Extract of a Letter from Mr. James Dodson to Mr. William Mountain , F.R.S. May 26 , 1753 .
THE world has , without dispute , been obliged to the invention of fluxions , for many concise methods of calculating the peripheries , areas , and solidities , of curvilinear figures ; but it must be confessed , at the same time , that the most useful , even of those , had been computed before , tho ' by methods more laborious ; and , consequently , since the truth of the principles of fluxions was long disputed , that art seems rather to have received , than to have afforded , any advantage , in those cases .
Read July 5 , 1753 .
Neper and Briggs calculated their several tables of logarithms , with almost insuperable labour ; and Van Ceulen was rendered famous for his approximation to the quadrature of the circle , on account of the acknowleged tediousness of its computation .
The methods of computing logarithms were indeed improved , by the assistance of the properties of the Hyperbola , and the use of infinite series , before the great inventer of fluxions was known in the world : but then , the business of logarithms being purely arithmetical , the Hyperbola was foreign to the subject ; and the nature of infinite series , tho ' well adapted to the purpose , was at that time but little understood .
I am sometimes induced to believe , that , if the latter ( i. e. the nature and use of infinite series ) had arrived at any degree of perfection , before the invention of fluxions , most of the series , which are given for the above-named kind of calculations , and have been deduced from fluxional processes , would have been discovered without the assistance of them : and I am of this opinion , because I am certain , that many of them might have so been : to instance in both the cases above quoted ; viz. the series , for making logarithms , and for rectifying the circle .
And first , the terms of one of the most simple series , for expressing the logarithm of a given number , is composed of the powers of the excess of that number , above unity , divided by their respective indices ; of which the first , third , fifth , &amp;c. terms are affirmative , and the second , fourth , sixth , &amp;c. terms are negative ; and the difference between the sums of the affirmative and the negative terms , is the Neperian , hyperbolic , or ( as some call it ) the natural logarithm of the given number .
Now a mathematician , who understands the nature and management of series ( altho ' wholly ignorant of fluxions , or what the justly celebrated Dr. Halley , in his admired investigation of this very series , published in No 216 of the Philosophical Transactions , calls ratiunculae , &amp;c. ) might arrive at the same conclusion , in the following manner : Since the logarithm of unity is universally determined to be nothing ; that of 2 , 3 , 4 , 10 , or any other number , considered as a root , is one ; that of 4 , 9 , 16 , 100 , &amp;c. considered as the square of that root , is two ; and so on ; it follows , that ( in all cases ) the logarithm of a greater number will exceed that of its lesser ; and each logarithm will have some relation to the excess of its number , above unity , the number , whose logarithm is nothing : the terms of the series , therefore , which will represent the logarithm of any number , will consist of the powers of the excess of that number , above unity , with some , yet unknown , but constant coefficients .
That the logarithm of the square of any number is twice the logarithm of its root , is a well-known property of those artificial numbers ; and therefore , the doubles of the particular terms of the assumed series will constitute a series , expressing the logarithm of the square of the given number .
But , by the fourth proposition of the second book of Euclid , the square of any quantity is equal the sum of the squares of its two parts , more a double rectangle of those parts ; which , in this case ( where the given number has been assumed , to consist of unity , and an excess ) will be unity , more twice that excess , more the square thereof .
If , therefore , the several powers of the compound quantity ( twice the excess of the given number above unity , more the square thereof ) be multiplied by the above assumed coefficients , and afterwards ranged under each other , according to the powers of the said excess ; their sums will again express the logarithm of the square of the given number .
Now , since the logarithm of the square of the given number may be thus expressed by two infinite series , each constituted of the excess thereof , above unity , and its powers ; it follows , that the coefficients of the like powers of that excess , in each series , will be equal between themselves ; and , consequently , the values of the unknown coefficients may be obtained , by simple equations ; and these coefficients will , by the process annexed , appear to be , the reciprocals of the several indexes of the powers of that excess , affected alternately with the signs + and - , as was before found , by the quadrature of the Hyperbola , by Dr. Halley in the above-cited Philosophical Transaction , and by many who have used a fluxional process .
But , there is another logarithmic series equally simple with the former , consisting of the same terms , but all affirmative .
This has been demonstrated to be the logarithm of that fraction , whose numerator is unity , and denominator a number , as much less than unity , as the former number exceeded it .
Now , if an infinite series be formed from that fraction , by actual division , it will consist of unity , and all the powers of that defect ; and if the several powers of the excess of this infinite series above unity , be multiplied by the coefficients above-found , and ranged according to the powers of that defect , their sums will exhibit the above-described series for the logarithm of that fraction , as appears by the operaration subjoined .
As to the application of these two series , and their sum , to the finding the logarithms of numbers ; the same , being copiously treated of by Dr. Halley ( in the Philosophical Transaction before quoted ) there is no occasion for the repetition thereof here .
