compputation of the Lengths of the Waves of Light corresponding to the Lines in the bispersion Spectrum measured by KIRCHIOFF .
By GEORGE BIDDELL AIRY , Astronomer Royal .
Received March 2 , -Read March 21 , 1867 .
THE greatest and most valuable system of measures which we possess of the lines in the Dispersion Spectrum , produced by the emanation of light from the Sun , by tran~smission through the Atmosphere , and by the combustion of Metals , is that published by Professor KIRCHHOFF in the Memoirs of the Berlin Academy for 1861 and 1862 .
This noble series includes about 1600 lines , taken in uninterrupted succession through the spectrum , beginning a little before FRAUNHOFER'S line A and ending a little after G. The same apparatus was used throughout , and in the same manner ; and the system .
followed in the whole process was so nearly constant that we may speak of it generally as one consistent series of measures .
This work has been made known to English readers by an accurate translation published by Professor ROSCOE .
The measures thus exhibited depend upon the form of the apparatus , the refractive and dispersive powers of the glass , the nature of the scale employed , and the value of its unit of measure .
They are not therefore available for physical research , until they are cleared of the effects of these circumstances of experiment .
There can be no hope of our arriving at , or even commencing , any mechanical theory on the formation of the spectral lines , until we have obtained the one natural measure of each , namely the length of the Wave of Light corresponding to each .
I have therefore undertaken the labour of computing from each spectral measure the corresponding length of the Wave of Light ; and I trust that the result may not be unacceptable to the Royal Society .
I will now describe the method which I have used .
When I commenced this work , I could not learn ( although I made inquiries ) that any careful measures had been made of the Diffraction Spectrum since the original measures of FRAUNHOFER .
I therefore determined to make FRAUNHOFER'S measures of the principal lines the basis of my computations .
From GILBERT'S ' Annalen der Physik und der Physikalischen Chemie , ' XIV .
Band , Leipzig 1823 , page 559 , I took the following values of the Length of a Wave , omitting the value for B , which apparently was not determined in the same series of measures with the others : in .
FRC =Length of wave for C , in Paris inch , =0 00002422 , FERD= , , , , D , , ==0-00002175 , F RE= , , - , , E , , _0 00001945 , MDOCCLXYIIL .
F in .
FR : =Length of wave for F , in Paris inch , =0-00001794 , FRG Ad , , G , , 000001587 .
The lines are not designated , in KIRCHiOFF 's tables of measures , by the received letters ; the letters , however , as well as KIRCHHOFF'S scale-numbers , are attached to his engravings of the spectrum , which appear to be carefully drawn .
From these I took the following numbers : KC =KiRClHloFF 's scale-measure for C 694 1 , KD= , , , D=1004 8 , KE= - , , , E-1527 7 , KF , , , , F=2080*0 , : KG= , , , , G-2854 3 .
The next step was to construct a formula , which , when applied to KIRCHd OFF'S values for C , D , E , F , G , should produce FPAUNHOFIER'S values for the same lines .
I know no physical reason for adopting one formula in preference to another ; and I therefore adopted the simplest algebraical formula .
Putting kl=T , X-K , andfg=100000000 x FR1 , I assumed fI ( + b.k+ 1c(k1)2+d(k1)3 +e(kj)4 , and formed five equations numerically by putting successively on the left side j % , fD , &amp;c. , and on the right side lc , , kl , &amp;c. , giving to those symbols the numerical values derived from the small tables above .
Solving the equations , I found fib= 3596688 2582 018 X e1 , +1587046 X ( IC)2 4760676 X ( k1)3 + 52-959 x ( kl)4 .
From this formula , values off and FFR1 were tabulated for every 0 01 of Il , or for every 10-0 of KIRCIIHOFF'S scale .
It will be sufficient to give here every tenth number .
Fraunhofer .
Fraunhofer .
Fraunhofer .
Kirchhoff .
in .
lKirchhoff- .
in .
Kirchhoff .
in .
00000 00000 00000 0 35967 1000 21780 2000 18148 100 33539 1100 211'99 2100 17888 200 31400 1200 20697 2200 17625 300 29525 1300 20262 2300 17358 400 27887 1400 19879 2400 17087 500 26462 1500 19538 2500 16814 600 25227 1600 19229 2600 16539 700 24162 1700 18942 2700 16269 800 23244 1800 18671 2800 16007 900 22456 1900 18407 2900 15760 3000 15535 Between the numbers computed for every 10 0 of KIRCHHOFF'S scale , the numbers were interpolated for every one of KIRCHHOFF'S line-mreasures by simple proportion of first differences .
Thus I had a complete table of the values of the length of the wave of light expressed in Paris inches for every one of KIRCHHOFF'S lines .
Each number was then multiplied by 27-0700 , and thus I had a complete table of the values of the length of the wave of light expressed in millimetres for every one of KIRCHHIOFF'S lines .
In this state I very carefully revised all the final numbers .
