]\gt ; Tables for the of an Equation .
Tables for Solution of the Equation By W. STEADlfAN ALDIS , M.A. Communicated by J. J. THOIISON , F.R.S. Beceived and June 16 , 1898 .
1 .
The object of the present paper is to exhibit the processes of calculation of the values of the two solutions of the tion ( 1 ) , for stlccessive values of in the two cases of and That is if be the complete integral of ( 1 ) , where is a function which becomcs zcro is indefinitely increased , our object is to calculate the values of , , , for successive equidistant of The values of and have been calculated and published by committee of tho British Association for the Advancement of Sciencc .
To the best of the writer 's knowledge , no steps have been towarcls the computation of and .
The Tables I and II , the end of this paper , give the values of these latter functions for of in the argument , to such a large number of places as make it a mere matter of difference calculation to determine intermediate values of and to any reasonable degrce of , at any for values of greater than umity ; and also ) means of the sequence laws to derive those of , , may be requisite .
It convenient to state a nnmber of well-known theorcms regard to the solution of ( 1 ) .
2 .
The fmtction is defined by the condition For ducs of is a solution of ( 1 ) , being the function defined by Gauss .
* ' Werke , ' vol. 3 , p. VOL. LXIV .
Mr. W. Steadman Aldis .
When is a positive integer , a second solution is given by ( 4 ) , whore - ( 5 ) , where denotes the series .
It is also necessary to assume zero as the value of the , in itself meaningless , symbol It is also known that if represent the quantity the function becomes indefinitely small as increases indefinitely .
since in virtue of ( 3 ) and ( 4 ) is iously a solution of ( 1 ) , it is allowable to write ( 6 ) .
3 .
The three functions I , , and ars all subject to the where is written for .
These equations hold when either or is substituted for I. When has the value zero , the two equations must be replaced by the single equation or the same with A or written for I. Tables for the Solution of .
an uation .
These laws give , for values of not less than unity , ( ) .
They are known and will be quoted as the sequence laws .
4 .
It can be shown that is expressible in two ways in terms of a definite integral , namely , ( 7 ) , ( 8 ) .
By putting in ( 7 ) , expanding the binomial and integrating the separate terms , another form can be obtained for , namely , ; where the series within the bracket can be brought to a close at any point by means of a remainder term , which , after a certain point in the series , is always numerically less than the next term given by the general law of the series .
5 .
It is now possible to explain the processes by which the Tables I and II tho end of this paper have been calculated .
The series actnally employed arc , for the smaller values of , the uItimately convergent serics ( 6 ) ; and for larger values the series ( 9 ) .
In the calculation of the functions , the latural logarithms of are required .
These the writer has taken from Wolfram 's table at the end of Vegtt 's ' Thesaurus Logarithmorum , ' having , in the numbers up to 20 for the prime numbers up to 59 , verified them to 30 places of decimals by calculation .
The qnantity has been derived from Using 's notation it follows that .
* Werke , ' vol. 3 , p. 155 .
Mr. W. Steadman Aldis .
From Wolfram 's table , taking thirty-six places , log2 121458176568 .
The value of is given in a note by Gauss as 901532860606512090082402431 .
The algebraical sum of these is which is , therefore , the value of to many more places than will be required .
The quanbity is , of course , Euler 's constant , and the above value is also to be derived from a paper by the late Professor J. C. Adams in the 'Proceedings of the Royal Society .
' 6 .
The calculations of , , , are best carried on in connection with one another .
We hay The first process is to find the values of and .
If a series of quantities , , , be determined by the successive relations coupled with the condition , it easily seen that Thus the successive terms of and are obtained by multiplying by a series of factors of the form ; the alternate terms when obtained ars written down underneath one another , the odd ones in one column , the even ones in another , and by addition of each column the values of and are obtained .
In working out the values of and given in Table I , all the Tcibles for the Solution of an .
