]\gt ; .
Tables of the Numerical Values of the Sine-integral , Cosine-integral , and Exponential-integral .
By J. W. L. GLAISHER , Trinity College , Cambridge .
Communi- cated by Professor CAYLIiY , F.R.S. Received February 10 , \mdash ; Read March 10 , 1870 .
IT has for a time been evident that the extension of the Integral Calculus would lequire the introduction of new functions ; or , lather , that certain functions should be regarded as primary , so that forms reduced to dependence on them might be considered known .
Thus , in the evaluation of Definite Integrals , the three transcendents called the sine-integral , the cosine-integral , and the exponential-integral , have become recognized elementary functions , and great use has been made of them to express the values of more complicated forms .
They were introduced by SCHLoMILCH to evaluate the integral , and several allied and denoted by him Si , Ci , Ei ARNDT also employed them in a similal manner about the same time .
The first two functions had , however , previously received some attention from BRETSCHNE1DER , who appears to have been led to their consideration by their analogy with the logarithm-integral , as in a paper in the 17th volume of CRELLE'S Journal he announces his intention of not only tabulating this integral , but also of forming " " tabulas aliarum quarundam functionum , cum logarithmo integlali arcte junctarum The Tables are published in the third volume of GBUNJlRT'S ' Archiv der Mathematik und Physik ' for 1843 , and contain ten positive and negative values of the logarithm-integral , and ten values of the sine-integral and cosine-integral , besides tables of several other functions .
The exponential-integral was introduced in its present form by SCHLoMILCH , though for all real values it is the same as the logarithm-integral li , the relation between the two forms being li The logarithm-integral appears to have been first discussed by MASCHERONI and in 1809 a work was published by SOLDNER at Munich concerning its theory , which also contained a Table of its values .
This Table is reprinted in DE MORGAN'S ' Differential and Integral Calculus , ' p. 662 .
* CRDLLE'S Journal , vol. xxxiii .
p. 316 .
Referred to by BRETSCHNEIDEn , CRELLE'S Journal , vol. xvu .
p. 257 .
DCCCLXX .
NUMERICAL VALUESM O in the The sine-integral and cosine-integral occur in aMemoir by BIDONE in the Ton Transactions for 1812 , where they are expanded in series , the same as those marked ( 1 ) on the next page .
From the moment of its introduction the logarithm-integral excited considerable interest , but it is only in the last twenty-five years that the other functions have become of importance .
A complete list of all the memoirs in which these functions are considered is given by Professor BIERENS DE HAAN , on Page 83 of his ' Supplement aux tables d'integrales definies , ' published in the tenth volume of the Transactions of the Royal Academy of Amsterdam ; and in the second volume reference is made to several other works in a memoir by the same author .
Since 1845 the three integrals have been practically regarded as primary functions in the integral calculus ; and how well suited they are to this purpose is evident from the success which has attended the labours of those analysts who have sought to reduce more complicated integrals to dependence on them .
Professor DE HAAN , in the fifth volume of the Amsterdam Transactions , has evaluated a very large number of integrals by means of them ; and in the great Tables*of the same author there are given nearly 450 functions dependent for their evaluation on that of these integrals .
Considering therefore their extreme importance as a means of extending the Integral Calculus , and th probable value of many of the integrals eyaluated in physical inquiries , it seemed very desirable that they should be systematically tabulated , so as to be known , not only by convention , but in reality ; and on this subject Professor DE HAAN has strongly expressed his opinion of the value of such Tables .
BRETSCHNEIDER , in the memoir previously cited , has computed Si , Ci , Ei for the values 1 , 2 , 3 . . .
10 of the argument to 20 places of decimals ( except the values for , which are extended to 35 places ) .
This Table is reprinted by SCHLoMILCH at the end of his ' Analytische Studien , ' and a portion of it is quoted by the same author at the end of a paper in the thirty-third volume of CRELLE'S Journal .
The Tables given in the present paper are the following:\mdash ; Tables , II .
, III .
, \mdash ; Si , Ci , Ei from to at intervals of 01 to 18 places of decimals , with differences to the third order .
Tables V. , VI .
, VII .
, VIII.\mdash ; Si , Ci , Ei from to at intervals of to 11 places of decimals , with differences to the third order .
Table IX.\mdash ; Si , Ci , Ei , and from to at intervals of unity , to 11 places of decimals .
Table X.\mdash ; Si and Ci from to at intervals of 5 , from to at intervals of 10 , from to at of 100 , and for several higher values of to 7 places of decimals .
Table XI.\mdash ; Maxima and minima values of Si to 7 places of decimals .
Table XII.\mdash ; Maxima and minima values of Ci to 7 places of decimals .
