Structure of the Polished Surface of Metals

A. V. Sushkin, Moscow

The initial ideas about the structure of the polished surface of metals are found already in the works of Gernetz¹, who noted that polishing is the result of the friction of microscopic particles against one another, and of Rayleigh², who pointed out the dependence of the quality of a polished surface on the kind of polishing materials used.

A broader assumption about the nature of a polished surface, made on the basis of investigation with an ordinary microscope, was put forward by Beilby³, according to whom the polished surface is amorphous and owes its origin to the flow of material from one part of the surface to another. Following the appearance of this hypothesis, which now bears its author’s name, many works appeared that examined this question from various points of view. However, all these works, devoted to an indirect consideration of the question, could neither confirm nor refute the original hypothesis. Only beginning with the work of G. Thomson⁴ was a new step taken in the study of the polished surface.

Thomson applied the now widely used method of electron diffraction, which, as is known, makes it possible to study very thin layers of matter. Investigating a polished surface by this method, Thomson, in agreement with Beilby’s hypothesis, did not obtain in the electron diffraction patterns the rings characteristic of polycrystalline structures.

Continuing Thomson’s work, French⁵ found that the diffraction pattern changes with the degree of polishing: the sharp diffraction rings from a ground surface, with increasing degree of polishing, gradually become blurred until, in the final result, two broad rings are obtained, only weakly visible against a comparatively bright background. This phenomenon could have been caused either by an extremely great diminution of the crystals, or by the complete destruction of the crystalline structure with the formation of an amorphous surface. By measuring the radii and relative intensities of these rings and comparing with them the corresponding data to which Wierl’s formulas⁶ lead for diffraction by molecules, French found satisfactory agreement between these quantities, from which he had to conclude that the assumption of an amalgam consisting of randomly arranged atoms is correct.

In opposition to the views of Thomson and French there appeared works in which their assertions were disputed. Randall and Rooksby⁷, on the basis of investigations of graphite and coal, asserted that the rings obtained by French corresponded to the strongest characteristic rings, blurred as a result of a decrease in crystal size; the less intense rings were thereby obscured by the background. The difference between the radii of the normal and the blurred rings, in the opinion of Randall and Rooksby, was due to a change in the lattice parameters as the crystals diminished in size. Germer⁸, considering the polished surface to be crystalline, asserted, contrary to the results of French’s experiments, that the inner potential of metals, reaching 15 V for nickel at a small angle of incidence—

...of the electron beam may cause, as a result of refraction, complete extinction of all the rings. In agreement with Germer’s views on the role of refraction, Kirchner^9 assumed that polishing is a chipping-off of protruding parts of crystals, whereby the polished surface consists of crystalline planes parallel to it, and the beam, instead of passing through protruding crystalline “spikes,” is forced to pass along a longer path in the body of the metal. Therefore, diffuse and weakly noticeable rings should be observed on electron diffraction patterns, since in this case multiple scattering and refraction must have a strong effect.

Retter,^10 who had previously been a supporter of the views of Kirchner set forth above, on the basis of his latest experiments came to the conviction of the correctness of Beilby’s hypothesis. He found that iron, copper, nickel, potassium, silver, and gold, when subjected to drawing, polishing, forging, or pressing with a polished matrix, gave identical diffuse rings on electron diffraction patterns. The position of these rings depended neither on the kind of treatment applied nor on the position of the corresponding characteristic rings, as would be the case if the diffuse rings were the result of coalescence of characteristic rings broadened, for example, as a result of pulverization of the crystals. As in French’s work, the radii of the rings obtained by Retter coincided with the values calculated on the assumption of the existence of an amorphous layer (it is interesting to note in passing that the polished surfaces of single crystals of rock salt, potassium sulfate, fluorite, and pyrite studied by Retter did not give diffraction patterns at all that differed in any way from the usual patterns typical of single crystals).

Later Derbyshire and Dixit^11 continued French’s investigations, extending them to metals of non-cubic structure, with particular attention to selecting the metals in such a way as to exclude the possibility of assuming a coalescence of the characteristic rings due to refraction. Having obtained specific electron diffraction patterns with two faintly visible rings and, using Wierl’s formula, calculating from their radii the interatomic distances, Derbyshire and Dixit compared the values they computed with the values of the interatomic distances \(b\) of metals in the molten state, calculated by Keesom’s formula

\[ b = 1.33 \sqrt[3]{\frac{A}{D}}, \]

where \(A\) is the atomic weight and \(D\) the density.