Note , The above tract of Dr. Halley 's is printed with the explication of Sherwin 's tables of logarithms ( fol. 10 ) with many examples of the use thereof annexed .
Secondly , The terms of one of the best series , for the rectification of the circle , are composed of the odd powers of the tangent of any arc , not exceeding 45 degrees , severally divided by their respective indexes ; of which the first , third , fifth , &amp;c. terms are affirmative ; and the second , fourth , sixth , &amp;c. terms are negative ; and the difference , between the sums of the affirmative and negative terms , is the length of that arc , of which the tangent , and its powers , constitute the series .
Now a mathematician , who understands the nature and management of series , altho ' wholly ignorant of fluxions , might investigate this series in the following manner : It has been geometrically demonstrated , that , the radius of a circle being unity , if the double of the tangent of any arc , be divided by the difference between unity , and the square of that tangent , the quotient will be the tangent of twice the arc .
Now if an infinite series be formed by actual division , its terms will consist of the doubles of the odd powers of the tangent , and will be all affirmative ; which series will express the length of the tangent of the double of that arc , whose tangent and its powers constitute the same .
If a series , consisting of the tangent and its powers , with unknown coefficients , be assumed ( as in the former case ) to express the length of the arc ; then the length of the double of that arc may be expressed two ways ; viz. either by multiplying each term of the series assumed by the number two ; or by finding the powers of the series above described ( which exhibits the length of the tangent of the double arc ) multiplying each power by its proper coefficient , ranging the products under each other ( according to the powers of the tangent of the single arc ) and finding their sum .
Now , since the length of the double arc may be thus expressed , by two infinite series , each constituted of the tangent of the single arc , and its powers ; therefore the coefficients of the like powers of that tangent , in each series , will be equal between themselves ; and consequently the values of the unknown coefficients may be obtained by simple equations .
Lastly , since the series , which gives the length of the tangent of the double arc consists only of the odd powers of the tangent of the single arc , therefore none of the even powers thereof can range therewith : now these will not occur in the odd powers of that series ; and therefore the series , assumed to express the length of the single arc , whose double is to be compared with the sum of the former , must consist only of the odd powers of that tangent ; and then the series first mentioned results from the operation , as will appear by examining the same , as hereto annexed .
I am thoroughly sensible , that , to the learned , who are already masters of the investigation of these series by other methods , this will appear to be an essay of more curiosity than use ; but , with regard to students , I humbly conceive it will have its advantages ; because by those , who are acquainted with the notation of algebra , and the manner of solving simple equations , the nature of the series , and the operations above used , may be easily obtained ; and they may be hereby enabled to make a scientific use of logarithms in arithmetical , and of the rectification of the circle , in geometrical calculations ; which at present cannot be done , till a much greater progress is made in mathematical knowlege , without great labour ; as well as to examine and correct the printed tables of both sorts , if dubious , or found to be erroneous .
The operation necessary to find the coefficients of a series , which will express the logarithm of a given number .
If the given number be represented by 1+n , then the following series may be assumed to represent its logarithm : n+xn2 + yn3 + zn4 + un5 , &amp;c. And 2n + 2xn2 + 2yn3 + 2zn4 + 2un5 , &amp;c. will represent the logarithm of the square of that number ; viz. of 1+ 2n + nn .
But , because 2n+nn is the excess of 1+ 2n + nn , above unity , therefore its logarithm will be also expressed by 2n+nn+x~2n+nn2+y~2n+nn3+z~2n+nn4 , &amp;c. Now 2n+nn2 = 4nn + 4n3 + n4 2n+nn3 = 8n~ + 12n4 + 6n5 , &amp;c. 2n+nn4 = 16n4 + 32n5 , &amp;c. 2n+nn5 = 32n5 , &amp;c. Therefore , 2n+nn =2n+nn ; x~ 2n+nn2 = 4xnn + 4xn3 + xn4 y~ 2n+nn3 = 8yn3 + 12yn4 + 6yn5 &amp;c. z~ 2n+nn4 = 16zn4 + 32zn5 &amp;c. u~ 2n+nn5 = 32un5 &amp;c. And the sum of these is equal to the logarithm of the square of 1+n .
If an equation be formed , of the coefficients of n2 , in each of these expressions of the logarithm of that square , Then 2x =1+ 4x ; whence 1/ 2 = x. And , by proceding in the same manner with the coefficients of n3 , n4 , n5 , &amp;c. and supplying the places of x , y , z , &amp;c. as they arise , by the numbers ( so found ) we shall have 2y = 4/ 2 +8y3 whence + 1/ 3 = y ; 2z = 1/ 2 + 12/ 3 + 16z ; whence 1/ 4 = z ; 2u = 6/ 3 32/ 4 + 32u ; whence + 1/ 5 = u ; Consequently , the logarithm of 1+n will be expressed by n-1/ 2 n2 + 1/ 3 n3 1/ 4 n4 + 1/ 5 n5 , &amp;c. as above asserted .