In the course of the work BRETSCHNEIDEB'S values have been verified as far as they * Nouvelles tables d'integrales deffiies , Leyden , 186 SINE-INTEGRAL , COSINE-INTEGRAL , AND EXPONENTIAL-INTEGRAL .
coincided with those in the present paper , though from the great care he used , and the mode of verification he adopted , there was little doubt of their accuracy .
One error was detected in , which should be .
instead .
This has doubtless arisen in the final copying or printing .
Expressed in series the fimctions are : Ci Ei I ( 1 ) 7 being EULER'S constant From these expressions it is evident that Si , Ci , and Ei are connected by the relation Si The logarithms are written as above to indicate that they are real when is negative , or of the form .
The logarithm-integral differs in this respect only from the exponential-integral , for li The series in ( 1 ) are clearly always convergent , however large may be .
The following series are easily obtained by integration by parts : Si Ci ( 2 ) 1 .
These series are ultimately divergent , though for values of greater than unity they begin by converging , and when , seven places of decimals are obtainable from them for Si and Ci Formulae ( 1 ) were used for values where the argument was less than 16 , formulae ( 2 ) where it was greater .
Tables I. to IX .
were calculated in the following manner .
The denominators of the terms in the series ( 1 ) were first computed , the first 20 figures of which , as far as the power , and the logarithms of their reciprocals , are given in the following Table .
The powers of all numbers from 1 to 100 were then formed up to the 20th , and in some extreme cases up to the 40th , the numbers under each power being entered in Tables* .
The divisions were then made , and the values of the following four senes , *The powers of numbers from 1 to 100 as far as the tenth are given in HUTTON'S 'Powers of Numbers , ' published by the Commissioners of Longitude in 1781 .
In this , which was only used as a check , the fifth power of 81 and the seventh power of 98 were found to be inaccurate .
SINE-INTEGRAL , COSINRINTEGRAL , AND EXPONENTIAL-INTEGRAL .
were formed , which , when suitably combined by additions and subtractions , gave the values of the functions .
In Tables I. to IV .
the highest power of included was In Table IX .
was required for the value 15 .
In the intermediate Tables was the highest power included .
The values of the functions when the argument was above 16 were calculated from formulae ( 2 ) , the sines and cosines being taken from TAYLOR'S logarithms .
As the formulae ( 2 ) are divergent , to remove every shade of doubt that attach to their use , the functions for were computed from both formulae , and the agreement was perfect to the eighth place , which was as far as the second formulae could give correct results for this value of The calculation of the values from to was extremely difficult and laborious , owing to the great number of terms and high powers necessary to be included .
The term involving was the first one rejected in the calculation for ; and to show how extremely unmanageable the formulae ( l ) had become , it may be stated that this value , calculated as before described , lequired the formation of about 22,000 figures exclusive of verifications .
Great confidence may , however , be placed in the truth of these results from to , as they were also calculated entirely independently by deducing each term from its predecessor , and in addition the value of Ei , which , on account of its extreme smallness , admitted of being obtained from formulae ( 2 ) , served as a rigorous verification of the whole process , excepting the final additions and subtractions .
The functions for were obtained correct to the 12th place , the values Si20 241701043 Ci20 Ei20 Having obtained the values for , it was a matter of comparative ease to give the values of the functions for to a great many places ; they are to 43 places as follows : Si2 720148198688940 8485834 Ci2 565153198255 8941357378 Ei2 001890163 3795051302270352755180536 061119567239 835228049522 3144922 agreeing with BRETSClINEIDER'S values to the first 20 places , which is as far as he has computed them .
The value of 7 was taken from a paper by Mr. SHANXS in No. 114 of the 'Proceedings of the Royal Society .
' BRETSCHNEIDER has calculated the functions for to 35 places .
and Si from Si The cosine-integral has its maxima and minima values for odd multiples .
Those above were calculated from formulae ( 2 ) , the others were deduced from previously calculated values , Ci from Ci ; Ci from Ci ; Ci from Ci 8 ; Ci from Ci 11 ; and Ci from Ci 14 .
The difference formulae in the form best adapted for logarithmic computation are : Si \mdash ; Si Ci \mdash ; Ci SINE-INTEGRAL , COSINBINTEGRAL , AND EXPONENTIAL-INTEGRAL .
Si was calculated both from Si 3 and Si 3.1 , thus verifying the formula as well as the value of Si In all cases the values calculated from the difference formulae on the preceding page were verified to as many places as could be obt , ained fxom formulae ( 2 ) .
In Table XI .
Si is tabulated in preference to Si , as the changes in the function are thus rendered more apparent .