Good agreement of these quantities was found for zinc, gold, silver, molybdenum, and some other metals; for bismuth, antimony, tellurium, cadmium, and lead the values of the interatomic distances in the polished layer deviated from the values of the interatomic distances in the liquid metal by approximately 20% in the direction of decrease. Later Miwa^12 found that exactly the same is obtained for tin, whereas copper, iron, chromium, cobalt, and nickel likewise give good agreement of the indicated values. Having drawn attention, on the basis of Jom-Rosery,^13 to a certain anomalous character of the metals giving noncoincident values of the interatomic distances for the polished layer and the liquid state, Derbyshire and Dixit came to the conclusion that the electron diffraction patterns they obtained are due precisely to an amorphous layer.

It is interesting in this connection that in a number of metals whose atomic volume lies at the maximum of the curve expressing Mendeleev’s periodic law, their volume and, consequently, density change considerably. The magnitude of the atomic volume of such a metal approaches the nearest minimum on the periodic curve. The atomic volume of the elements lying close to the minimum on the aforementioned curve does not change essentially. The significance of these conclusions, in view of their agreement with Bridgman’s observations on the densification of metals under pressure, on the one hand, and observations on the formation of an amorphous layer by pressure, on the other, is difficult to overestimate.

In the further development of views on the structure of a polished surface, the investigations of Finch, Quarrell, and Roebuck played a particularly important role.1 These investigations produced the following experiments: by evaporation in a high vacuum, metallic films were deposited in an electronograph onto well-polished or etched metal plates. Electronograms first from the substrate itself, and then from the films, were observed on the fluorescent screen of the instrument. At the same time it was noticed that the diffraction pattern obtained immediately after the deposition of zinc onto a polished copper surface and corresponding to crystalline zinc began gradually to disappear and, after several seconds, vanished without a trace, which testified to the disappearance of the crystalline structure of the zinc. Electronograms from subsequent deposits went one after another (sometimes up to 12), disappearing in the same way as the first, and the rate of disappearance decreased with each time. Thus, if the pattern from the first layer of zinc disappeared in a few seconds, the patterns from the last layers were faintly visible after 4 min, disappearing completely after 5 min, and only the last deposit gave a pattern still perceptible after four hours. The same zinc films deposited on crystalline (etched) copper surfaces gave corresponding diffraction patterns that did not change over the course of one and a half hours either in brightness or in configuration. Similar results were also obtained with layers of zinc, lead, silver, tin, cadmium, and lithium on copper, iron, zinc, lead, and gold. In addition to this, a somewhat later work by Finch, Quarrell, and Wilman[^15] showed, in confirmation of the supposition regarding the difference between a polished surface and a polycrystalline one, that in electrolysis with a current of 0.1 A/dm² on electrodes of crystalline copper, 30 sec were sufficient for the formation of a definite zinc layer, whereas for the formation of the same layer on polished copper electrodes 3 min were required. On the basis of the stated facts, the authors of the investigations concluded that the disappearance of the diffraction pattern of the deposited films is evidence of their dissolution by the polished layer, with the formation of a solid solution, which, as the number of deposited layers increases, gradually reaches saturation. This phenomenon shows that the polished surface has the property of dissolving, considerably more rapidly, the layers of metal deposited on it than does a polycrystalline surface, and this is strong evidence in favor of Beilby’s hypothesis (it should be noted that the described effect is observed only on sufficiently thick amorphous layers).

However, the question of the structure of the polished surface remained not finally decided even after the appearance of the investigations just described, and its significance was so great that at the discussion devoted to the structure of metallic coatings and films, organized in March 1935 by the Faraday Society in England, it was subjected to serious discussion.[^15] [^16] [^17] [^18] [^19] In a report devoted to the state of electronography, Finch, Quarrell, and Wilman[^15] pointed to the appearance of an amorphous layer on the inner surfaces of the cylinders of aircraft engines from the friction of their piston rings during prolonged operation. Electronograms were taken from the working surfaces; in motors that had been run in, these gave blurred rings characteristic of an amorphous surface. After the tested specimens were rubbed with No. 0000 emery paper, ordinary rings appeared on the electronograms (as with the surfaces of engines that had not been run in), characteristic of cast $\alpha$-Fe.

Speaking at this discussion, Hopkins[^16] reported in his paper the results of measurements of the thickness of the amorphous layer. His work, giving additional evidence for Beilby’s hypothesis, showed that when the upper layers were removed from polished gold by cathodic sputtering, alongside the pattern from the amorphous state, fine lines of crystalline gold were also observed. These lines, barely noticeable when a layer of 10 Å was removed, became brighter and brighter as further layers of gold were removed from the surface, until, finally,

after removal of a layer of \(400\text{--}500\) Å, a distinct normal electron diffraction pattern of gold was not obtained.