This is the table of results now offered to the Royal Society .
It did not appear necessary to publish a long series of results expressed by a measure which is now so nearly obsolete as the Paris inch ; although , as will appear below , the possession of those numbers in my own hands has proved to be of considerable utility .
From the description which I have given of the process by which I have computed the wave-lengths from KIRCHHOFF'S measures , it will be perceived that the results are not rigorously accurate , unless the same system ( whatever it may be ) has been rigorously followed by KIRCHHOFF through the whole of his measures .
Now KIRCHHOFF has distinctly stated that he did not in every case adapt his prisms carefully to the angle of minimum deviation .
So far as I could learn by communication ( through Professor RoscoE ) with M. KIRCHHOFF , no memorandum was preserved on the possible amount of the error thus introduced .
I was confident , however , as well from the character of Professor KIRCHHOFF as from the absence of any apparent sctltus in the measures , that the error must be small , perhaps imperceptible .
It was a matter of great satisfaction to me therefore , after the completion of my work , to be referred to two late series of direct measures of wave-length for numerous lines .
One is that by ANGSTR6M , published in POGGENDORFF'S ' Annalen der Physik und Chemie , ' CXXIII .
Band , p. 498 , in which I have found 51 lines apparently identifiable with lines measured by KIRCHHOFF .
The measures in these series are expressed by the Paris inch ; and the possession of my own calculations in terms of the Paris inch , to which I have alluded above , greatly facilitated the comparison of ANGSTROM'S measures with my calculations .
The other is that by DITSCHEINER , published in the 'Sitzungsberichte der Mathematisch-Naturwissenschaftlichen Class der Kaiserlichen ( Wiener ) Academy der Wissenschaften , ' L. Band , II .
Abtheilung , 1864 , page 340 : it contains direct measures of the wave-length , expressed in millimetres , for 107 of the lines measured by KIRCHHOFF .
The comparison of these direct measures with my computed values of the wave-length exhibited small discrepancies , of which the greater part arises from discordances between the values for the fundamental lines found by FRAUNHOFER and those determined by ANGSTROM and DITSCHEINER ( which those philosophers ascribe to an erroneous evaluation by FRAUNHOFER of his interference-grating ) ; a part apparently is produced by faults inherent in every process of interpolation among a limited number of values , especially when that interpolation is so extended as to become extrapolation ; but , as I believe , no sensible part , or a very small part , can arise from the suspected class of errors in the measures of the prismatic dispersion-spectrum .
I had arrived at this stage of my work , when I became acquainted with DITSCHEINER'S Essay contained in the 'Sitzungsberichte der IMathematisch-Naturwissenscha~ftlichen Class der Kaiserlichen Akademie der Wissenschaften , ' LII .
Band , II .
Abtheilung , page 289 , &amp;c. This paper gives the details of a new measure of the grating with which FRAUNHOFER'S original measures of wave-lengths were made ( as I understand M. DITSCHEINER'S words ) ; and by means of this correction of measure , and the corresponding correction of wave-length for the line D common to FRAUNHOFER'S and DITSCHEINER'S measures , DITSCHEINER computed new values of wave-lengths for the entire series of lines , primary and secondary , which he had himself observed .
On careful consideration of all the circumstances of formation of these new values , it appeared to me that it was indispensable for me to base my reduction of KIRCInHOFF'S measures upon DITSCHEINER'S new values .
I should thus have the opportunity of introducing a value for , B , which would enable me to adopt a function of the 5th order instead of the 4th order ( upon which all my preceding calculations were founded ) ; and I could use the same opportunity for making correction for a petty error which I had committed in the line E ( I had inadvertently used 1527T7 instead of 1523-2 for KIRCHHOFF'S measure ) .
My immediate object , in calculation , now was , to make such an addition to the function embodied in my former calculations as would produce definite alterations in the wave-lengths computed for B , C , D , E , F , G. This , it was evident , could be done by a function or , a sum of functions of the 5th order ; and I had only to adopt the most convenient form .
Remarking that KIRCHROFF'S measures for B , C DP , E , F ( with the decimal point thrown back three places ) differ very little from 06 , 0T7,1 0 , 1'5 , 2'1 , and that the measure for G is not very different from 2'9 , I adopted the following form for general correction ; where by [ 06 ] , [ 0'7 , [ 10 ] , [ I'5 ] , [ 2'1 ] , [ 2'9 ] , are meant certain constants whose values will shortly be seen .
General correction for all values of k. correction for k= 06 [ O'6 X(k-O07 ) X ( k-1'0)X(k--1'5)x ( k--21)X ( k--2'9 ) correction for k= 0'7 + [ o ] X ( k-0'6 ) X(k-17O)x ( k-15)x ( k-2'1)x ( k-2'9 ) correction for k =1IO +X ( k-O06)X(k-O07)X ( k-1'5)X ( k--2'1 ) X(k-2'9 ) +correction for k= 15 [ 1'5 ] + correction for k*1-5 " '1-)~ --7xkIOxk--~~ 29 +orect[2'1 ] X ( k-0'6 ) X ( k07 ) X ( k -1O ) X ( k-1'5 ) X ( k-2'9 ) correction for k=2'9 + [ 2'9 ] X(k-0'6)X(k-O07)X(k-1O)X(k--1'5)X(k-2'1 ) .
