because, from the existence of several 'critical points,' and great variations due to other causes, the results obtained were not constant.

In a subsequent research on the gold-copper alloys used in the coinage, Mr. Roberts, the Chemist of the Mint, and myself were able to show that the shortening in the length of the lines by reduced quantity was such a definite physical effect following upon reduced quantity, that a difference of 1000 part of copper in gold could be detected.

VI.

We are now in possession of the facts utilised in the work which has led up to the subject discussed in the present paper.

They have been utilised along two perfectly distinct lines of thought.

(1) They have been used in an attempt to enable us to produce a spectrum of a substance free from lines due to the impurities which are almost always present.

(2) They have been used to indicate the existence and amount of dissociation when acknowledged compounds have been submitted to the action of different and increasing temperatures.

I will deal with (1) first.

The purification of the spectrum is conducted as follows:-The spectrum of the element is first confronted with the spectra of the substances most likely to be present as impurities. This is most conveniently done by photographing the spectra on the same plate one above the other, so that common lines are continuous.

The retention or rejection of lines coincident in two or more spectra is determined by observing in which spectrum the line is thickest; where several elements are mapped at once, all their spectra are photographed on the same plate, as by this means the presence of one of the substances as an impurity in the others can be at once detected.

Lines due to impurities, if any are thus traced, are marked for omission from the map and their true sources recorded, while any line that is observed to vary in length and thickness in the various photographs is at once suspected to be an impurity line, and if traced to such is likewise marked for omission. I give a case.

The two lines H and K (3933 and 3968), assigned both to iron and calcium by Ångström, are proved to belong to calcium in the following way :

a. The lines are well represented in the spectrum of commercial wrought iron, but are absolutely coincident with two thick lines in the spectrum of calcium chloride with which the iron spectrum was confronted.

b. The lines are represented by mere traces in the spectrum of a specimen of pure iron prepared by the late Dr. Matthiessen. Both

the poles between which the arc passed were of iron, one consisting of an ingot of the metal which had been cast in a lime-mould.

c. The lines are altogether absent in a photograph of pure iron, where both the poles were of the pure metal not cast in lime, and they are likewise absent in a photograph of the spectrum of the Lenarto meteorite.

By eliminating lines due to impurities in the manner just described, a spectrum is at length obtained, of which every line is assignable to the particular element photographed, the same temperature being employed in the case of all the elements observed.

With regard to the second line of work, I should commence by stating that from a beautiful series of researches carried on by several methods, Mitscherlich concluded in 1864, that every compound of the first order heated to a temperature adequate for the production of light, which is not decomposed, exhibits a spectrum peculiar to this compound.

In some experiments of my own, communicated to the Royal Society in 1873, I observed:

First. That whether the spectra of iodides, bromides, &c., be observed in the flame or a weak spark, only the longest lines of the metals are visible, showing that only a small quantity of the simple metal is present as a result of partial dissociation, and that by increasing the temperature, and consequently the amount of dissociation, the other lines of the metal appear in the order of their length with each rise of temperature.

Secondly. I convinced myself that while in air, after the first application of heat, the spectra and metallic lines are in the main the same, in hydrogen the spectra are different for each compound, and true metallic lines are represented according to the volatility of the compound, only the very longest lines being visible in the spectrum of the least volatile compound.

Thirdly. I found that with a considerable elevation of temperature, the spectrum of the compound faded almost into invisibility.

These results enable us to make the following statement :

A compound body, such as a salt of calcium, has as definite a spectrum as that given by the so-called elements; but while the spectrum of the metallic element itself consists of lines, the number and thickness of some of which increase with increased quantity, the spectrum of the compound consists in the main of flutings and bands, which increase in like manner.

In short, the molecules of a simple body and a compound one are affected in the same manner by an increase in their quantity in so far as their spectra are concerned; in other words, both spectra have their long and short lines, the lines in the spectrum of the element being represented by bands or fluted lines in the spectrum of the

compound; and in each case the greatest simplicity of the spectrum depends upon the smallest quantity, and the greatest complexity upon the greatest.

The heat required to act upon such a compound as a salt of calcium, so as to render its spectrum visible, dissociates the compound according to its volatility; the number of true metallic lines which thus appear is a measure of the quantity of the metal resulting from the dissociation, and as the metal lines increase in number, the compound bands thin out.

VII.

These results bring us face to face with the subject-matter of the recent work.

First with regard to impurity elimination. I find that although the method is good for detecting and eliminating impurities, there are still short line coincidences between metals which are pure.

This 'higher law' has come out in the following manner :—

For the last four years I have been engaged upon the preparation of a map of the solar spectrum on a large scale, the work including a comparison of the Fraunhofer lines with those visible in the spectrum of the vapour of each of the metallic elements in the electric arc.

To give an idea of the thoroughness of the work, at all events in intention, I may state that the complete spectrum of the sun, on the scale of the working map, will be half a furlong long; that to map the metallic lines and purify the spectra in the manner described, more than 100,000 observations have been made and about two thousand photographs taken.

In some of these photographs we have vapours compared with the sun; in others vapours compared with each other; and others again have been taken to show which lines are long and which are short in the spectra.