Analogous results were obtained and reported in a paper at the same discussion by Liss\({}^{17}\). Working with gold and copper, in addition to a similar change in the electron diffraction patterns, he discovered the presence beneath the amorphous layer of crystals oriented by the (110) plane parallel to the surface. On the basis of these observations, Liss raised an objection to Kirchner’s views. If one assumes that crystalline faces are present on the surface, then, according to his observations, one must conclude that it will consist of (110) planes, for which the corresponding maxima in an electron diffraction pattern at an internal potential of about \(15\ V\) and an electron velocity of \(40\ kV\), as calculation shows, should, owing to refraction, be noticeably elongated in the direction normal to the surface of the specimen. In reality, however, Liss did not observe these elongated spots. Both of these works on measuring the thickness of the amorphous layer, while mutually confirming one another, also confirm Beilby’s hypothesis in that, with successive removal of layers, i.e. in the transition from an amorphous structure to a crystalline one, the electron diffraction patterns change, passing in reverse order through all the phases previously observed by French.

Despite the success of all the evidence considered above, Kirchner\({}^{18}\), continuing at the discussion to defend a standpoint opposite to Beilby’s hypothesis, presented a number of objections deserving detailed analysis. In his opinion, his own observations—by which he found a difference between electron diffraction patterns obtained by the transmission method of an electron beam through thin metallic films and electron diffraction patterns obtained by “reflection” from a surface—do not agree with the hypothesis of the amorphous nature of the polished surface.

The experiments consisted in the following: from a gold film about \(\sim 10^{-6}\ \text{cm}\) thick, obtained by condensation in vacuum, electron diffraction patterns were taken by the transmission method and by the “reflection” method. In the first case sharp rings were obtained on the electron diffraction pattern, and the rings (111) and (200) were clearly separated. In the second case two indistinct rings of the same appearance as those usually obtained in electron diffraction patterns from a polished surface were obtained. More extensive experiments by Pansdorf\({}^{19}\) confirm this phenomenon. It was further found that, as the layer thickness increases, the difference in the sharpness of the interference rings is smoothed out. However, the normal sharpness of the rings in “reflection” is obtained at a greater layer thickness than in the case of transmission of an electron beam. This phenomenon is explained as follows: assuming an ideally smooth substrate (for example, mica) with irregularities only of atomic dimensions, at small values of the thickness of the deposited metal one can obtain a layer with irregularities on the order of the dimensions of the crystals. Consequently, in taking electron diffraction patterns by the reflection method, i.e. with glancing passage of the electron beam over the surface under investigation, the crystals must almost cover one another, so that only the protruding parts (“peaks”) should take part in forming the pattern. Taking the effective size of a crystal to be \(1/4\) of its actual size, one should expect that, in reflection, some decrease in the sharpness of the rings will be observed as compared with the case of transmission. The smaller the crystals, the more strongly this effect is manifested, and if in transmission the true size of the crystal can be better judged from the width of the rings, in the case of reflection, even with natural growth of the crystals, it is difficult to judge their true size. Therefore the broadening of the rings in electron diffraction patterns taken from a polished surface cannot be sufficient proof of a strong reduction in crystal size or of the formation of an amorphous layer. Pansdorf’s investigations showed, further, that an amorphous surface in fact does not exist here, since the deliberately amorphous film of cesium studied on electron diffraction patterns taken by the transmission method—

... gave strongly broadened interference rings, while on electronograms taken by “reflection” it gave only a background uniformly decreasing from the center to the edges; the investigated surface of liquid mercury likewise did not give interference rings. Hence it follows that, from the point of view of amorphous surface scattering in reflection, one cannot in general expect any maxima on an electronogram, and their appearance in such cases should be taken as an argument for the presence of comparatively large irregularities, larger than the irregularities of an amorphous surface or the surface of a monatomic liquid.

Raising objections to the above conceptions of Knipping, Thomson^20 stated that, when electronograms are taken by the reflection method, there is no true reflection, since at ordinary angles of incidence of the electron beam, not exceeding \(5^\circ\), and with a maximum path of the electrons in the metal of about \(4\cdot 10^{-6}\) cm, the depth of penetration of the electrons from the surface must be not less than \(2\cdot 10^{-7}\) cm, and this should be sufficient, in the presence of a truly crystalline material, for the appearance of sharper rings than those usually obtained on electronograms from a polished surface. Summing up the discussion, Thomson, as well as Finch and others, stated that he considered the existence of an amorphous layer on polished metal to be an established fact.

