progressed, she would move across its interior, until she finally reached its other edge, at which instant the totality would end." Dr. Curtis remarks, that "the notion that the corona may be the luminosity of our own sunlit atmosphere beyond the belt of totality is also both theoretically impossible and practically proven false by the testimony of the photographs. As to the theoretical considerations, it is sufficient to point out that with the diameter of the moon's shadow upon the earth two or three times as great as the vertical extent of our own atmosphere, it is geometrically impossible for an observer near the centre of that shadow to see any portions of our atmosphere which lie beyond the cone of darkness-which portions alone, of course, could under the circumstances be illuminated -in apparent contiguity with the moon's limb."

The evidence against the theory derived from this simple consideration,-first pointed out by Mr. Baxendell, I believe, -is so simple and so convincing, that it seems useless to consider further arguments.

But then there is another mode in which the theory has recently been defended. It is pointed out that it is a mistake altogether to imagine that the advocates of the atmosphericglare theory had overlooked considerations so simple and so obvious. They had not imagined, it would seem, that the solar rays pass directly into the moon's shadow-cone, but that these rays are introduced into the shadow-cone by means of a possible action exerted near the moon's limb.* The theory is thus made to resemble La Hire's, described in these words by Prof. Grant:-"La Hire suggested that the corona might be produced by the reflection of the solar rays from the inequali

Mr. Lockyer remarks, that both Dr. Gould and M. Faye have expressed such an opinion. It is possible that M. Faye may; but I may venture to say, very confidently, that Dr. Gould is not an advocate of the atmosphericglare theory of the corona, and has advanced no line of reasoning in its support. He has pointed to certain peculiarities observed during the American eclipse (respecting which, however, Dr. Curtis remarks that all other observers are at issue with him); and he remarks respecting these, that they seem to point to parts of the corona as belonging to our atmosphere. But he told me distinctly, after the last meeting of the Royal Astronomical Society, at which he had been present as a visitor, that he had no theory of the corona, and was content with stating what he had seen.

What Dr. Gould has remarked about the moon's possible action is, that the apparent encroachment of the prominence-bases on the lunar disc may be due to specular reflection of the moon's surface. This view has no bearing whatever on the subject of the corona; but if it had, then Dr. Curtis's proof that the encroachment referred to is a purely photographic phenomenon―a view confirmed by Dr. Mayer-would serve to dispose of any reasoning founded on the observed fact.

ties of the moon's surface contiguous to the edge of her disc, combined with their subsequent passage through the terrestrial atmosphere." Prof. Grant, after discussing De Lisle's evidence against La Hire's view, remarks, that clearly "the hypothesis was untenable."

Let us, however, apply what has been already inferred respecting the moon's shadow-cone. Replacing our observer at the bottom of the imagined well of shadow, we have now to consider the case of light making its way into this well through the deflection of the solar rays. But we have one certain fact about the region of shadow. The moon looks black during total eclipse, and therefore it is abundantly evident that lines taken from the eye of the observer to the edge of the moon's limb include within them a cone which is not illuminated. Now we have seen that, taking only the case of undeflected

A

C HE

rays, we should have a shadow 170 miles in width at the top and 150 at the bottom, while the cross-section at the top would subtend about 93 degrees, the moon only subtending about half a degree. Hence the cone of blackness between the eye and the moon's disc occupies a position quite clear of the imagined walls of our shadow-well. The shadow-well, in fact, is shaped somewhat as ABCD in the figure, while the shadow-cone extends from the eye of the observer at E upwards through the middle of the shadow-well, as shown by the black region EF.

Now, before the theory we are dealing with can be accepted, it must be shown how the solar rays, whose direct course would keep them outside ABCD, can be brought within ABCD without trenching at all upon the black cone EF. It is obvious that this is wholly impossible. If we get the rays within ABCD, they can only avoid EF by travelling parallel to its surface (it is to be remembered that they must come quite up to its surface if the corona is to be accounted for); and it would thus follow that the rays which pass the moon's edge are deflected exactly to E. But obviously, though this might happen for a moment during some particular eclipse, it could not by any possibility happen in all eclipses and throughout their continuance, since there is absolutely no fixed relation whatever between the point E and the boundary ABCD of the geometrical shadow-cone. E is simply the station where the observer places himself, and is not necessarily on the axis of the shadow-cone at any moment, and is necessarily away from the axis at all moments save

one.

There is, in fact, very little to choose between this form of the theory and the former. It will doubtless be deemed wholly unnecessary to discuss further a theory against which such decisive geometrical arguments may be adduced, especially as, according to its most ardent advocates, the theory has for its main support so vague a conception as that of "a possible action at the moon's surface."

We are reduced, therefore, to accept the sole remaining theory that the corona is a solar appendage; and our subject is proportionately enhanced in interest, since so viewed the corona becomes the most extensive region within the solar system. Granting to it no greater apparent extension than half a degree from the moon's limb, it must yet have a diameter three times as great as the sun's, and therefore a volume (enclosing his globe) twenty-seven times as great. But we may fairly assume that the greatest observed extension of the corona falls far short of its true boundary, since the state of the atmosphere tends importantly to affect the extent of the corona's outline, always reducing it, but not always by the same amount. And since the corona has been observed to extend so far as 8 degrees from the eclipsed moon, or to a distance exceeding some fifteen times the moon's diameter, we should have to assign to the corona a volume exceeding about 30,000 times that of the sun.

