evident by the critical potential methods, since the number of electrons losing this energy increases gradually as the velocity is increased, without there being any well-defined threshold potential. It was at first suggested that this effect might correspond to the excitation of H3, as relatively large quantities of H3+ are known to exist under the discharge conditions employed. The rare appearance of the loss below about 15 volts suggested that He was in some way connected with its occurrence, and this it is known is necessary to the formation of H3. It seems more likely, however, that this line appearing at low velocities is intimately connected with the dissociation of the diatomic hydrogen molecule. The fact that the most. probable loss, i. e., the loss corresponding to the point of maximum intensity, is not a definite amount, but increases as the velocity of the primary electrons increases from 16 to 20 volts, makes it unlikely that it corresponds to any particular excitation state of any kind of molecule. When the initial velocity of the electron, suffering collision with the molecule, corresponds very nearly to the energy separation of the two states between which the transition occurs, it might be expected that the electron and molecule. will remain in interaction for a time sufficiently long to allow the nuclear motion to affect the amount of energy finally lost by the electron. The energy lost under consideration may be due, therefore, to an excitation process, which finally results in the dissociation of the molecule and the emission of radiation which forms part of the continuous spectrum. It is interesting to notice in this connexion that Oldenberg † has found that the continuous spectrum, but not the Lyman bands, is strongly excited by electron impacts in pure hydrogen, but when argon is present Lyman bands appear strongly and the continuous spectrum very weak. The two thus appear to be closely related and probably due to the same initial electronic excitation, i. e., to the B state. The essential difference is, according to Oldenberg, that with argon present (the excitation occurring then by collisions of the second kind with the excited argon atoms) the vibrational energy is limited to that of the third state, whilst in the other case high vibrational states may occur, and therefore dissociation in the resulting transition which produces the radiation of the continuous spectrum. The work of * Jones and Whiddington, Proc. Leeds Phil. & Lit. Soc., July 1928. + Oldenberg, Zs. f. Phys. xli. p. 1 (1927). Hughes and Skellett shows that dissociation occurs almost certainly as a result of a simple process, i. e., not a secondary effect, and, moreover, that the dissociation does not set in until the electrons have a velocity of 11 or 12 volts, and increases for higher potentials. These results fit with the observation first made by Horton and Davies † of the excitation of the continuous spectrum at low velocities. They gave tentatively as the lower limit for excitation 12.6 volts. Finally, then, it appears that (1) the loss not being a definite one, independent of the excitation conditions, suggests that it is connected with dissociation; (2) results like those of Hughes and Skellett indicate that dissociation is connected with the emission of the continuous spectrum, since there appears no other way of accounting for the balance of energy, and does not occur until the electrons have a velocity of about 12 volts; (3) Oldenberg's results may be taken to indicate that the continuous spectrum is intimately connected with the excitation of the B electronic state, and this, as has been previously mentioned, would apparently require the impacting electron to have an initial velocity somewhat greater than the 111 volts necessary to stimulate the B, state. This fits, then, with the non-occurrence of the loss at very low potentials, and may be taken to support the attribution of this loss to a dissociation process. Summary. Experiments are described in which the energy losses of electrons in hydrogen of controllable velocity are measured, and also the relative number in the different energy groups. The results of the measurement of the photographs and their photometric curves may be summarized as follows:-(1) By far the most probable effect of a collision between an electron of 50 volts or more velocity and a hydrogen molecule is the excitation of the C state with a certain amount of vibrational energy. (2) The probability of effective collision at 150 volts. is in the neighbourhood of 1 or 2 per cent. (3) At low velocities, smaller losses, the upper limit of which is 8 or 9 volts, varying slightly with the primary electron velocity, are observed, and the suggestion is made that these are connected with the dissociation of the molecule and the excitation of the continuous spectrum. (4) There is no loss of about 111 volts to be found, indicating that the B, and other low vibrational B states cannot be stimulated directly * Hughes and Skellett, Phys. Rev. xxx. p. 11 (1927). by electron impact. (5) There is no indication, except at very low velocities (8 to 10 volts) when it is very faint, of a loss occurring which might be connected with direct dissociation. The cost of part of the apparatus used in this investigation has been defrayed by a grant from the Royal Society. The University, Leeds. LXXXIX. The K X-Ray Absorption Edge of Iron. RE [Plate XVI.] Introduction. ECENT work in X-ray absorption spectra has shown that on the short wave-length side of the edge there often exists a complicated appearance which has been called the fine structure, or multiple structure of the edge. Early investigators were Stenström †, Fricke ‡, and Hertz §. Lindh, in an extended study of the K absorption edges of sulphur, chlorine, and phosphorus, found one or two secondary edges. Coster