image is cut off sharply by the shadow of the shutter. Strong lines extend into the shadow more or less, and if there is a real reversal the extension of the reversed part into the shadow is trumpet-shaped (a), whereas if it is only a pseudo-reversal it is closed (b).

We would call especial attention to the enormous expansibility of the magnesium line w. l. 2852. The limits are quite indefinable either up or down the spectrum and the absorption (i.e. the reversed line) is a broad band, sometimes reaching up to the solar line U as seen in Plate I. fig. 2, which is a negative photograph of the same part of the arc as fig. 1, but had a fragment of magnesium dropped into it at the moment it was taken. You might fancy that magnesium were the substance which produces the absorption of the sun's rays above U, if it were not that the light reappears in the case of magnesium above the dark band as well as below it while the sun's light seems to be quite cut off above U. The cessation of the solar spectrum above the magnesium line may be due to the general absorption of the lines of iron and other metals which are far more numerous in that region than in the visible part of the spectrum, and it is just possible that a very long exposure of sensitive plates may yet reveal some extension of the solar spectrum beyond U.

In these cases of self-reversals the light of the line is only partially absorbed but there are other cases of

II. Complete reversal, where the lines are seen only as dark lines. To this class belong the reversals before described when the lines of volatile metals were seen dark against the bright background of the hot bottom and sides of the tubes heated in a furnace. But the arc itself does not generally give such a background. It gives generally a discontinuous spectrum consisting of bright lines with only a very faint continuous spectrum. There are however some parts of the arc, as we used it, which are so full of fine closely set lines that it serves as a good background against which a dark reversed line can be seen. The tail of fine lines formed by the ultra-violet Cyanogen bands is such a region. Another is in the neighbourhood of Q where there are a great many fine iron, manganese and chromium lines, and where also the fine lines of the water spectrum add to the effect so as to give something like a continuous spectrum.

One of our photographs shews the reversal of a Thallium line near L. In the negative photograph it is a simple absorption line. Another photograph shews silver lines near Q reversed as mere white lines on the general spectrum. In some cases we have succeeded in getting a good background by bringing up the hot carbon pole into the line of view. In that case the larger part of the bright metallic lines disappear and the con

tinuous spectrum becomes much stronger, so that the reversed lines are more readily seen. We have various plates shewing complete reversals of the lines of several metals in one or other of these ways. The ultra violet lines of potassium form a series extending at decreasing intervals to about the solar line U. They are all reversed, appearing only as white lines in the (negative) plates. The least refrangible is a double line just below O. Probably they are all double in reality, and harmonically related to the violet pair. Plate II. shews most of these potassium lines reversed.

In other plates are seen complete reversals of copper and silver, and lead lines. Some of our plates shew complete reversals of a great many iron lines. In two of these plates the reversals were obtained by putting an iron wire into the arc through one of the carbons, which was perforated for the purpose. Some of the lines are greatly expanded, but they appear only as absorption lines; they have no bright wings. In another case the iron was introduced into the arc as ferrocyanide of potassium, and the lines are sharp clean fine lines just like the Fraunhofer lines in the solar spectrum, the continuous background being heightened by the potassium, which seems to give some continuous spectrum in the ultra-violet region. Plate III. shews a great many iron lines so reversed.

III. Another class of reversals are produced by the vapour of one metal seen against a bright background given by the expansion of a bright line of some other metal. For this purpose the expanded lines of magnesium are most effective from their great breadth. We have several plates shewing sharp clean reversals of iron lines against the continuous light of the expanded magnesium line, w. 1. 2852; one shews lines reversed in this way against the bright expansions on both sides of the absorption band of magnesium. One of the plates shews some iron lines near M and chromium lines reversed in this way. Plate I. fig. 2 shews the principal iron lines between S and U reversed against the expanded magnesium line w. 1. 2852.

IV. A fourth mode of obtaining reversals we cannot so easily explain. It is by allowing a very gentle stream of hydrogen, coal gas, or ammonia, to pass into the arc either through one of the carbons perforated or through a separate opening into the crucible. The effect of such a current is generally to sweep away most of the metallic lines, make those which remain fainter, and increase, relatively at least, the continuous spectrum, and reverse many lines. The character of the effect is fairly seen in a photograph, Plate IV., of the part of the spectrum near U, in which the magnesium line (2852) appears about as strong as usual in a magnesia crucible, but wholly as an absorption, while the other magnesium lines are still bright. Many of the iron lines are reversed in this photo

graph, and so are some less refrangible calcium lines. No doubt many of the other lines which have disappeared are not seen because absorption just balances emission. Indeed by varying the current different effects may sometimes be produced, the same line appearing at one time bright, at another time dark, and at another time not at all. We have seen the calcium lines at H and K, and the strong blue triplet just below the well-known (indigo) calcium line at wave length 4226 all bright, while the indigo line itself was quite invisible, it was neither bright nor reversed. This and similar effects are no doubt due to differences in emissive power at the different temperatures of the emitting and absorbing vapour. At the higher temperature the emissive power of calcium for H and K and the blue triplet is relatively greater than for the indigo line, while at the lower temperature of the absorbent vapour the reverse holds; so that the emission for the indigo light is balanced by absorption, while that of the other rays is not balanced. In like manner we have frequently seen H when K was quite invisible, sometimes K reversed when H was not reversed.