The numerical values of the products following the fraction in each line .
are given , for every 0'1 of k from 0'0 to 3'0 , in the last six columns of the following Table .
numbers is necessarily ( from the nature of the process ) almost perfect .
But between each of these principal lines and the next the discordance rises gradually from 0 to a considerable value and then drops gradually to 0 .
Between F and G the value is large .
The discordances , however , are systematic .
There is no appearance of any saltus between one principal line and the next ; but there is nothing to enable us to decide positively whether there is a saltus at each principal line .
It appears to me that we must look to one of the three following causes for explanation:1 .
The actual inability of a formula of the fifth order to represent the wave-length in terms of the spectrum-measure with sufficient approximation .
2 .
A change in Professor KIRCHHOFF'S method of observation at each successive principal line .
3 .
A change in DITSCHEINER'S and ANGSTROM'S methods at each successive principal line .
The third of these suggestions is extremely improbable : and after remarking the harmonious flow in the course of the discordances in every interval between principal lines ( a harmony which gives no small testimony to the care employed in both classes of measures , and to the accuracy of the interpolation-details in the computation which connects them ) , I am hardly inclined to advocate the second suggestion .
I think it more probable that the real cause is to be found in the first .
I remark , however , that this doubt affects only the physical question on the broad scale .
It does not affect the possibility of computing with accuracy the wave-length for any one of KiRCHHOFF 's lines .
By means of the comparison between the computed wave-lengths and DITSCHEINER'S measures ( at the end of the Tables ) there is no difficulty in computing for any other line the correction that ought to be applied to the wavelength in the principal Tables , in order to exhibit the true wave-lengths on DITSCHEINER'S scale , without appreciable error .
I have now only to explain the Tables which follow .
The principal Table , headed " DConversion of KIRnCHoFF 's Spectral Measures into Wave-lengths in terms of the Millimetre , " corresponds , in extent , to KIRCdHOFF 's Tables in the Berlin Memoirs , 1861 and 1862 .
Instead , however , of adopting the broken arangement of these Memoirs , I have placed the numbers in a continuous series , and I have omitted the repetitions in KiRCIIHOFF'S successive Tables .
The Table consists throughout of four columns ( one of which is occasionally blank ) .
The first column contains simply KincHHOFF'S measures ; and the second contains the corresponding wave-lengths , computed by the process described above .
The third column contains KIRCHHOFF 's symbols expressing the darkness and the breadth ' of the lines ; 1 expresses a slight darkness , 2a little darker , and so on up to 6 which is very dark ; a denotes a very narrow line , ba little broader , and so on up to g , which is a line of considerable breadth ( the measure of the breadth is not given ) .
The fourth column ( where it is used ) contains the contracted name of the metal which , by combustion , produces a line corresponding in spectral position to the line here measured .
This Table is , in fact , KIRCH11OFF 's Table , with the insertion throughout of one column containing the computed values of the wave-lengths .
In copying the first column , I have endeavoured to copy the braces and brackets of KIRCdIIOFF 's Table , without in every case understanding their application .
The Table which next follows is KIRCHHOFF'S special Table for the lines produced by the metals Cerium , Didymium , Lanthanum , Palladium , Platinum , Ruthenium , Iridium , with the corresponding wave-lengths .
Two of its compartments contain the lines in a combination of Lanthanum and Didymium , and in a combination of Ruthenium and Iridium .
It is accompanied by a similar Table of lines produced by the atmosphere .
The next Table ( formed by myself ) contains the wave-lengths for each of the metals whose lines have been measured by KIRCHdOFF , and are contained either in the General Table of lines of the solar sppctrum , or in the smaller special Table which I have just described .
The numbers are simply extracted from those Tables .
I may mention that my original intention was to confine my computations to these metal-lines , in the hope that they might give some assistance to the formation of a mechanical theory .
The two last Tables contain the comparison of direct measures of the wave-length made by ANGSTROM and DITSCHEINER with the computed measures given in the General Table , as far as the identity of the lines appears to be established .
ANGSTRO6M'S measures were given in terms of the Paris inch : I have preserved these ; and have also given their equivalents in terms of the millimetre .
I have only to add that , in my original communication to the Royal Society , dated 1867 , March 2 , the modifications of the first computed numbers , now introduced after consideration of DITSCHEINER'S later measures , had not been applied .
I have now incorporated them in the finished Tables , with the permission of the President and Council of the Society .
G. B. AIRY .
Royal Observatory , Greeniwich , 1867 , October 115 .