The curves \mdash ; Si were easily drawn from the values in the Tables .
It is thus seen how rapidly the two former curves become flattened , and it is worth notice that the radius ofcurvature at any maximum or minimum point is equal to the abscissa of that point .
The point where the exponential-integral curve cuts The Exponenhal-integral Curve , the axis of has the abscissa , in other words , this is the only real root of the equation Ei It is my intention to determine the points where the sine-integral and cosine-integral curves cut the lines with which they ultimately coincide , and with this view I have already determined the ordinates corresponding to points midway between the maxima and minima values as data for a first approximation .
This will amount to finding the roots of the equations Si ; Ci In Tables I. to , the work has been performed correct to the 20th place , and the last two figures have been finally rejected ; in Tables .
to , the twelfth place has been corrected throughout the work , and the last figure only rejected .
In the other Tables two figures have been generally rejected .
In Tables I. to VIII .
, all the results were verified as far as -figure logarithms were available , and in some cases 10-figure logarithms have been used .
A very large portion of the work was done in duplicate , and every figure has been carefully examined either by my father or myself ; in all cases great pains were taken to ensure accuracy , by independent methods as by logarithms , or in the way by which the values from 10 to 20 were verified .
It may be mentioned also that differences as far as the ninth order have been taken of the numbers in Tables I. to , thus affording a rigid test of the accuracy of more than the first ten , and in the sine-integral of the whole eighteen figures .
Values of SINE-INTEGRAL , COSINE-INTEGRAL , AND EXPONENTIAL-INTEGRAL .
TABLE I. ( continued ) .
Values of the from to 1 at intervals of 01 .
SINE-INTEGRAL , COSINE-INTEGRAL , AND -INTEGRAL .
TABLE II , ( continued ) .
Values of the from to 1 at intervals SINB-INTEGRAL , , AND -INTEGRAL .
TABLE .
( continued ) .
Values of the Exponential-integral from to 1 at intervals 5 S01 SINE-INTEGRAL , -INTEGBAL , AND -INTEGRAL .
TABLE .
( continued ) .
Values of the Exponential.integral from to 1 at intervals l2686003 103514 95 ] 41o 097992 S90 116899 64150 !
SO 083567812 I846400 881 081437002 116519 061317 ( ; 11600 169158 SINBINTEGRAL , COSINE-INTEGRAL , AND -INTEGRAL .
Values of the Cosine-integral from 1 to 5 at intervals of 06 06140 } 05 04 04 088 03 509320 02 02 01536420 96 0111415923 00210 00 71715713 00 00 627583 34 00 01155 01382 01585 89195 02 049500 36 02155 02 238 85017 02 02 335 02 35134919 02 319 02 02131800 65 02 Values of the SINRINTEGRAL , COSINE-INTEGRAL , AND EXPONENTIAL-INTEGRAL .
Values of the Exponential-integral from 1 to 5 at intervals of TABLE X. Table of the Sine-integral and Cosine-integral for values of the argument above 20 .
SINE-INTEGRAL , COSINE-INTEGRAL , AND EXPONENTIAL-INTEGRAL .
TABLE XL\mdash ; Maxima and Minima Values of the Sine-integral .
TABLE XII.\mdash ; Maxima and Minima Values of the Cosine-integral .
The Sine-integral Curve , , for positive abscissa .
The Cosine-integral Curve , Note added July 30 , 1870 .
Professor OPPEBNANN , of Copenhagen , who was present at the reading of this paper , shortly afterwards presented to the Royal Society two pamphlets , " " Tabulae logarithmi alis , auctore L. STENBERG , Malmogiae , Pars I. 1861 , Pars II .
1867 contaimng values of li 10 ' from to at intervals of 01 to 18 places of decimals ; the arguments differ therefore from those in this paper by the modulus of the common logarithms as a factor .
From a reference in the second of these tracts the author found that Tables of Si , Ci , Ei , and from to 1 at intervals and from 1 to at intervals of , had been computed by BnETSCHNHiIDER , and published in the 6th volume of SCHLoMILCH'S 'Zeitschrift fiix Mathematik und Physik .
' The referees recommended the comparison of the parts common to these Tables and those given in this paper ; this has been made , and the following errors have been found in BRETSCHNEIDER'S values:\mdash ; li for should be 32936 instead of 32936 li for should be : instead of 43780 : 06037 : should be the error previouly alluded to in is corrected in this paper .
BRETSCHNEIDER has indicated by dots certain limits between which the eleventh figure must lie , and the agreement between these and the eleventh figure in the Tables V. to VIII .
was so close , that it seemed worth while to retain this figure , on the understanding that it may be in error to the extent of a unit .