207 multiplications by Sx/ rl have been conducted in two different forms to avoid the possibility of mistakes .
Thus , for lnstance , working out and , the factor Sx/ 8 , or , can be used as it stands , or as , and also put into the form .
The adoption in all cases of two quite different processes is an almost infallible guide to the detection of a mistake .
7 .
The values of and being thus obtained , and can be derived .
1Ve have But Adding It will be convenient to denote by the Hence ( 10 ) The value of is known and that of can be found from Wolfram 's Table .
The quantities must be derived , each from the ponding For earlier values of the multiplier is most easily used in the natural form For later valucs , it is simpler to use the decimalized values of given V. In using the primary form , many simplifications are possible , thus , and the multiplication is effected by shifting the decimal point one )lace to the right , and dividing by 12 .
Again , , , and so on .
In the computation of the values in Table I , two Mr. W. Steadman Aldis .
different processes for computing each have been employed , so any mistake is almost certain to have been detected .
For the lower values of a second process of calculating the of has been found from the obvious fact that if be the when becomes Thus the values of the quantities for can be deduced those for by a series of divisions by 2 or powers of 2 .
For the values of from upwards , this process was available , but either two different transformations of the sum of vulgar fractions , or one such transformation , and the decimalized have been used in every case .
Another process , which has been occasionally used , when the happened to be in low terms , is based on the easily formula It only remains to multiply by and , adding unity to this product , to subtract the sum from .
The value of , which is always a positive quantity , is then obtainsd .
8 .
The second function can be readily expressed in terms of quantities already found .
For adding but Tables for the Solution of Equation .
Hence ( 11 ) The calculation therefore only involves two operations which entail much labour , namely , the muItiplication of by and the computation of the series .
9 .
The quantities given in Table I have been computed by these formlae .
The nlnltiplications of and by have } ) worked , taking as multiplicand , which saves a good deal of labonr , as many of the lines of multiplication used in finding occur again in finding .
Inall cases the multiplications have been carrisd to several places further than are used , or given , in the final results .
10 .
A process of verification has been applied to alues giver .
, based on the following theorem .
It is easy to show that if and be any two different solutions of the fundamental equation ( 1 ) , Mr. W. Steadman Aldis .
In the particular case of and may have the values and respectively .
Hence or , by means of the sequence laws , Multiplying by , and putting , it is easily seen that consequently for all values of .
The writer has to thank his friend Captain Makgill , R.E. , of Waiuku , Auckland , for verifying by this formula most of the results obtained before the writer left New Zealand .
For the others he has had to rely on his own verification .
The formula is an infallible indicator of any mistake in the values of or , or in the process of multiplication of and into .
It obviously will not indicate an erroneous value of this last quantity .
The values of have been all calculated in two different ways , so as to avoid the possibility of mistake , but in order to give the greatest security , a table of the values employed is appended , and the writer hopes that if any mistake is detected , information of it may be sent to him , as it would be a very easy matter to supply the requisite correction to the values of and .
As a final test of the accuracy of the results , the differences of the column for have been calculated up to those of the seventeenth order .
Up to this point they present in each set of differences a series of regularly decreasing quantities .
In the differences of the eighteenth order this ceases to be the case with regard to the quantities at the lower end of the column .
This is due to the accumulation of the effect of residual error in the last figures of the column of values of .
The differences of the seventeenth order at the lower end of the column quantities consisting of fifteen ciphers followed by six signfficant figures .
Now since is greater than a million , it follows that a residual error of four-tenths of a unit in the last figure , in opposite directions in two consecutive values of might possibly , after eighteen differentiations , produce an error of a unit in the sixth place from the end , consequently completely disorganise the sequence of the eighteenth differences which consist only of five figures .
That this has actually happened in this case the writer has shown by examining the effect of adding to the values of given in Table I the three additional Tables for the Solution of an Equation .
figures , two of them certainly correct , which he has calculated .
The differences at the lower end of the table then become regular up to the twentieth order .
