In the last class of stars to which I have referred, the fourth, the lines have given place to fluted bands, at the same time that the light and colour of the star indicate that we have almost reached the stage of extinction.

These facts have long been familiar to students of solar and stellar physics; indeed, in a letter written to M. Dumas, December 3, 1873, and printed in the Comptes Rendus, I thus summarised a memoir which has since appeared in the Philosophical Transactions:

Il semble que plus une étoile est chaude plus son spectre est simple, et que les éléments métalliques se font voir dans l'ordre de leurs poids atomiques.1

Ainsi nous avons :

1. Des étoiles très-brillantes où nous ne voyons que l'hydrogène, en quantité énorme, et le magnésium;

2. Des étoiles plus froides, comme notre Soleil, où nous trouvons :

Hydrogène + Magnésium + Sodium

Hydrogène + Magnésium + Sodium + Calcium + Fer, . . . ;

dans ces étoiles, pas de metalloïdes;

3. Des étoiles plus froides encore, dans lesquelles tous les éléments métalliques sont ASSOCIÉS, où leurs lignes ne sont plus visibles, et où nous n'avons que les spectres des métalloïdes et des composés.

4. Plus une étoile est âgée, plus l'hydrogène libre disparaît; sur la terre, nous ne trouvons plus d'hydrogène en liberté.

Il me semble que ces faits sont les preuves de plusieurs idées émises par vous. J'ai pensé que nous pouvions imaginer une dissociation céleste,' qui continue le' travail de nos fourneaux, et que le metalloïdes sont des composés qui sont dissociés par la température solaire, pendant que les éléments métalliques monatomiques, dont les poids atomiques sont les moindres, sont précisément ceux qui résistent même à la température des étoiles les plus chaudes.

Before I proceed further, I should state that while observations of the sun have since shown that calcium should be introduced between hydrogen and magnesium for that luminary, Dr. Huggins' photographs have demonstrated the same fact for the stars, so that in the present state of our knowledge, independent of all hypotheses, the facts may be represented as follows: 2

I have no hesitation in stating my opinion that in this line of facts we have the most important outcome of solar work during the last ten years, and if there were none others in support of them the

1 The old system of atomic weights was the one referred to.

2 Symbols are used here to save space. H-Hydrogen, Ca Calcium, Mg= Mag. nesium, Na Sodium, Fe Iron, Bi= Bismuth, Hg= Mercury.

=

=

conclusion would still stare us in the face that the running down of temperature in a mass of matter which is eventually to form a star, is accompanied by a gradually increasing complexity of chemical forms.

This then is the result of one branch of the inquiry, which has consisted in a careful chronicling of the spectroscopic phenomena presented to our study by the various stars.

Experimentalists have observed the spectrum of hydrogen, of calcium and so forth in their laboratories, and have compared the bright lines visible in the spectra with the dark ones in the stars, and on this ground they have announced the discovery of calcium in the sun or of hydrogen in Sirius.

In all this work they have taken for granted that in the spectrum thus produced in their laboratories, they have been dealing with the vibration of one specific thing, call it atom, molecule, or what you will; that the vibrations of these specific molecules have produced all the lines visible, which they have persistently seen and mapped in each instance.

II.

It is at this point that my recent work comes in, and raises the question whether what has been thus taken for granted is really true. And now that the question is raised, the striking thing about it is that it was not asked long ago.

One reason is this. Time out of mind-or, rather, ever since Nicolas Le Fèvre, who was sent over here by the French king at the request of our English one at the time the Royal Society was established, pointed out that chemistry was the art of separations as well as of transmutations-it has been recognised that with every increase of temperature, or dissociating power, bodies were separated from each other. In this way Priestley, from his 'plomb rouge' separated oxygen, and Davy from potass separated potassium; and as a final result of the labour of generations of chemists, the millionfold chemical complexity of natural bodies in the three kingdoms of nature has been reduced by separations till only some sixty socalled elements are left.

Now this magnificent simplification has been brought about by the employment of moderate temperatures-moderate, that is to say, in comparison with the transcendental dissociating energies of electricity as employed in our modern voltaic arcs and electric sparks.

But, in the observations made during the last thirty years on the spectra of bodies rendered incandescent by electricity, we have actually, though yet scarcely consciously, been employing these transcendental temperatures, and if it be that this higher grade of heat does what all other lower grades have done, then the spectrum

we have observed in each case is not the record of the vibrations of the particular substance which we have put into the arc and with which we have imagined ourselves to be working alone, but of all the simpler substances produced by the short or long series of the 'separations' effected.

The question, then, it will be seen, is an appeal to the law of continuity, nothing more and nothing less. Is a temperature higher than any yet applied to act in the same way as each higher temperature, which has hitherto been applied, has done? Or is there to be some unexplained break in the uniformity of nature's processes?

The definite reason for my asking the question at the present time has been this. The final reduction of four years' work at a special branch of the subject to which I will refer presently, on the assumption that at the temperature of the electric arc we do not get such simplifications,' has landed me in the most hopeless confusion, and if I do not succeed in finding a higher law than that on which I have been working, my four years' work, in this direction at all events, will have missed its aim.

III.