But it seems clear that we cannot regard the corona as a solar atmosphere. The arguments founded by Mr. Lockyer on the laboratory experiments of his eminent ally, Professor Frankland, seem to me to be wholly convincing on this point. The pressure of the solar atmosphere at the level of the summits of the prominences must be exceedingly small, since, even near the base of the chromosphere, the pressure is not considerable. From the researches of Wüllner it would seem that, near the base of the chromosphere, the pressure corresponds to a barometric height of between 50 and 500 millimètres-that is, roughly, between two and twenty inches.* Eighty or a hundred thousand miles or so above this level, the pressure

* It seems to me open to doubt whether at the real base of the chromosphere—that is, at the very level where the photosphere and the chromosphere meet-the atmospheric pressure may not exceed, and that enormously (even many thousand fold), the pressure estimated from the observed width of the hydrogen lines at the apparent base of the chromosphere; for the most powerful telescope yet constructed could not recognise as a sensible quantity, a stratum next the photosphere of even 100 miles in thickness. Yet remembering the compound character of the solar atmosphere at its lowest levels, it would be within such a stratum that by far the greatest increase of pressure would take place.

would be inconceivably minute; so that we can scarcely conceive that, above that level, any atmosphere capable of supplying an appreciable amount of light can exist, still less that any atmosphere can extend to the enormous distances at which portions of the corona have been seen to lie above the sun's surface.

It seems to follow, as an inevitable conclusion from this reasoning, that the corona must consist of some sort of matter -discrete solid or liquid bodies, vaporous masses, or groups in which solid or liquid bodies are intermixed with vaporous masses-travelling around the sun. There appears to me absolutely no escape from this conclusion, when we merely consider the evidence against other theories; for I can conceive no other theory of the corona besides this one and those others which we have been forced to reject.

The negative evidence thus bringing us to this particular theory, however, it will be well that we should enquire how far we have positive evidence in its favour.*

In the first place, let me invite attention to what Leverrier has demonstrated respecting the motions of Mercury. The secular motion of the perihelion is not such as it should be if the whole mass of matter within the orbit of Mercury were within the visible boundary of the sun. The conclusion, to which Leverrier has been led by this circumstance, is that a zone or band of small planets exists within the orbit of Mercury. A single considerable planet would produce changes in the motions of Mercury in his orbit, not a change such as is actually observed in the orbit itself. A single planet, then, is not in question, but many. The change, in fact, resembles that produced on the orbit of Mars by the family of asteroidal planets.

Now we cannot neglect such evidence as this, because it is precisely that sort of evidence which has been found most reliable. Unless observation has erred in a systematic and most singular manner, there is a family of small bodies within the orbit of Mercury. What the nature of the family may be Leverrier's researches do not tell us; but, as observations

In a review of my "Other Worlds" in the Quarterly Journal of Science (a review with which I have every reason to feel satisfied, and which has obviously been written by a thoughtful student of science), it is urged that I am too anxious to show that every point of evidence favours a theory of mine-though some points may be urged as well in favour of other theories. I cannot too strongly express my conviction that this anxiety is the very essence of safe theorising. If a theory is clearly opposed by one single point of evidence, it must go overboard, even though a thousand other points seem to favour it.

made during eclipses have revealed no bright points of light,* we may fairly conclude that there are many very small bodies instead of a few bodies of considerable dimensions. Crowds of such small bodies could scarcely fail to be discernible by their combined light during the total obscuration of the sun.

Thus, by an independent line of reasoning, it has been shown that light might be looked for where the corona is actually seen. But, again, Mr. Baxendell has been led to infer, from observed meteorological and magnetic changes on the earth, that a zone or ring or disc of matter surrounds the sun, extending to a distance closely corresponding to that assigned by Leverrier to the family of planets. The evidence adduced by Baxendell is of a very striking character, and seems almost inexplicable save on the hypothesis he adopts. And clearly, if such a zone or disc of matter exists, we might expect to see it during total eclipse.

Here, then, is a totally distinct reason for expecting that some such object as the corona would appear when the sun is eclipsed.

Then there is the zodiacal light-demonstrably not a terrestrial phenomenon, since it rises and sets with the celestial bodies as the earth rotates. We see this light during the twilight hours; we recognise a gradual condensation towards its core, and still more markedly towards the place of the concealed sun. Is it reasonable to believe that this condensation is suddenly checked before the sun is reached? Or if we assumed this, must we not yet believe that the zodiacal matter forms at least a zone around the sun? so that, despite the imagined gap between it and the sun, its light would in appearance reach to the sun's place, did the background of dark sky but reach so far.

Here, then, is a new reason for expecting that during totality light would be seen around the sun.

Then, again, there are the meteor systems, of which the earth encounters more than a hundred, all whose perihelia must necessarily lie within the earth's orbit. The chances are so enormous against the earth's encountering one such system, unless many millions existed, that we are forced to conclude that there are many millions for each system actually encountered by the earth. This is, be it remarked, not a hypothesis, but an inevitable conclusion. These meteor systems, illuminated as they must be by the sun, would be visible where the corona is, even if the perihelia lay as far from the sun as the earth's orbit. But passing, as many must, quite close to him, they must be illuminated yet more brilliantly-probably rendered incandescent, or even vaporised, by intense heat.

* With perhaps one exception during the American eclipse,