Despite this outcome of the discussion, Dobinsky^21, in a later work, asserted that the ordinary electronogram from a polished surface should be regarded as due to oxide films that readily arise on metals exposed to air. In support of this assumption Dobinsky, referring to the absence in all works on this question of any special precautionary measures, cited his own observations. Using benzine as a lubricant during polishing, immediately after polishing he obtained (from copper specimens) electronograms with diffuse half-rings corresponding to interatomic distances smaller than those usually obtained for polished copper. The same specimen, left in air, gradually oxidizing, after some time began to give electronograms that did not differ from the ordinary electronograms of polished copper. Interpreting the obtained electronograms as the result of a mixture of characteristic copper rings \((111)\) and \((200)\) for the inner diffuse ring and \((220)\), \((331)\), and \((222)\) for the outer ring, Dobinsky concluded that the ordinary electronograms of polished copper may be assigned to cuprous oxide \(Cu_2O\). Since many metals (for example nickel and iron), while having different crystalline structures, nevertheless almost all have the same structure for their oxides as \(Cu_2O\), the above considerations on the interpretation of electronograms may be extended to most metals. Callender^22 also agreed with these conclusions of Dobinsky, basing himself on his own experiments on dissolution carried out on polished aluminium. Callender maintained that all previous investigations had in fact been carried out on an oxide film^1).

The last of the large series of works devoted to this question is Finch’s article^23, published in 1937. In it, summing up, Finch writes that Beilby’s views are confirmed by analysis and comparison of all information available up to the present time. Only in the part of the hypothesis that concerns the polished surface of nonmetals was Beilby wrong: the polished surface of calcite, for example, is (as is confirmed especially by the latest exper—

^1) How serious the conclusions of Dobinsky and Callender are can also be seen from a whole series of works connected with the discovery of an oxide film arising at room temperature on other metals as well. Such a film was observed by [[unclear: Einetingeil]]^24 on bismuth^25, by [[unclear: Nemoli]] and Riedmüller^26 on nickel, by Finch and Quarrell^27 on zinc, and, finally, by Nelson^28 and many other authors on iron. Ed.

STRUCTURE OF THE POLISHED SURFACE OF METALS

experiments of Finch himself) a monocrystalline structure appearing during recrystallization of the amorphous layer.

One may fully agree with the authoritative statement of the well-known investigator, since consideration of all the viewpoints set forth above shows that the assertions of the opponents of Beilby’s hypothesis are refuted by a number of extremely weighty facts discovered by its numerous supporters. Even the objections of Dobinsky and Kalender, which are essentially modifications of Kisler’s objections, find their refutation in the works analyzed above. Indeed, quite apart from the impossibility of coincidence of the oxide structures of all metals and the obligatory removal of them from the surfaces under study during cathodic and electrolytic removal of layers, which makes it possible to purify the true structure of the polished layer, even Kisler’s proposed replacement of the entire amorphous oxide layer can hardly explain the broadening of the rings in the electron diffraction pattern otherwise than by the comminution of crystals.

Thus it may be considered that Beilby’s hypothesis has at present apparently won a firm position among the majority of investigators working on the question of the polished surface. The work of numerous investigators has not only supported Beilby’s proposal, but has also provided material for the formation of a modern view of the structure of the polished surface and the polishing process. This view may be formulated as follows.

The polished surface of metals consists either of strongly comminuted crystals (down to the size of an elementary cell), or, more probably, of separate atoms brought as close as possible together, forming a certain analogue of a liquid—an amorphous layer.

In addition to the judgments expressed about the structure of the polished surface, it is also possible to establish its thickness and to characterize the structure of the metal layers nearest to it, which is directly connected with the structure of the polished surface.

The initial attempts to determine the thickness of the amorphous layer, made by Beilby with the aid of a microscope, consisted in the successive etching of the polished surface, but they did not make it possible to establish even the order of magnitude of the quantities that might be of interest. Retter, on the basis of a generalization of his experiments, gave for the amorphous layer of metals a thickness of the order of \(10^{-6}\)—\(10^{-7}\) cm.

