November i, 1894] 



NA TURE 



not three, but ten or more plates, we may well judge 

 (since there seems no possibility here of systematic error) 

 that a result, which all confirm, is reliable, and that, on 

 the other hand, a single inflection on one plate, which 

 the other nine unite in repudiating, is due to some 

 fortuitous cause. 



But there is still a higher certainty to be obtained, by 

 a method independent even of comparison or the 

 exercise of judgment. It is founded on the well- 

 known process of composite-photography, where, in 

 photographing the successive members of an assem- 

 blage of persons, having similar general characteristics, 

 as of race, character or education, the individual 

 disappears, and the normal type alone remains. In 

 order to apply this method to such results as ours, 

 however, another step in the process must be intro- 

 duced, and this is an interesting one, for the energy 

 curve itself, however valuable, is a comparatively un- 

 familiar method of showing variations in the energy, 

 which we are all alike used to seeing in the visible spec- 

 trum given by linear representations, and not by a system 

 of inflections. 



In describing this new step, which is to give us a 

 linear spectrum in addition to the original curve, it will 

 be desirable to also give evidence of the statement now 

 made, that the present method is capable of recording 

 far minuter inflections than those shown in the curves 

 here exhibited, which, as has just been stated, have been 

 taken only for the purpose of illustrating such more im- 

 portant features, as can be seen and verified by the 

 audience, and especially for showing the agreement of 

 independent observations. The evidence of the capacity 

 of the apparatus to show this detail will best be illus- 

 trated by applying our purely thermometric method to 

 some well-known lines in the visible spectrum, such as 

 the familiar " D '' lines of sodium. I have already stated 

 that ten years ago the bolometer was barely able to dis- 

 tinguish this as a single line. At the present time our 

 little thermometer, as you here see, now shows not only 

 the two "D's" as separate lines, but the nickel line 

 between them. First we have the complex energy curve 

 (Fig. i), where we see successively the inflections due to 

 the motions of the galvanometer caused by the cold m D,, 

 then to the smaller chill from the nickel line (aided per- 

 haps by that from some of the close atmospheric lines), 

 then the chill from Dj. 



Immediately below this curve is the more familiar linear 

 representation of the same subject (Fig. -)■ Now this 

 linear representation, it is most important to observe, has 

 been obtained, not by drawing, but by the subsequent 

 application to the curve of an automatic process, by means 

 of which its indications are reproduced by photogravure, 

 as separate lines, while by the same automatic process 

 the most complex spectral curves can be rendered into 

 their linear equivalents. 



I have no space to enter here on a description of this 

 process, further than to say it is effected by means of a 

 systematically distorted image of the curve, obtained by 

 a special combination of spherical and cylindric lenses. 

 You will see, on minute inspection, that the inflections 

 of the galvanometer curve have been slightly " loaded," 

 to produce a more ettective contrast of light and dark. 

 Except for this, which can in no way affect the position 

 of a line, but only its intensity, the whole process is as 

 absolutely automatic as any photograph of the visible 

 spectrum. 



This thermograph of the " D " lines has been chosen 

 to indicate the grasp of this new thermometric method, 

 by applying it to the test of an object in the visible 

 spectrum, with which every physicist here is doubtless 

 familiar. He may then be invited to recall that the 

 distance between the " U's " in a rock-salt Oo-degree 

 prism is about eleven seconds of arc, and to observe that 

 two lines about half this distance apart are here shown 



NO. 1305. VOL. 51] 



as sharply divided by this thermal method, as, for 

 instance, are the components of the double star 

 a Geminorum by a three-inch achromatic. Obviously, 

 then, our method could indicate the existence of two 

 lines, little, if any, more than one-quarter the distance 

 between the " D's." Lines 3' or less apart can then 

 evidently be indicated by this method, even in its present 

 stage of development. 



And now, returning to what has been said about the 

 evidence obtainable as to the perfect coincidence of 

 these inflections in different energy curves obtained at 

 different times, and to the consequent evidence that each 

 inflection so given is real, and not the product of an 



Fig. :. 



t'. 



accidental variation in the curve, we may conceive that 

 from any number of such independent curves, any 

 number of such linear representations of the spectra 

 have been obtained ; for example, that ten such linear 

 representations of the whole spectrum as are nere given 

 of the U lines only, have been so found from ten com- 

 plete energy curves taken on as many different days. 

 From these ten linear representations, by the well-known 

 processes of composite photography, one final photo- 

 graph of the spectrum is formed, and on this it is 

 evident we may expect to find only what is permanent 

 and not what is accidental, granting that a rare accident 

 may have introduced an occasional abnormal deflection. 

 Xow, considering that the part of the infra-red solar 

 spectrum of rock-salt under review extends through 



