August 21, 1919] 



NATURE 



497 



The contours Essociated with the latter should be thin, 

 symmetrical parabolas. Those we found were wedge- 

 shaped, with definite kinks indicating the introduction 

 of new component lines when the condenser was put 

 into the circuit. The wedge shape indicates that the 

 law of decrease of intensity from the centre of a com- 

 ix)ncnt is exponential, and not the law of error as in 

 the Doppler effect. By measuring the distances be- 

 tween the kinks, and knowing the magnification and a 

 previous calibration of the wedge, all necessary quan- 

 titative data of the spectrum line can be calculated. 

 It was possible to show in particular that the separa- 

 tions in wave-length of the components of Ha wei-e 

 those found by Stark when new components were 

 called into being by the existence of an enormous ex- 

 ternal potential gradient. As we had suspected, the 

 origin of this exceptional broadening under the con- 

 densed discharge is the. "electric Zeeman effect," the 

 origin of the large electric field on any atom being 

 the close proximity of other charged atoms. We thus 

 have a new means of studying the electrical resolution 

 of spectral lines, more convenient in many ways than 

 the older methods, and capable of much greater gener- 

 ality and accuracy. A large number of observations 

 »of the same phenomenon were made also on the spec- 

 tral lines of helium and lithium, and the correspond- 

 ence with the Stark effect always held good. 



The examination of an individual line has also been 

 applied in the case of an "ordinary" discharge, and 

 has given the first direct proof of the probability dis- 

 tribution of velocities in the radiating atoms of a gas. 

 This distribution was taken as a basis bv Lord Ray- 

 leigh in the elaboration of a precise method of deter- 

 mining the mass of a radiating atom from the breadth 

 of the spectrum lines — a method applied by Buisson 

 and Fabry with great success, when the breadth is 

 measured by methods of interference of light. Our 

 experiments have defined very closely the circum- 

 stances in which this method is practicable, and shown 

 that it fails altogether if condensed discharges are em- 

 ployed. In the ordinary uncondensed discharge under 

 low pressure, however, our contours are very accurately 

 parabolic, which fact can be shown to imply a very 

 rigorous probabilitv law of velocities in the atoms, and 

 no other important source of broadening of lines in 

 these circumstances. 



The only other application to an individual line, 

 which I shall mention _ concerns the nature of the 

 Balmer series of hydrogen, long believed to be a diffuse 

 series, with each line consisting of two close com- 

 ponents, scarcely separable or not separable at all, with 

 rho same intei-val in frequency between them for every 

 line. We have shoAvn that it is in fact a principal 

 series, with ihe separations decreasing in a calculable 

 way required by theory, confirming also the value of 

 the separation in Ha given by Fabry and Buisson. The 

 method was to use the neutral wedge in combination 

 with another apparatus of extreme resolving power — 

 in this case a Lummer-Gehrcke plate. We can in this 

 way obtain contours for a pair of close components 

 which cannot be detected visually as a pair, and the 

 actual interval can be deduced by a series of measure- 

 ments of the joint contour, consisting of two over- 

 lapping parabolas. We calculate the position of the 

 vertex of the inner one, and thence the separation, 

 which can. in Hg. be determined within about oooi of 

 an Angstrom unit. The actual separation in this line is 

 as small as 0030 A., and the present method could* 

 measure separations accurately, even if they were much 

 smaller. 



We pass now from the phenomena of structure and 

 intensity of a single line to those involving a compari- 

 son of different lines. Here the behaviour of the plate 

 NO. 2599, VOL. 103] 



for different wave-lengths must be dealt with. But it 

 so happens that every plate can be calibrated by throw- 

 ing on tp it, not only the whole spectrum under exam- 

 ination, but also the radiation — a continuous spectrum 

 — from the positive crater of the carbon arc. The energy 

 distribution in this case is known from Wien's law 

 when the temperature of the arc is known, as it is very 

 closely. On the slide "you will notice the curious con- 

 tour bounding this spectrum, lar'gely due to vagaries 

 in the sensitivity of the plate. Above it is the helium 

 sjx^ctrum on the same plate. To obtain an absolute 

 scale of intensities down the helium spectrum, inde- 

 pendent of all sources of error due to apparatus, we 

 only need to compare the heights of the lines with the 

 corresponding heights directly below them in the 

 carbon-arc spectrum. It is, in fact, logarithms of 

 intensity which the heights represent, and differences 

 of height represent powers of a definite factor entenng 

 into the intensities, so that the photographs give no 

 visual impression of the enormous differences of in- 

 tensity which occur. For example, the line of wave- 

 length A 3888, a principal line in the helium spectrum, 

 appears quite short on the photographs, but is actually 

 the rtiost intense in the spectrum. It happens to be 

 in a region of wave-length where the plate is not sensi- 

 tive. One of our conclusions is, in fact, that principal 

 series deserve their name even in elements which 

 appeared hitherto to be exceptions, in that they do 

 contain, for the visible region, a preponderant part of 

 the energy radiated 



It is not necessary to use the carbon arc in every 

 subsequent experiment. We can, by its means, cali- 

 brate the helium spectrum under conditions easily 

 reproduced, and afterwards take this as our standard, 

 especially when the work projected is the variation of 

 the helium spectrum under changing conditions of 

 excitation. Some of the remaining slides indicate the 

 unexpected character of some of these variations. It 

 would not be possible, in this discourse, to give any- 

 thing like a complete account of the phenomena of 

 this class already investisfated, and I shall therefore 

 confine myself to some of those which are most strik- 

 ing. In the first place, we may notice the spectrum 

 of a mixture of hydrogen and helium or neon. The 

 fundamental phenomenon which this method has de- 

 tected is what we have called " transfer of energy 

 along a series." For instance, in the Balmer series of 

 hydrogen, produced from pure hydrogen under "ordin- 

 arv " conditions, there is a perfectly definite intensity 

 relation among the lines H,,, H^, and so on, but in 

 the presence of helium this is disturbed, and the ratios 

 H,j/H„, H,,/H,j, are notably increased. In other 

 words, more energy tends to be emitted in the form of 

 the more refrangible rays, at the expense of the less 

 refrangible. It is interesting to speculate as to how- 

 far this process can be carried, for its logical extreme 

 is a radiation from hydrogen concentrated at A 3646, 

 the limit of the Balmer series. We have not, in fact, 

 examined the matter from this point of view. 



Neon produces the same effect on the hydrogen spec- 

 trum, the first recorded evidence being an experiment 

 of Liveing and Dewar, who found in iqoo that it was 

 Dossible to observe more of the violet members of the 

 Balmer series when neon was present. We have made 

 quantitative measurements of the effect in various 

 cases, and in one experiment, for instance, the neon 

 was found to make Ha 6/5 as strong, H, 0/5 as strong", 

 and H^ 11/5 as strong in comparison with H^. But 1 

 shall not enter into further numerical detail, A parti- 

 cularly interesting fact is that the effect of a small 

 trace of an impurity is often diametrically opposite to 

 that of a larjre quantitv, and causes the transfer of 

 energy to take place in the opposite direction along the 