This process has not been applied to the column , because the writer believes that , granted correct , the verification formula above sufficiently proves the accuracy of .
The values of the quantities in Table I are believed to be correct to the last figure given .
A dot after the last figure indicates that it has been increased by unity , the first figure omitted being equal to or greater than 5 .
11 .
Table II has been computed by means of the formula ( 9 ) .
The remainder after terms in the series involves the integral\mdash ; \ldquo ; where is some proper fraction .
Now whatever may be , after a time becomes negative .
has reached such a value , inspection of ( 9 ) shows that the terms in the series thereafter are alternately positive and negative , as much as a new negative factor is introduced in forming each successive coefficient .
It is also evident that , from and after that point in the series , the quantity is numerically less than and the remainder required at any point to give the value of is numerically less than the next term in the series .
Consequently , after the alternation of signs has begun , the sums of terms , terms , terms , \amp ; c. , will be a series of quantities lternately greater and less than the value of .
As long as the te1ms of the series diminish , it is possible in this way to obtain a set of quantities , continually approaching one another , between alternate pairs of which must lie .
For the values , ( 9 ) gives\mdash ; ( 13 ) ( 13 ) 12 .
In the alternation of signs begins with the first term .
Hence the stlm of 1 , 3 , 5 , terms is numerically greater than the value of , while the snm of 2 , 4 , 6 , terms is less .
The is derived from the by multiplying by .
As long as this factor is less than unity , the ternl is less than the , and the terms continue to diminish .
The Mr. W. Steadman Aldis .
term is least when has the largest value , which makes less than .
This gives , where is the integral part of Hence the nearest approach of the limits , within which ( 12 ) confines the value of , is ( 14 ) It is evident that as lies between the sum of terms , and the sum of terms , the mean of these two sums is as near an approximation to the actual value of as ( 12 ) will give .
This mean cannot differ from by quite half the quantity ( 14 ) .
If be an integer , the value of is ; thus , if the third term is the smallest : when the eleventh , when the seventeenth , and so on .
The limit of error , estimated by half the least term , is for ; for ; for 000022 and for larger values of the limit becomes rapidly smaller .
For values of as great as , or greater than , five , can thus be determined with accuracy to seven or more places of decimals .
Very similar statements oan be made with respect to the determination of from ( 13 ) .
13 .
From ( 12 ) The multipliers , disregarding the sign , by which the coefficients of the powers of within the bracket are derived , each from the preceding , are Let these numbers be denoted by the symbols , and let bs called .
Then if a series of quantities , : be derived by the successive relations , ( 15 ) it is evident that The relations are adapted to logarithmic computation .
For the value of two logarithms beside that of are required .
These are ; .
Tables for the Solution of an .
213 With the help of these and the logarithm of , that of can be easily ascertained , and then , if the logarithms of , be tabulated , it is easy to derive those of , in succession .
The logarithms of , as far as it has been necessary to use them in the construction of Table II , are given at the end of this paper in Table VI .
14 .
In going through the calculation , it is , of course , useless to take the values of the quantities , to a decimal place fnrther than the last one which can be accurately obtained in .
If ten-figure logarithms be used , ten significant figures can be ordinarily obtained with accuracy from the logarithm .
Of this the writer has satisfied himself by working out the value of by elementary metic and the exponential theorem , for one or two simple values of , as , and comparing the result so obtained with that derived from the logarithms .
They always agree for ten places , sometimes for eleven , if account be taken of the second differences of the logarithms .
It follows that for larger values of , for which the smallest term in the series is less than , the value of can be obtained with accuracy , probably for ten , and pretty certainly for nine significant lres .
The tenth figure may be in error owing to the accumulation , in addition , of the errors in the last places of the quantities , 15 .
quation ( gives The multipliers , disregarding sign , by which the coefficients of the successive powers of within the bracket are derived , each from the preceding , are Let these be denoted by the symbols , and let a series of quantities , , be obtained from by the suocessive elntions , ( 16 ) having the same value as in Article ( 13 ) .