It is not at first sight easy to explain the action of the gas. Probably hydrogen is the chief agent in the case, for a mere current of air has no such effect. The gas may act by helping to diffuse the metallic vapours, diminishing their density in the arc, and increasing it in the tubular part of the crucible without too much lowering the temperature. Hydrogen gas also maintains a reducing atmosphere, .preventing the oxidation of the metallic vapour. It also forms compounds with some metals, notably with magnesium and the alkali metals, as well as with the carbon of the electrodes, which are probably all endothermic, and whether that be so or no, must affect the distribution of the temperature in the arc and tube. In the case of expanded bright lines, when they are not wholly swept out by the gas the effect is to diminish the wings, and as absorption does not begin at the wings, this can only be by diminishing the range of emission, probably by diluting the metallic vapour. The amount of heat absorbed by such a small mass of gas will not lower the temperature much. Moreover a current of chlorine has usually the opposite effect to that of a current of hydrogen, increasing the strength of the bright lines, probably by assisting the volatility of the metals, and so increasing the quantity in the arc. That the explanation here offered of the action of hydrogen is correct, is borne out by the behaviour of mixtures of metals. Thus an alloy of zinc with a little lead gives far sharper and cleaner reversals of the lead lines than lead alone does. When lead alone is put into the arc in the crucible the lead lines come out very strong and diffuse, and the emission is not nearly balanced by the absorption in the (comparatively) cool tube; but when the alloy with zinc is used the lead lines are

much less expanded and the emission much more nearly balanced by the absorption.

By inclosing the arc in a block of lime or magnesia we have found its steadiness very greatly increased, and the mass of metallic vapour which can be maintained at a temperature approaching to that of the arc much enlarged, but it cannot be said that that temperature is at all under control, and the walls of the crucible are almost always cooler than the contents.

V. We have lately used a carbon tube, passed quite through a block of lime, and made it the positive electrode of the current, while the other electrode is a carbon rod passed into the lime at right angles to the tube so as to meet it in the middle.

In this way the outside of the tube becomes intensely heated, the heat is retained by the jacket of lime, and the interior of the tube gradually rises in temperature, and attains in the central part a very high point. By stopping the arc it can be made to pass through the same stages of temperature in the inverse order. Observations are made by looking down the perforation. When the light issuing from the tube is projected by a lens on to the slit of a spectroscope, the heated walls of the tube give at top and bottom a continuous spectrum, against which various metallic lines are seen reversed, while in the central part, when the tube is open at the farther end, the spectrum is discontinuous, and the metallic lines seen reversed against the walls at top and bottom appear as bright lines. Plate V. is a negative photograph shewing this effect.

The apparatus employed is thus constructed: A rod of carbon, a in the figure, 15 millims..in diameter, perforated down its axis

with a cylindrical hole 4 millims. in diameter, is passed through a hole in a lime block d, and is connected by means of a copper clip with the positive electrode of a Siemens dynamo-electric machine;

lines H and K is peculiar. These lines are often absent altogether, when the line wave-length 4226 and the two near M are well seen, and when the two aluminium lines between them and many of the iron lines are sharply reversed. Even the introduction of a small quantity of metallic calcium or calcium chloride into the tube did not bring them out reversed. They were only seen as bright lines, not very strong, when the small rod was removed. The lithium lines at 4603 and 4131 are often bright when many other lines in the neighbourhood are reversed, and must therefore be regarded as relatively difficult of reversal. As a rule the lines less refrangible than 4226 are balanced as to their emissive and absorptive power and, therefore, disappear, while the more refrangible are reversed.

Both the indium lines 4101 and 4509 are persistently reversed, and so are several lead lines. Tin gives lines, of which some are reversed, in highly refrangible regions besides a channelled spectrum, and silver gives a fine fluted-looking spectrum in the blue. Chloride of calcium gives a striking set of six or seven bands between L and M, which may be seen both bright and reversed.

VI. Occasionally a double self-reversal of lines occurs, i.e. in the middle of the reversed or absorption line a bright line appears. We have many times seen this but not often photographed it. We have a plate which shews it in the case of the magnesium triplet between K and L. The photograph is that of the arc itself projected on to the slit of the spectroscope by a lens, and was taken at the moment when a piece of magnesium was dropped into the crucible. This has caused such an expansion of the bright lines that they have extended beyond the nearest two cyanogen bands and caused them to be reversed. On this bright background the magnesium lines themselves are reversed as a dark band in the centre of the field, and on this dark band are three narrow bright lines forming the second reversal. It is not difficult to understand how this occurs. The magnesium is very quickly evaporated and part of the vapour is driven by the expansion up the tube of the crucible before it has got very hot, and it drives before it some hotter vapour previously in the hottest part of the crucible. The cool vapour stops the radiation of the arc and the hotter vapour in front of it of course appears bright. The effect is little more than momentary. In this case the cool vapour was only just around the arc, for at top and bottom the same narrow lines which are bright in the middle of the field (black in the negative photograph) are reversed in the top and bottom.

It is worthy of remark that the facility of reversal of lines appears, speaking generally, to increase as the wave-length diminishes. This is very notable in the case of iron. It seems