This and other reasons compel me to hold that the answer to the question put is, that these transcendental temperatures do dissociate, and that therefore what has hitherto been taken for granted is, in all probability, not true. But before I proceed to give the reasons for the faith that is in me, I must, at the risk of being both technical and tedious when I should wish to be neither, lead up to the understanding of the arguments I have used.

IV.

The spectroscope, however simple or complex it may be, is an instrument which allows us to observe the image of the slit through which the light enters it, in the most perfect manner. If the light contains rays of every wave length, then the images formed by each will be so close together that the spectrum will be continuous, that is, without break. If the light contains only certain wave lengths, then we shall get certain, and not all, of the possible images of the slit, and the spectrum will be discontinuous.

Again, if we have an extremely complex light source, let us say a solid and a mixture of gases giving us light, and we allow the light to enter, so to speak, indiscriminately into the spectroscope, then in each part of the spectrum we shall get a summation-a complex record-of the light of the same wave length proceeding from all the different light waves. But if by means of a lens we form an image of the light-source, so that each particular part shall be impressed in its proper place on the slit plate, then in the spectrum the different kinds of light will be sorted out.

There is a simple experiment which shows clearly the different results obtained. If we observe the light of a candle with the spectroscope in the ordinary manner, that is by placing the candle in front of the slit at some little distance from it, we see a band of colour-a continuous spectrum-and in one particular part of the band we see a yellow line, and occasionally in the green and in the blue parts of the band other lines are observable. Now, if we throw an image of the candle on to the slit-the slit being horizontal and the image of the candle vertical—we then get three perfectly distinct spectra. We find that the interior of the candle, that is the blue part (best observed at the bottom of the candle), gives us one spectrum, the white part gives us another, while on the outside of the candle, so faint as to be almost invisible to the eye, there is a region which gives us a perfectly distinct spectrum with a line in the yellow. In this way there is no difficulty whatever in determining the coexistence of three light-sources, each with its proper spectrum, in the light of a common candle.

We see in a moment that much the same condition of affairs will be brought about, if, instead of using a candle, we use an electric arc, in which the pure vapour of the substance which is being rendered incandescent fills the whole interval between the poles, the number of particles being smaller and the degree of incandescence being less intense at the sides of the arc. By an easily understood artifice we can throw an image of a horizontal arc on a vertical slit; the slit will give then the spectrum of a section of the arc at right angles to its length. The vapour which exists furthest from the core of the arc has a much more simple spectrum than that of the core of the arc itself. The spectrum of the core consists of a large number of lines, all of which die out until the part of it furthest from the centre gives but one line.

In this way the spectrum of each substance furnishes us with long and short lines, the long lines being common to the more and the less intensely heated parts of the arc, and the short lines confined to the more heated ones only. This is the first step.

V.

It has been necessary to enter thus at length into the origin of the terms long and short lines, because almost all the subsequent work which need be referred to now has had for its object the investigation of the phenomena presented by them under different conditions. The first results obtained were as follows:

-

1. When a metallic vapour was subjected to admixture with another gas or vapour, or to reduced pressure, I found that its spectrum became simplified by the abstraction of the shortest lines and by the thinning of many of the remaining ones. To obtain reduction of pressure, the metals were enclosed in tubes in which a partial

vacuum was produced. In all these experiments it was found that the longest lines invariably remained most persistently.1

2. When we use metals chemically combined with a metalloidin other words, when we pass from a metal to one of its salts (I used chlorine)-only the longest lines of the metal remain. Their number is large in the case of elements of low atomic weight, and small in the case of elements of high atomic weight.

3. When we use metals mechanically mixed, only the longest lines of the smallest constituent remain. On this point I must enlarge somewhat by referring to a series of experiments recorded in the Philosophical Transactions (1873).

A quantity of the larger constituent, generally from five to ten grammes, was weighed out, the weighing being accurate to the fraction of a milligramme; and the requisite quantity of the smaller constituent was calculated to give, when combined, a mixture of a definite percentage composition by weight (this being more easily obtainable than a percentage composition by volume).

The quantities generally chosen were 10, 5, 1, and 0.1 per cent. In a few cases, with metals known to have very delicate spectral reactions, a mixture of 0.01 per cent. was prepared.

Observations were then made of the spectrum of each specimen, and the result was recorded in maps in the following manner :First, the pure spectrum of the smallest constituent was observed, and the lines laid down from Thalén's map.

The series thus mapped were as follows:--

The observations showed that the lines of the smallest constituent disappeared as the quantity got less. Although we had here the germs of a quantitative spectrum analysis, the germs only were present,

i In the case of zinc the effect of these circumstances was very marked, and they may be given as a sample of the phenomena generally observed. When the pressure-gauge connected with a Sprengel pump stood at from 35 to 40 millimètres, the spectrum at the part observed was normal, except that the two lines 4924 and 4911 (both of which, when the spectrum is observed under the normal pressure, are lines with thick wings) were considerably reduced in width. On the pump being started these lines rapidly decreased in length, as did the line at 4679,-4810 and 4721 being almost unaffected; at last the two at 4924 and 4911 vanished, as did 4679, and appeared only at intervals as spots on the poles, the two 4810 and 4721 remaining little changed in length, though much in brilliancy. This experiment was repeated four times, and on each occasion the gauge was found to be almost at the same point, viz.:

1st observation, when the lines 4924 and 4911 were gone, the gauge