that, although the appearance of the spectrum may be very different under different circumstances, yet the position of the lines does not depend on the temperature.

"Even the alteration of the mass of the incandescent gas is sufficient to effect a change in the character of the spectrum. If the thickness of the film of vapour whose lines are being examined be increased, the luminous intensities of all the lines increase, but in different ratios. The intensity of the bright lines increases more slowly than that of the less visible rays. The impression which a line produces on the eye depends on its breadth as well as its brightness. Hence, it may happen that one line being less bright although broader than a second is less visible when the mass of incandescent gas is small, but becomes more distinctly seen than the second line when the thickness of the vapour is increased. Indeed, if the luminosity of the whole spectrum be so lowered that only the most striking of the lines are seen, it may happen that the spectrum appears to be totally changed when the mass of the gas is altered. Change of temperature appears to produce an effect similar to this alteration in the mass of the glowing vapour, no deviation in the maxima of light being observed, but the intensities of the lines increasing so differently that those most visible at a high temperature are not those most readily seen at a low temperature."

These conclusions of Bunsen and Kirchhoff are now known to be true only within certain limits. The spectrum of a substance may be very considerably altered by change of temperature, and these changes in the spectrum do not consist merely in the alteration of the relative intensities of the lines, but are caused both by the addition of new lines and by the actual disappearance of lines present in the spectrum produced at the lower temperature. We have in the lithium and sodium spectra examples in which the change caused by increase of temperature consists simply in the addition of new lines, and the higher the temperature the greater becomes the complexity of the spectrum. A bead of lithium chloride in the Bunsen flame gives a spectrum consisting of only one red line, whose wave-length is about 6684 ten-millionths of a millimetre, which corresponds to 32 of the scale to which the spectra accompanying this article are drawn. If the temperature be slightly raised by employing the blowpipe, an orange line at 44 (wave length, 6107) makes its appearance. At the higher temperature of the oxyhydrogen jet a blue line at 105 (wave length, 4605) is added, while at the intense temperature

obtained by using the electric light the spectrum gives a fourth line at 86. The sodium spectrum at the temperature of the Bunsen flame consists only of the double yellow line of the same refrangibility as the solar line D, but if the sodium compound be ignited in the electric arc the spectrum contains four other lines, each also double.

The high temperature spectrum of sodium is represented in Fig. 1, on the plan proposed by Bunsen. The position of the bright bands on the illuminated millimetre scale of the spectroscope is shown by the position of the black lines, while the intensity is indicated by the relative height of the lines. In regard to the other conclusion of Bunsen and Kirchhoff, that all compounds of a metal give the same spectrum, we now know that this is true only if the metal is one whose compounds are decomposed even at the low temperature of the flame. It is well known that a sufficiently high temperature causes the decomposition of many chemical substances into their elements, and that these re-combine when the temperature is allowed to fall again. Thus, sodium carbonate and sodium nitrate give the same spectrum, because each is decomposed in the flame yielding metallic sodium by the incandescent vapour of which the yellow line is produced.

But we have compounds which do not split up into their elements in the flame, although they are decomposed in the intense heat of the electric spark. Such a substance is copper chloride, which volatilises in the flame without being decomposed, and gives a spectrum which is altogether different from the true spectrum of copper obtained when the electric spark is allowed to pass between copper poles. There is no doubt that the spectrum is that of the compound copper chloride. Fig. 2 shows this spectrum compared with that of copper.

The spectra of barium, strontium, and calcium show differences at different temperatures, which are probably to be explained in the same way. A reference to Fig. 3 will show the great difference observed when the calcium spectrum is produced by taking the electric spark in an atmosphere containing calcium, and when it is produced by bringing a bead of calcium chloride into the Bunsen flame. Figs. 4 and 5 show the spectra of strontium and barium respectively under the same circumstances. It is supposed that in the flame spectrum the incandescent substance to which the lines of the spectrum are due is calcium or strontium oxide, so that we have the spectrum of a compound, but when the electric spark is used, the

oxide is decomposed, and we have the true spectrum of the metal. It should be remarked here that the true metal spectrum consists invariably of sharply defined lines, while a compound gives a spectrum containing broad bands, which show a family resemblance amongst themselves and are often repetitions of each other. This is seen in Figs. 2, 3, 4, and 5. The spectra of those metals whose salts are easily decomposed in the flame-for example, sodium, lithium, and thallium-give spectra containing only lines, and the only change producible by increase of temperature is the addition of new lines, while, in the case of metals whose salts are not so easily decomposed, the increase of temperature not only adds new lines, but also splits up the bands into groups of fine lines.

We observe the same thing amongst the non-metallic elements in the case of cyanogen. If the flame of cyanogen burning in air be examined with the spectroscope, a magnificent spectrum is seen, which is obviously made up of two different spectra; one stretching from the light green into the blue exhibits a series of groups of lines, in each of which the brightest line is towards the red, and each group fades away on the side towards the blue; but the red end of the spectrum shows a series of groups of lines of exactly the opposite character-the brightest line of each group being on the side towards the blue, and each group fading away on the red side. We shall see afterwards that the blue portion of the spectrum is due to the element carbon, but the red end is produced by the compound cyanogen; and if the cyanogen be burnt in oxygen instead of air, the two of these groups most towards the blue become replaced by carbon lines, while if the gas be ignited by the electric spark, the whole of the cyanogen bands disappear, and the spectrum consists altogether of carbon lines.