To establish precisely the thickness of the amorphous layer and the deep-lying structures of metals for different polishing conditions is still not possible, owing to the small number of experimental data. However, the available data make it possible in individual cases to determine the limits of variation in the depth of penetration of polishing, the dependence of which on many factors is now beyond doubt. Observations were made on the lower layers of surfaces of copper and gold polished by various methods. Removal of the outer layers of metal was carried out successively in small portions, in one case by electrolysis (Lees), in another by cathodic sputtering (Tolking). Irrespective of the method by which this was done, different results of observation were not obtained. It was found that the upper layer, about 20–40 Å thick, is an amorphous layer (or a layer of disoriented crystals smaller than 15 Å). Immediately behind this layer there was found a layer, with a thickness from 200 to 800 Å for copper and up to 10,000 Å (1 μ) for gold, in which an orientation of the crystals was observed with the \((110)\) plane almost parallel to the surface (the angle between the normals to the 110 plane and to the surface under investigation was equal to 20°), while the size of the crystals in this layer was smaller than in the deeper layers. A very remarkable property of this layer is the independence of the orientation of the crystals in it from the direction of motion of the metal specimen being polished during polishing. Behind the layer of oriented crystals there was observed a layer of somewhat comminuted crystals,

directly passing into the ordinary structure of the metal. Thus it is evident that, with increasing distance from the surface, the magnitude of the deformation decreases and, naturally, for soft metals the depth of penetration of the deformation must be greater, despite the fact that the overall picture of deformation for metals of the same group should be the same.

The structural features described above undoubtedly determine—though not yet in an explicit form—the properties of the polished surface, which, in addition to the universally known (but not specific) property of “specularity,” possesses a number of remarkable properties. It has long been known that a polished surface is comparatively weakly oxidized and that its hardness and electrical resistance are greater than those of a crystalline surface. Some of these properties can be explained more or less reliably in the following way. Specularity is a consequence of the smoothing of irregularities larger than the wavelength of light rays; the chemical passivity of the polished surface is not an inherent property of it, but is the result of the passivation of the polished surface—the formation of oxide films, which readily arise on a polished surface in air; the polished surface itself, however, is extremely chemically active; the increased hardness is explained by the disordered and dense packing of atoms in the amorphous layer, which has, from the thermodynamic point of view, a high potential and is in a very stressed state. In a special place stands the property described above of the dissolution, by a polished surface, of thin layers of metal deposited upon it, for which no complete explanation can be given; the form of existence of a solid solution in a polished surface is likewise unknown.

The connection of the properties of a polished surface with the conditions of polishing is beyond doubt, and therefore a clear conception of the mechanism of this process is extremely important. Beginning with Beilby, the question of the mechanism of polishing has received attention in almost every investigation, but, despite this, much remains unclear in this question. Beilby believed that the process of polishing proceeds as follows: under the pressure and friction of the microscopic protrusions of the particles of the polishing material, the upper layer of the metal being polished experiences large deforming forces, from which the projecting angles of the crystals break off, the crystal lattice is destroyed, the metal is, as it were, smeared out, flows, filling depressions on the surface. Lis, in the work mentioned above, in confirmation of what has been set forth, expressed the supposition that pressure plays the principal role in the formation of the amorphous layer and of the subsequent layer of oriented crystals. This conviction is based on the fact that, no matter how the direction of motion of the specimen over the plate changes during polishing, the direction of orientation remains constant; moreover, the orientation proves to be the same as that observed by Taylor^23 and others in strongly compressed crystals. True, in order to draw this parallel and accept this interpretation, it is necessary to assume that the area of the metal in contact with the total area of the protrusions of the microscopic particles of the polishing material is equal to \(1/5000\) of the total area of the specimen, since the pressure applied in polishing is, by hand, \(\sim 0.12\ \mathrm{kg/cm^2}\), while the pressure used by Taylor is of the order of \(\sim 620\ \mathrm{kg/cm^2}\). It is also necessary to assume that friction at the places of contact of the polishing particles and the metal is insignificant, since otherwise, in identifying the experimental conditions, the pressure force normal to the surface cannot be taken into account. The important role of the pressure that produces plastic deformation is indicated by the experiments of Rettger, who obtained an amorphous surface by drawing, polishing, and burnishing, the structures of the surfaces obtained not differing from one another.^1)

It should be noted that softer and more viscous metals are more readily polished. It is also curious that in the case of work with liquid—

^1) In addition to the data presented here, one should also refer to the work of Bowden and Ridler^29, who showed that the mean tempera-

with lubricating substances the specular quality appears more often when the polishing material becomes somewhat dull and the motion of the specimen requires considerable effort.