Then evidently . .
; the summations being carried on , either until the smallest term of the series is reached , in the case of the lower values of , or until a term is rrived at which is less than , which will happen first for larger values of Mr. W. Steadman Aldis .
The relations ( 16 ) are adapted to logarithmic computation .
The logarithms of , are given in Table 16 .
The verification of the values of , in Table II , cannot be conducted on the method applied to those in Table I , because the values of , are wanting .
A certain amount of check is given by the values of the four functions , calculated for the integral values of , by the former method , given in Table III .
Two other checks , in addition to the useful one of performing all additions and multiplications in two different ways , have been applied throughout .
The first depends on a very simple relation between the quantities and It is easily seen , from the general formula for the term in , that which , since , easily reduces to .
Thus When the quantities and have been calcuIated from the logarithmic formulae , this result gives an easy method of verification .
It detects any mistake in the computation of the logarithms , or in the derivation of the number from the logarithm .
This formula leaves untouched the possibility of a mistake in the value of .
To check another process has been used .
17 .
If be any continuous function of , whose differential coefficients are also finite and continuous for the values of considered , Taylor 's Theorem gives Let be the value of corresponding to any particular value of , and let , denote the values of , , , Similarly , let denote , , , Then If be so small that the terms of the ssries on the right hand not written down may be neglected , and the three terms written be denoted by , respectively , it follows that Tabtes for the Solution of an quation .
say ; writing for , this gives and putting for From these three equations can be foumd in terms of , and its value can be put into the convenient shape ( l8 ) From the sequence law it follows that Consequently , if values of for seven equidistant values of x.be taken , the quantities , and can be derived , and ( 18 ) ought to give value of equal to that of for the middle value of .
This test has been frcely applied throughout Table II with very satisfactory results .
18 .
If the values of and be taken into account , still more stringent test is afforded .
As before , let these quantities be denoted by and .
Let be used to denote and let Then ; hencc it is not difficult to show that This can be used independently , or it can be made to yield a correction to ( 18 ) .
In the latter case the quantity has to ] ) subtracted from the value of given by ( 18 ) .
Mr. W. Steadman Aldis .
19 .
As an example , suppose the value of is taken as .
Table II gives whence , remembering that , it easily follows that , a minus sign being understood before all the numbers , whence Hence , using the formula ( 18 ) , The value in the table for is If we apply the correction ( 20 ) , the above values give Correction This has to be subtracted from the former value , and the result agrees exactly with the value of in Table II .
The agreement is not in all cases quite so exact as in this example , as may be expected from the necessary existence of more or less of error in the last figures taken into account .
Tables for the Solution of an Equaticn .
A slight additional verification of the general accuracy of Table II has been gained by the calculation of the term for the values 8 , 9 , 10 , 11 , and 12 by elementary arithmetic and the exponential theorem without the use of logarithms .
The last figure of the quantities in Table II oannot be depended on for strict accuracy , in which respect the table differs from Table I. 20 .
A farther exbension of the formulae of Articles 17 and 18 has some interest .
If , with the same notation extended , the quantity be denoted by , it is not difficult to prove that This value of can most easily be computed by subtracting from the value of given in ( 18 ) .
This farther correction is too small to be applied with any oertainty to the lues of derived from in Table II .
Obviously however , all these formulae may be equally well applied to Table I , and throughout the range of that table , this formula deduces a value of more accurate to one or two places than that given in ( 19 ) .
To two examples ; one from the earlier part of the table .
If Equation ( 18 ) gives ( 19 ) , , ( 21 ) , , hile the correct value is taking the largest value of in Table I which admits of the application of ( 21 ) , namely Equation ( 18 ) gives ( 19 ) , , ( 21 ) , , hile the correct vadue is of these formulae is sufficient for Yerifioation of the values in I to the last figure given .