It is a coincidence which doubtless has its own significance that, in all cases where by increase of temperature the bands of the compound are made to give way to the lines of the element, the change takes place earliest in the blue end of the spectrum, and proceeds gradually towards the red. The flame-spectrum of strontium contains one line Srd of the metal which is seen also in the spark-spectrum. The calcium flame-spectrum also contains one line (135) which remains unchanged on increasing the temperature to that of the spark, and in both cases these lines of the metal terminate the spectrum towards the blue end.* The flame-spectrum

* I owe to my friend Mr. Aldis the probable explanation of this peculiarity as also of the way in which the carbon bands shade off uniformly towards the blue.

of barium contains no line of the metal, since barium oxide is less easily decomposed than either calcium oxide or strontium oxide.

But amongst all the additions made to our knowledge of spectrum analysis within these ten years, none is so startling as the discovery, which we owe to Plücker, that a substance may give two totally different spectra which have no line or band in common. In a paper published in the "Philosophical Transactions" for 1865, Plücker and Hittorf describe double spectra of nitrogen, sulphur, selenium, hydrogen, and iodine. Nitrogen exhibits this peculiarity in a marked manner. In order to obtain its spectrum it is necessary to employ electricity, as no flame is hot enough. If an ordinary vacuum-tube containing nitrogen have the current from an induction coil sent through it, the narrow part of the tube gives out a purple light, which is resolved by the prism into the spectrum represented in the chromolithograph-a spectrum consisting of an immense number of shaded bands.

If, instead of using nitrogen at low pressures, we let the spark pass in the gas at the ordinary pressure, and intensify it by connecting the two wires with the outer and inner coatings of a moderate sized Leyden jar, we obtain an intensely bright light, which gives a spectrum also represented in the chromo-lithograph. This second spectrum is entirely different from the other, consisting only of sharply defined bright lines. Plücker terms these spectra, spectra of the first order, and of the second order, respectively. It will be observed that these spectra possess, respectively, the

The vibrations which produce light depend, so far as we can see, on the manner in which the atoms of a molecule are in equilibrium. We see from the occurrence of lines in all parts of the spectrum, that there are in the molecule several different vibrations executed simultaneously, and these correspond to the greater or less intensity of the force by which the atoms are maintained in their position of equilibrium. The more intense the force by which an atom is held in equilibrium, the faster it will vibrate when set in motion.

When cyanogen is moderately heated its molecules vibrate in the regular way indicated by the cyanogen spectrum; but when the temperature is raised the compound is dissociated, and the carbon atoms vibrate uninfluenced by those of nitrogen. Now, when this takes place, the vibrations of the cyanogen which first disappear must be those due to the closest intimacy-that is, the most rapid. Hence the carbon spectrum comes in at the blue end.

Further, the vibrations of an atom about its position of equilibrium will not all be of equal length, and so will produce light of varying intensity. If the vibrations are quite cycloidal, all will be executed in the same time, and we shall have a sharp bright line, but if the vibrations are like those of an ordinary pendulum, the smaller vibrations will be performed in less time than the larger, and the result will be a band fading off into the blue. If, on the other hand, the vibrations of small amplitude are executed more slowly than those of larger amplitude, the result will be a band fading off into the red.

characteristics of those produced by compounds and by elements. It remains for future experiments to confirm or modify the indication thus given of the compound nature of nitrogen. Similar results were obtained by Plücker with sulphur.

In order to experiment with sulphur, a tube of difficultly fusible glass was employed. Sulphur was introduced into the tube, which was then completely exhausted. When the narrow part of the tube is gently warmed by a spirit-lamp, and the platinum wires are connected with the induction coil, a spectrum of the first order is obtained. It consists of thirty-seven well-defined bands extending from the red into the extreme blue. Seven lie between the solar lines C and D, eighteen between D and F, and eleven between F and G. On heating the tube still more a quite different set of bright lines makes its appearance, and on introducing a Leyden jar into the circuit the second spectrum becomes fully established and no trace of the spectrum of bands remains. This second spectrum consists entirely of sharply defined bright linestwo red lines are especially noticeable, each of them triple.

It is thus established that certain gases may, under altered circumstances, vibrate in an entirely different manner, and Plücker believes that the necessary difference of circumstance is simply difference of temperature, the spectra of the first order belonging always to the lower temperature. The Leyden jar increases the temperature of the gas, for it necessitates the accumulation of a larger quantity of electricity preparatory to each discharge, so that the temperature of the spark with the Leyden jar is much higher than that of the simple discharge. Thus we see that, on heating the sulphur-tube and employing the Leyden jar, the low temperature spectrum gives way to the high temperature spectrum.

The spectrum obtained from an ordinary vacuum-tube. containing hydrogen (under a pressure of 5 to 10 millimetres), consists of three lines only-Ha, coincident with Fraunhofer's line c in the red; HB, coincident with F in the blue; and Hy, nearly coincident with G in the violet.

Plücker, who first observed this spectrum (Fig. 6), described, in the paper already referred to, a second spectrum of hydrogen, corresponding to a lower temperature. This observation has been abundantly confirmed by important experiments, made since the date of Plücker's paper, by Prof. Wüllner, of Bonn; and his results are so remarkable that it will be well to describe them somewhat at length. His apparatus consisted of a vacuum-tube of the ordinary