ON OUR KNOWLEDGE OF SPECTRUM ANALYSIS. 293 



hypothesis, which may explain the behaviour of long and short lines as 

 the hypothesis of molecular shocks, for according to it the short lines are 

 brought out by a greater intensity of molecular shocks. 



There is, however, another way of looking at these phenomena, which 

 is advocated by Mr. Lockyer. If we carefully examine the spectnim of a 

 metal, an image of the spark being projected on the slit of the spectro- 

 scope, and if we observe the changes which a spectrum undergoes when 

 the temperature or pressure is altered, we cannot fail to be struck by the 

 fact that we can generally divide the lines into two, or sometimes perhaps 

 more than two sets, the lines in each set varying together. _ The question 

 forces itself on the observer whether we have not to deal in one and the 

 same spark with two or more overlapping spectra which vary relatively to 

 each other. The faint lines which stretch sometimes like ghosts away 

 from the poles into the centre of the spark, would belong to one spectrum, 

 the short brilliant, often winged, lines, which are only confined to the 

 neighbourhood of the pole, would belong to another. 



A spark would show that set of lines strongest which belongs to the 

 molecular grouping which is present in the largest quantity within the 

 spark. The relative quantity of different molecules may vary with the 

 distance from the pole, and thus a line which is strongest in the centre of 

 the spark may be weakest near the pole. If a spark is taken from different 

 chemical compounds, or from alloys, that set would show which is due to 

 the particular grouping in which the element is contained in the com- 

 pound. We shall speak of this hypothesis as the hypothesis of molecular 

 combination. Both suppositions which we have mentioned, and to which 

 we shall have to refer again, express the facts fairly well, but neither of 

 them is free from difficulty. 



V. Other Changes in the Relative Intensity of Lines. 



"We have given in the preceding pages a method by means of which 

 the study of a spectrum shown in one given spark, will indicate to us its 

 behaviour under a great many diffei'ent circumstances. What we have 

 said is true within sufficiently wide limits to render the method an 

 extremely valuable one, but if the range of temperature or pressure within 

 which our experiments are made is pushed beyond a certain point, further 

 considerations will have to be taken into account. That the method 

 must or may ultimately break down appears both from the experimental 

 results wliich we have given and from the two possible theoretical ex- 

 planations which we have mentioned. We have quoted, for instance, the 

 behaviour of some zinc lines, when the pressure at which the spark is 

 taken is reduced. We have seen that while three long zinc lines have 

 remained comparatively unaffected, two equally strong but shorter lines 

 near it rapidly decreased in length and finally disappeared. Now suppos- 

 ing that instead of decreasing the pressure we had increased it, the lines 

 Avhich rapidly decreased in length would increase more rapidly than the 

 others, and thus, unless all lines tend towards one fixed limit, the shorter 

 lines might finally outgrow the long ones. At that point the method of 

 long and short lines would fail to give us correct results. Mr. Lockyer * 

 has drawn attention to another cause which renders the method unsafe, 

 if the temperature is pushed beyond a certain point. According to the 

 two hypotheses, which, as we have seen, fairly well account for the facts, 



' Proc. Boy. Soc. xxviii. p. 157 (1879). 



