On the other hand, the light of the prominence-matter within p p' is spread only over the three lines shown in the figure (and a few fainter ones), and is therefore proportionately but very little reduced. Hence if we only have enough dispersive power, we can make sure of rendering the prominence-lines visible, for we get the same luminosity for them whatever the length of the spectrum, the only effect of an increase of length being to throw the bright lines farther apart; whereas the atmospheric spectrum which forms the background

FIG. 37.

will obviously be so much the fainter as we spread its light over a longer range.

By this plan we get a certain number of images of a portion of a prominence-a mere strip so to speak; and we can get any number of such portions, and in any direction as compared with the Sun's limb. For example, if s s' (fig. 37) be the Sun's limb, P P' a prominence, we can get from such a strip as s s' the spectrum R V. And obviously since the length of the bright lines tell us the length of the part p p' in figs. 36 and 37, we can, by combining a number of such parallel

strips as s s', learn what is the true shape of the prominence P P'.

But the plan can be applied to show the whole of a prominence. For let us suppose that in place of a narrow strip as s s', in figs. 36 and 37, we have a space such as is shown in fig. 38, through which, but for the intense brightness of the illuminated air, the prominence P P' would be visible. Then the part s s' of the Sun will produce a solar spectrum-altogether impure, of course, on account of the great width of s s', and brighter than the solar spectrum produced by s'p' in the case illustrated by fig. 36 in precisely the proportion that s s' in fig. 38 is greater than s'p' in fig. 36.

All the remainder of the space, including the prominence P P', will give an impure solar spectrum due to the illuminated air, and very much brighter than in the cases illustrated in figs. 36 and 37, because so much more of this light is admitted through the open slit. Three coloured images will be formed of the prominences (other fainter ones need not be considered), one red at C, one orange near D, the other greenishblue near F. These images will be as bright (neglecting variations in the intrinsic brilliancy of the prominence) as the corresponding lines in the cases illustrated by figs. 36 and 37; but they will of course not

L

be so well seen, since the background, as I have said, will be very much brighter than in those cases. They can be made as conspicuous only by an increase of dispersive power; hence the importance of constructing prismatic batteries of great dispersive power.

In connection with this portion of my subject it is necessary to remark that the bright lines seen in the prominence-spectrum are not uniformly wide throughout, but commonly are wider close to the Sun's limb. This circumstance will be referred to more at length further on; but it is proper to state in this place, that this increase of width is held to indicate an increase of pressure, because the researches of Plucker, Hittorf, Huggins, and Frankland have shown that the spectral lines of hydrogen grow wider as the pressure at which the gas exists is increased.

And now, lastly, it remains that I should explain what has been justly regarded as the most wonderful of all applications of the powers of spectroscopic analysis —the measurement of the velocity of recess or approach of stars, or other self-luminous objects moving with very great rapidity.

Reverting to fig. 21, the reader will see that the violet rays are most affected by their passage through the prism, the red rays least. Now, it has been demonstrated by the careful mathematical analysis of the

*

* I refer here to the investigations of Cauchy, and Baden Powell, and others (see specially Cauchy's Mémoire sur la Dispersion), not, of course, to the proof that when the differences of velocity are admitted, differences of refrangibility are accounted for. The latter may be regarded in fact as self-evident.

motions of light-waves that this difference of refrangibility is due to the different velocities with which the longer light-waves forming red light, and the shorter light-waves forming violet light, travel (respectively) through material media. The shorter waves travel more slowly than the long ones, and the difference is the greater according to the density (or approach to opacity) of the medium. So that, in fine, the part of the spectrum formed by light of any order depends on the wave-length of that light; and if under any circumstances the wave-length could be altered, then the light of that order would no longer occupy the same portion of the spectrum, but would pass nearer to the violet end if the waves were shortened, and nearer to the red end if they were lengthened.

Now, so far as we know, it never happens that lightwaves of a certain length are really modified. Precisely as waves of a certain breadth propagated along a canal are not found to change their breadth as they proceed, or, again, precisely as a sound of a certain tone does not change in tone as it travels onwards, so lightwaves of a certain length or order do not as they travel through ether, or through material media, become changed into light-waves of some other order.†

* In the ether of space they travel of course with appreciably equal velocities; otherwise the satellites of Jupiter, after emerging from eclipse, would show the same changes of colour that we see in a metal heated from a red to a white heat.

I have sometimes been inclined to suspect, however, that under certain circumstances of excessive agitation within the substance of the source of light, the wave-length might be altered, precisely as waves travelling along a canal might be modified in length by the action of

But there is a circumstance which may cause the light-waves to appear to change in length. Supposing the source of light is approaching or receding at a very rapid rate—at a rate which bears an appreciable proportion to that of light-then the length of the light-waves must needs appear modified-shortened when the source of light is approaching, lengthened when it is receding. The same will also hold if the observer be carried very rapidly towards or from the source of light. To see that this is so, it is only necessary to consider that more light-waves must necessarily reach the observer in a given time when the source of light is approaching, than when it is at rest (with respect to him), and fewer when it is receding. They must then in one case succeed each other more rapidly, and so seem to be separated by shorter intervals, while in the other they must succeed each other more slowly, and so seem to be separated by longer intervals.*

the cause which gave them birth. When we know that the c line of the prominences has been observed to be tranquil, while the F line has been broken, the idea is certainly suggested that those molecular motions within the substance of the hydrogen of the prominences, which produce that part of the light corresponding to the F line, may by some violent action be so far modified that the observed disturbance of the wave-length corresponding to that particular line may be brought about. It seems difficult to understand how, under any other circumstances, one line of the hydrogen should be undistorted, while the other is, to use Professor Young's description, 'absolutely shattered.'

* The principle on which this brief but sufficient explanation depends admits of several illustrations. I do not know of any which more clearly exhibits the true nature of the principle than one which I employed in the first matter I ever wrote for publication, a paper on 'The Colours of the Double Stars,' which appeared in the Cornhill for December, 1863. I quote the portion referred to:- Let the reader imagine