The work carried out up to the present time does not provide material for a final judgment about the mechanism of polishing and for establishing practical dependences of the quality of the surface obtained on pressure, polishing material, and duration of polishing. It is therefore necessary to confine oneself to a few particular cases. Thus, for example, it is known that, in order to obtain an effect of dissolution in the polished surface of a layer thickness sufficient for observation, a chemically active polishing wheel with rosin is necessary; satisfactory results could not be obtained with hand polishing. According to Lyss’s data, it is known that when copper is polished by hand with crocus and water, the depth of the polished layer (amorphous and oriented together) was approximately \(200\ \text{Å}\), the size of the crystals in the oriented layer was \(\sim 160\ \text{Å}\); when polished with crocus on a flannel wheel, the total depth of the layer was \(\sim 500\ \text{Å}\), the crystal size \(\sim 100\ \text{Å}\); when working with a lubricating liquid polishing substance, the amorphous layer was not entirely perfect, and orientation of the crystals was observed on the surface. The role of the lubricant in polishing is not limited to its influence on the quality of the amorphous layer, but, depending on the kind of substance used for lubrication, after polishing one can obtain either an oxidized or an unoxidized surface. Thus, for example, when using gasoline (or a similar substance), oxidation during polishing could be avoided.

It will be appropriate to note here that different polishing conditions, which determine one or another surface quality, may lead to doubtful conclusions about the nature of the observed effect. It must be expected that the widespread opinion at present that a surface polished by the method of acad. Grebenshchikov is a surface that has preserved its crystalline character arose precisely because treatment by this method continues for a very short time, and therefore the thickness of the amorphous layer necessary for observing many characteristic phenomena is not obtained; instead a very thin layer is obtained, sufficient only to give the object a specular gloss.

LITERATURE

1. Herschel, Proc. Roy. Inst., 16, 563, 1901.
2. Rayleigh, Proc. Roy. Inst., 16, 563, 1901.
3. Beilby, Aggregation and Flow of Solids, Macmillan, London, 1921.
4. Thomson, Proc. Roy. Soc., A 128, 649, 1930.
5. French, Proc. Roy. Soc., A 140, 637, 1933.
6. R. Wierl, Ann. d. Phys., 8, 521, 1931.
7. Randall a. Rooksby, Nature, 129, 280, 1932.
8. Germer, Phys. Rev., 43, 724, 1938.
9. Kirchner, Nature, 129, 545, 1932; Trans. Farad. Soc., 31, 1114, 1935.
10. Raether, Z. Physik, 86, 82, 1933.

the polished layer, during its formation, rapidly increases to the yield point of the metal and then remains independent of the force of the polishing action.

According to Finch, this fact, if considered in connection with the flow of material during polishing discovered by Beilby (in particular by revealing old scratches from grinding after etching the polished layer), once again indicates that flow occurs not from the liquid phase, but from fragments of crystals. —Ed.

1. Darbyshire and Dixit, Phil. Mag., 108, 961, 1933.
2. Miwa, Sci. Pap. Tohoku Imp. Univ., 24, 222, 1935.
3. Hume-Rothery, The Metallic State, p. 307, Clarendon Press, 1931.
4. Finch, Quarell and Roebuck, Proc. Roy. Soc., A 145, 676, 1934.
5. Finch, Quarell and Willman, Trans. Farad. Soc., 31, 1935.
6. Hopkins, Trans. Farad. Soc., 31, 1935.
7. Lees, Trans. Farad. Soc., 31, 1935.
8. Kirchner, Trans. Farad. Soc., 31, 1114, 1935; Ann. d. Phys., 28, 21, 1937.
9. Papsdorf, Ann. d. Phys., 28, 555, 1937.
10. Thomson, Trans. Farad. Soc., 31, 1935.
11. Dobinsky, Nature, 138, 31, 1936.
12. Kalendar, Nature, 138, 291, 1936.
13. Finch, Trans. Farad. Soc., 33, 425, 1937.
    23a. Taylor and Farren, Proc. Roy. Soc., 111, 529, 1926.
14. Steinheil, Ann. d. Phys., 19, 465, 1934.
15. Beeching, Phil. Mag., 22, 933, 1936.
16. R. Riedmiller, Z. Physik, 102, 408, 1936.
17. Finch and Quarrell, Proc. Phys. Soc., 46, 148, 1934.
18. Nelson, J. Chem. Phys., 5, 252, 1937.
19. Bowden and Ridler, Proc. Roy. Soc., A 154, 640, 1936.

1. Finch, Quarrell, and Roebuck. ↩