the equator, as I was able to observe directly in the spectroscope with certainty. At the nearest place to the nucleus there appeared a significant displacement of the spectral line toward the red. This was measured at 4h 55m and indicated a recession of 470 km per second. While I was showing all this to my well

informed visitor, a protuberance rose up at 118°, which was at first also visible in the line 6677. It was inclined toward the equator at an angle of 60°, thus directed away from the spot region, and at 5h 3m had already attained a height of 128". Further observations were impossible. At no time, either on September 2 at 2 P. M., or at three observations on the 3d, could I see a prominence over the spot region, a fact which has

already been reported by others. I find in this only a confirmation of the known fact that quiescent prominences do not occur over areas of spots-which are the scene of transitory eruptive phenomena.

As to the further development of the spot I would only remark that the nucleus made a rotation of about 60° in the positive direction from the 5th to the 12th of September, and that the group did not develop strongly until the 9th. The observations of September 5 seem to me to be of especial significance. Between 11 and 12 and again at 1, I let the spot run over the slit. The Ha line was very bright over a portion of the nucleus and out beyond it. I could readily recognize the parts of the nucleus, and sketch the form of the red mass. Fig. 2 is intended to represent the outline of the spot and the parts of the nucleus. The reversed portion is indicated by the shading. This appeared at the same time to be brown in the helioscope, in contrast to the bluish-black color which spot nuclei usually have in my refractor. This reversal of the Ha could also be seen beyond the spot at a distance of a few degrees, particularly in the places marked e and f. The location of the latter were not accurately fixed by measures, but hastily estimated in the field of view by the eye. These two points e and ƒ, as well as d, near the nucleus, exhibited the rare phenomenon of a considerable and regular displacement of the dark Ha line on the Sun's disk. The peculiarity of this disturbance is best shown by reproducing the observed forms themselves, as they are given in Fig. 3, with the letters in question designated. The shaded portions denote the bright reversal which occurred simultaneously. If we should seek an explanation of these bright reversals, probably the simplest would be the assumption at these points of violent metallic eruptions, which are known to be extremely luminous, and thus could overpower the photosphere in monochromatic light. At these bright places in fact the dark line was not merely enlarged and broken, but the place was brighter than the surrounding spectrum, especially over the nucleus. We are thus also obliged to transfer the origin of eruption within the

spot to the very bright bridge there, for the reversal did not extend beyond the limits of the spot.

The very remarkable displacement of the dark Ha line and its doubling on the solar disk is confirmed by the measurement near d, where it corresponded to a recession of 108 km. These peculiar appearances were seen on this day only, from 10 to 12h, nothing of the sort being visible on the days following. I find it expressly noted in the observing book that no unusual phenomena and no reversals were visible at three times of observation on September 6th; on the afternoon of the 9th, the day of the aurora, no reversals were seen; and similarly on the 12th and 14th. The spot disappeared on the 15th, and on the particular point on the Sun's limb could be seen only the bright flames and stripes which as a rule characterize the departure of a region of spots.

The simultaneous occurrence of such extraordinary phenomena of a similar nature at a considerable distance from the spot area justifies the conclusion that the cause and source of these powerful disturbances, as well as the common rotation of the group of spots, are not to be sought in a comparatively shallow layer of the surface, but rather in decided depths of the body of the Sun.

In the course of the year 1899 I have otherwise observed such a reversal on the disk only on March 12 and 14, and then less strongly marked. It was situated over the last spot of the large group then visible. A great magnetic disturbance was also reported at that time on March 14. A close relation between these two phenomena is indicated by the fact that according to the magnetic observations at Stonyhurst only two very large magnetic disturbances have occurred during this year, and that both fell at the time when I observed these extraordinary phenomena over the spots each time following a few days thereafter. This is also confirmed by the magnetic observations at Pola, which registered at these times the only great disturbances in declination.

HAYNALD OBSERVATORIUM.
June 1899.

A SPECTROSCOPE OF FIXED DEVIATION.'

BY PH. PELLIN AND ANDRÉ BROCA.

It is well known that when a ray of light has traversed a prism at minimum deviation its direction, after reflection on a mirror attached to the prism, is independent of its color. This is readily understood when it is remembered that the direction of a ray which has traversed a prism at minimum deviation is the same as if it were reflected from its base. If it is then made to undergo reflection the result will be the same as if the ray were reflected on two mirrors making an angle with each other. The final ray will make with the incident ray an angle double that between the two mirrors.

This principle has frequently been employed in bolometric researches where a definite dispersion and a fixed receiving apparatus were required. To realize these conditions it is only necessary to rotate a system composed of a prism and an attached mirror in order to cause the dispersed radiations, at minimum deviation, to pass in succession over the bolometer.

In this form the apparatus is not suitable for practical use in a spectroscope, for silvered glass mirrors give a bad image and metallic mirrors are soon unfit for use. It is therefore necessary to devise a prism giving total reflection on one of its interior surfaces in such a manner as to realize a fixed deviation without the use of a perishable mirror. It must be shown at the outset that under these conditions the property of fixed deviation for a ray twice refracted at the desired angle is preserved.

Imagine a prism of 60°, A B C, and a mirror M (Fig. 1). Consider the ray x P Q Ry, and let us endeavor to replace the external reflection M by a total reflection. The emergent ray must take the direction Ry; since we may replace QR by its symmetrical with respect to M, and since we wish to preserve 'Séances de la Société Française de Physique, No. 1, 1899.

the same angle of refraction Q, we must take as the surface of emergence a surface symmetrical to AC with respect to the mirror M. We must therefore take a face parallel to AT. Under these conditions it is evident that the ray which is twice refracted at the same angle at P and Q' will have precisely the same properties as if it had been refracted at P and Q in the prism BA C.

Let us now consider a 60° prism to be cut by a plane A' C, such that the angle B C A' shall be 45°, which gives B A' C-75°;

let us join to it a rectangular prism BCD, having acute angles of 30° and 60°, and we shall obtain a prism having an internal reflection and a fixed deviation of 90° which is equivalent to a prism of 60°. It is moreover evident, since the angles DQ'y, x PB are equal, that the faces A' B, B D, must be at right angles. In the actual construction the faces may be limited to the parts actually used. If we assume an aperture B A', we shall obtain a useful prism by joining to B A' C the prism E' CD'. But this requires the use of cement, which should be avoided if possible. We are thus led to the solution which consists in cutting a single block BACD. This requires a slight increase in the thickness of glass to be traversed, but it avoids the necessity of cementing the surfaces, which is far more important.

Such a prism renders possible the construction of a very