A rigorous application of the system of impurity elimination formed, of course, a large part of the work.

The final reduction of the photographs of all the metallic elements in the region 39-40-a reduction I began in the early part of last year-summarised all the observations of metallic spectra compared with the Fraunhofer lines accumulated during the whole period of observation, and all the results of the impurity elimination.

Now this reduction has shown me that the hypothesis that identical lines in different spectra are due to impurities is not sufficient. I show in detail in the paper the hopeless confusion in which I have been landed.

I find short line coincidences between many metals the impurities of which have been eliminated, or in which the freedom from mutual impurity has been demonstrated by the absence of the longest lines.

The explanation of this result on the hypothesis that the elements are elementary, does not lie on the surface, but it does on the assumption that they are compounds and behave like them.

This is the first point. We now pass from the results brought about at the same temperature with different substances to those observed at different temperatures with the same substance.

I find that when the temperature is greatly varied, the elements behave spectroscopically exactly as compound bodies do, as we have already seen. New lines are developed with increasing temperatures, and others fade in precisely the same way as the metallic lines made their appearance in the salts at the expense of the latter, which faded too.

In short, the observations and reasoning which I formerly employed to show how acknowledged compounds behaved in the spectroscope are now seen to indicate the compound nature of the chemical elements themselves.

In a paper communicated to the Royal Society in 1874, referring, among other matters, to the reversal of some lines in the solar spectrum, I remarked:

It is obvious that greater attention will have to be given to the precise character as well as to the position of each of the Fraunhofer lines, in the thickness of which I have already observed several anomalies. I may refer more particularly at present to the two H lines 3933 and 3968 belonging to calcium, which are much thicker in all photographs of the solar spectrum (I might have added that they were by far the thickest lines in the solar spectrum) than the largest calcium line of this region (4226-3), this latter being invariably thicker than the H lines in all photographs of the calcium spectrum, and remaining, moreover, visible in the spectrum of substances containing calcium in such small quantities as not to show any traces of the H lines.

How far this and similar variations between photographic records and the solar spectrum are due to causes incident to the photographic record itself, or to variations in the intensities of the various molecular vibrations under solar and terrestrial conditions, are questions which up to the present time I have been unable to discuss. The progress of the work has shown that the differences here indicated are not exceptions, but are truly typical when the minute anatomy of the solar spectrum is studied.

Kirchhoff, indeed, as early as 1869 seems to have got a glimpse of the same thing, for in his memorable paper, which may justly be regarded as the basis of all subsequent work, he is careful to state that the sixty iron lines in the sun to which he referred, agree as a rule' in intensity with those observed in the electric spark. Those who have given an account of his work have not always been so cautious. Indeed, I find my friend Professor Roscoe 2 running beyond the record in the following sentence:—

In order to map and determine the positions of the bright lines found in the electric spectra of the various metals, Kirchhoff, as I have already stated, employed the dark lines in the solar spectrum as his guides. Judge of his astonishment when he observed that dark solar lines occur in positions connected with those of all the

2 Spectrum Analysis, 3rd edit. p. 240.

bright iron lines! Exactly as the sodium lines were identical with Fraunhofer's lines, so for each of the iron lines of which Kirchhoff and Ångström have mapped no less than 460, a dark solar line was seen to correspond. Not only had each line its dark representative in the solar spectrum, but the breadth and degree of shade of the two sets of lines were seen to agree in the most perfect manner, the brightest iron lines corresponding to the darkest solar lines.

This statement (the italics are mine) was made to prove the absolutely identical nature of the iron vapour in the sun's atmosphere and in the electric spark. As the statement is erroneous, the vapours can hardly be identical.

VIII.

Such, then, is the reasoning on which I base the two counts in the indictment against the simple nature of the elementary bodies. First, the common lines visible in the spectra of different elements at high identical temperatures point to a common origin.

Secondly, the different lines visible in the spectra of the same substance at high and low temperatures indicate that at high temperatures dissociation goes on as continuously as it is generally recognised to do at all lower temperatures.

In my paper I attempt to show that if we grant that the highest temperatures produce common bases-in other words, if the elements are really compounds-all the phenomena so difficult to account for on the received hypothesis find a simple and sufficient explanation. And with regard to the second count, I discuss the cases of calcium, iron, lithium, and hydrogen. I might have brought, and shall subsequently bring, other cases forward. In all these I show that the lines most strongly developed at the highest temperatures are precisely those which are seen almost alone in the spectra of the hottest stars, and which alone are obviously present in the spectrum of our own sun. This last fact suggests that the sun ought to be brought into court on this matter, because if it be true that the temperature of the arc breaks up the elements, then the higher temperature of the sun should do this in a still more effective manner.

I have put this question to the sun, and I have communicated a second paper to the Royal Society, stating my grounds for the belief that all the solar phenomena we have been watching with our spectroscopes for the last ten years cannot be explained on the existing hypothesis, and that they are simply and sufficiently explained on the

new one.

IX.

What I have next to do, then, is to show how the facts available for this appeal have been got together.