and Van der Tuuk¶ reported a secondary edge very close to the K edge of argon. Dyke and one of the writers found a more complicated structure for the K edge of calcium than had previously been observed for any element. Nuttall ††, working in this laboratory, has recently shown a structure for the potassium and chlorine K edges which is also very complicated. Van The valence of the element affects the position of the absorption edge. In the work of Lindh there appeared a progressive shifting of the edge towards shorter wave-lengths as the valence of the element became higher. The present work was undertaken for the purpose of extending the knowledge of the structure of the K absorp Communicated by J. M. Nuttall, M.Sc. Stenström, Diss. Lund, 1919. Fricke, Phys. Rev. xvi. p. 202 (1920). § Hertz, Zeits. für Phys. iii. p. 19 (1920). Lindh, Diss. Lund, 1923. ¶ Coster and Van der Tuuk, Zeits. für Phys. xxxvii. p. 367 (1926). **Lindsay and Van Dyke, Phys. Rev. xxviii. p. 613 (1926). Nuttall, Phys. Rev. xxxi. p. 742 (1928). tion region and also with the hope of showing the effect of valence on this structure. With this in view, iron was chosen because of its convenient wave-length, because of the fact that crystals containing iron are plentiful, and also because iron has two valencies. Experimental Procedure. Two methods were used to obtain the absorption: first, crystals containing iron were used both as diffracting crystal and as absorbing medium; second, screens were made and placed in the path of the beam in the usual manner. The first method has the advantage of forming a very uniform absorber, and of permitting more radiation to reach the plate. It is a very desirable method for long wave-lengths and when good crystals are obtainable. On the other hand, if the crystals are poor and if suitable absorbing screens can be made, the second method gives better results. While crystals containing iron are plentiful, they have not especially good surfaces, and hence in this work the better results were obtained by the second method. Many of the crystals used were not common in X-ray work, and it was therefore necessary to obtain the crystalgrating constants. This was done by precision measurements on the CuKa line, as outlined by Siegbahn *. The natural faces of the crystals were used, but in some cases these were polished in an effort to eliminate some of the surface irregularities. This polishing no doubt sacrifices to some extent the sharpness of the fine structure. The absorbing screens were made by mixing the finely-pulverized compound with collodion and spreading the mixture out in thin films to dry. In this manner screens of suitable thickness could readily be obtained. One of the crystals used in the first method was lepidomelane, a mica which contains iron, and which can be split into sheets as thin as 001 mm. These sheets served as excellent absorption screens for the second method. The screens for metallic iron were made by rolling a piece of pure electrolytic iron which was obtained from the General Electric Co., through Professor W. P. Davey. The rolling was done between pieces of copper, and fairly uniform screens as thin as 0.01 mm. were made. A sylvite (KC) crystal was used in the absorbing screen method. This crystal gave very good reflexion. Its dispersion is somewhat greater than that of calcite. Siegbahn, 'Spectroscopy of X-Rays,' pp. 62-64. A Siegbahn vacuum spectrograph was used for photographing the edges. The reference line used on all the plates was the FeK8 line. The WL2 line was also near, and served as a check. The distance from absorption edge to reference line was measured on a comparator having a least count of 0.005 mm. In measuring the edge the cross hair of the comparator microscope was brought up just to the point where the blackening begins to decrease. The practice has commonly been to add to this reading one-half the width of the slit. It was found, however, that this correction gave abnormally low wave-lengths in cases where the dispersion of the crystal was small. If no corrections were made, the values for the edge by the two methods were found to agree much better. It was also found that the edge was much less sharp for metallic iron than for the compounds, even though other conditions were the same. This shows that factors other than slit-width may determine the sharpness of the edge. From these considerations it was thought best to omit the correction for the width of the slit. The width of the slit in In the screen method, the crystal method was 0.126 mm. where stronger reflexion could be obtained, it was narrowed to 0.043 mm. This aided in the resolution of some of the details of the structure. The curves shown in this work are microphotograms taken on a Moll microphotometer. The lens system was modified so as to utilize more of the height of the plate, thus eliminating most of the fluctuations due to the coarse grain of the X-ray plates. Experimental Details. The predominant features of the multiple structure are as follows: first, a very sharp principal absorption edge; second, an intense white-line absorption immediately to shorter wavelengths; third, several secondary regions of absorption, which also sometimes appear as white lines of less intensity. The more striking appearance of the first white line is, however, partly due to the fact that there is less contrast at the secondary edges. The microphotograms show that in many cases the secondary absorptions are really as intense as the principal white line. The blackening between the secondary absorptions by no means approaches in intensity that on the long wave-length side of the principal edge. In the tables this principal edge, which is the one always measured, is designated simply as the K edge. It forms the long wavelength boundary of the principal white line. The short |