by x8 , we have z ' x ' _ o026584,70242 x 0-43046,7291 = 0-1 1443,84267,92dly .
Putting v+ z ' x8 = z " , and multiplying by 8 , we get z " X8s 0 15172,24360,1 x o043o46,721 = o'o6531,15337,2 .
Sdly .
Putting C+E+ z"x8= Z " ' , and multiplying by x$ , we get z"/ x%= o'20827,01655,4 x O043046,721 = o'o8965,34770,9 .
4thly .
Putting +D+ z " ' ?
8 _ ziv , and multiplying byx8 , we get z vX8_o028046,43895,9 x 0.43046,72 1 -012073,07232,90 .
5thly .
Putting C+ zivxS= Zv , and multiplying by z , we get zV =s 00 38085,19524,75 x 0-43046,721 = 016394,42774,05 .
6thly .
Putting B+ zv x8 = zvi , and multiplying by x8 , we get rVi x8 = o'6o8o6,5o444 ; 24 x 043046,721-= 026175,2o631,73 .
Lastly , to this product add A= 2'04083,30298,21 , andwe have thevalue of theseries proposed 2%3o258,50929,94 , which is true to twelve places of figures .
14 .
It may not be improper to remark , that this degree of accuracy is much greater than is requisite , even in astronomy ; for which , as well as -for most other purposes , six or seven places of figures are sufficientt : and , to that degree of exactness the value of the proposedseries might have been -obtained , by less than a fourth part of the labour that has been taken above .
But it was my intention to show , that the value of a very slowly converging series , of which all the terms are affirmative , may , by the method now described , be computed to ten or twelve places of figures , in the space of a few hours .