Again , since 1/ 1-n = 1+n+n2+n3+n4+n5 , &amp;c. as appears by the following division : 1-n ) 1 ( 1+n+n2 , &amp;c. 1-n +n +n-nn +nn And , since the excess of that series , above unity , is the series n+ n2 + n3 + n4 , &amp;c. Therefore the logarithm of 1/ 1-n will consist of the sums of the powers of that series , multiplied by the above-found coefficients 1/ 1 , -1/ 2 , +1/ 3 , -1/ 4 , +1/ 5 , &amp;c. 2 n2 + 2n3 + 3n4 + 4n5 , &amp;c. 3 n3 + 3n4 + 6n5 , &amp;c. Now the 4 n4 + 4n5 , &amp;c. 5 n5 , &amp;c. Power of that Series will be And , ~/ 1 1n+ n2 + n3 + n4 + n5 , &amp;c. -1/ 2 2 -1/ 2n2-2/ 2n3-3/ 2n4-4/ 2n5 , &amp;c. +1/ 3 3 +1/ 3n3+3/ 3n4+6/ 3n5 , &amp;c. -1/ 4 4 -1/ 4n4-4/ 4n5 , &amp;c. +1/ 5 5 +1/ 5n5 , &amp;c. of that Power will be The sums of which , viz. n+1/ 2n2+1/ 3n3+~/ 4n4+1/ 5n5 , &amp;c. will be the logarithm of 1/ 1-n , as above affirmed .
The operation necessary , to find the coefficients of a series , which will express the length of the arc of a circle , by the tangent of that arc , and its powers .
Let a represent the length of the arc , and t the length of the tangent thereof ; then the tangent of that arc whose length is 2a , will be 2t/ 1-tt ; which fraction is equal to the infinite series , 2t+2t3+2t5+ 2t7 + 2t9 , &amp;c. See the division .
1-tt ) 2t ( 2t + 2t3 + 2t5 , &amp;c. 2t 2t3 + 2t3 + 2t3 2t5 + 2t5 And by performing the necessary multiplications , or divisions , it will also appear , that ~2t/ 1-tt~3= 8t3 + 24t5 + 48t7 + 80t9 , &amp;c. ~2t/ 1-tt~5= 32t5 + 160t7 + 480t9 , &amp;c. ~2t/ 1-tt~7= 128t7 + 896t9 , &amp;c. ~2t/ 1-tt~9= 512t9 , &amp;c. Now if we assume , for the value of a , the following series , t+ xt2 + yt5 + zt7 + ut9 , &amp;c. Then 2t + 2xt3 + 2yt5 + 2zt7 + 2ut9 , &amp;c. =2a .
And because 2t/ 1-tt is the tangent of the arc , whose length is 2a , therefore 2t/ 1-tt +x~~2t/ 1-tt~3 +y~~2t/ 1-tt~5 +z~~2t/ 1-tt~7 &amp;c. =2a .
Which expression is equivalent to the sum of the following series ; for 2t/ 1-tt = 2t + 2t3 + 2t5 + 2t7 + 2t9 , &amp;c. xX~ 2t/ 1-tt3 = 8xt3+24xt5+48xt7+80xt9 , &amp;c. yX~ 2t/ 1-tt = 32yt5+160yt7+480yt9 , &amp;c. zX~ 2t/ 1-tt = 128zt7+896zt9 , &amp;c. uX~ 2t/ 1-tt = 512ut9 , &amp;c. And ( by making an equation , between 2x , the coefficient of t3 in the first-found value of 2a , and 2+8x , the sum of the coefficients of t3 in the latter ) 2x =2+ 8x ; whence 1/ 3 = x. And , by proceding in the same manner , with the coefficients of t5 , t7 , t9 , &amp;c. and supplying the places of x , y , z , &amp;c. ( as they arise ) by the numbers ( so found ) we shall have 2y=2 2~/ 3 + 32y ; whence y=+~ .
2z=2 48/ 3+160/ 5+128z ; whence z=-1/ 7 .
And 2u=2-80/ 3+~-~+512u ; whence u=+1/ 9 .
Therefore we may conclude , that t-1/ 3t3+1/ 5t5~ t7+1/ 9 t9 , &amp;c. = a. The application of this series to the rectification of the circle is extant in many authors ; particularly in Sherwin 's tables of logarithms , above-quoted , folio 55 .
When the arc is just 45 degrees , then t=I , and the series becomes 1/ 1-1/ 3+1/ 5-1/ 7+1/ 9 , &amp;c. which converges exceedingly slow ; but , by the assistance of a method , given in the appendix to M. De Moivre 's Miscellanea Analytica , it may be transformed to another , converging quicker ; which method is applied to this very series , in folio 362 of the Mathematical Repository , Vol. I.

