On the rotating stage under the microscope, the edges of the lamina give a singularly pleasing effect. They prove them to be more of the nature of steps down facing away from the shorter diagonals of the rhombic faces, than of definite protrusions; and their undeviating straightness, so different from the vacillating composition plane of the macle, contrasts sharply with the rippled surfaces they cut.

FIG. II.-Laminated Wesselton Diamond,

enlarged.

Geometrically, the shorter diagonal of a rhomb of a dodecahedron coincides with an edge of the derived tetrahexahedron. The same rule is approximately true for such diamonds as are of good symmetry, and are not laminated; but when diamonds are laminated the rule is varied, in that the tetrahexahedron edge tends to slant away from its geometrical position into parallelism with the edges of the laminæ. This is, perhaps, the most important change made by the lamination on the aspect of a stone.

Fersmann and Goldschmidt have described ("Der Diamant," 1911) three laminated diamonds, all Brazilian, namely: Crystal No. 26, an irregular lump of a weak violet colour, showing lamination in three directions; Crystal 32, a brownish regular dodecahedron, whose surface is covered by a delicate network formed of innumerable twin lamellæ in four sets; and Crystal 33, a brown dodecahedron with coarse lamination. They argue that most diamonds may be presumed to be composed of such lamellæ, and that the lamination is brought into view by the process of solution. They seem further to hold that hemitropic twinning is the cause of the phenomenon. It is curious that their drawings of crystals 32 and 33 show the protruding edges of the lamellæ as running nearly parallel with the edges of the rhombs, a feature not actually possible in nature. Indeed, in this respect their version of the aspect of 33 differs absolutely from that of Rose and Sadebeck, made some 40 years earlier, and which they reproduce for comparison with their own.

Boutan also regards these laminated diamonds as macles by hemitrophy. He further regards Brewster's celebrated lens of diamond as having been cut from a multiple twin of this kind-a re

mote possibility, maybe, though not quite a pro. bability, if only for the reason that a laminated stone is not likely to be transparent enough to serve as a good lens.

One reason against hemitrophy is that one sometimes comes across laminated stones which could just as well be called in the French way "macles by penetration," i.e., tending to conform geometrically to the interpenetrating twins of the plus and minus tetrahedra; the lamination giving a terraced aspect to the blunt protruding pyramidal bosses very like the terraced diamonds from Jagersfontein in miniature. Another reason is that lamination is limited to four definite directions each of which is parallel to an octahedron face, so that in any laminated stone each multiple twin set intersects another at a constant angle; and this is the case even when the specimen is a macle. But macles when they intersect one another are not limited to definite directions at all; their seams crossing almost at random much as irregularly twinned simple crystals interpene

trate.

Lamination has an intimate correlation with colour. With very few exceptions, all laminated diamonds, no matter where they come from, are coloured-brown, mauve, green, or blue-white. Slight lamination may also be seen once in a way on poor cape-white diamonds, and on poor yellow ones. Reflexively, nearly every brown, and nearly every mauve, dodecahedral diamond, either macle or simple crystal, whether the tint be light or dark, is plainly laminated. (The featureless brown diamonds of resinous lustre from SouthWest Africa do not show lamination so often as brown diamonds from other sources. For that matter South-West African diamonds seldom show any surface detail at all to speak of, saving either an exceedingly high polish or a roughening due perhaps to wind erosion). So are most bluewhites when they are of a milky transparency or when their tint inclines to mauve. Lamination also occurs among the green diamonds from the Rand banket.

Hitherto I have only succeeded once in seeing signs of lamination in the interior of a diamond. This was in an ugly dark-brown brilliant in which the lamella were marked out in alternating lighter and darker brown streaks. A fair inference seems to be that a diamond going in general through stages of growth and quiescence may, during the latter, become covered with colouring matter which is enclosed in a later growth. The apparent uniform tint of the final whole stone will be largely a refraction effect (like the grey colour imparted to coated stones by indefinitely thin layers of tiny black spots), and is the happier as the laminæ are thinner. Thus the colouring matter is an overgrowth which establishes the lamination. There is no doubt that Heddle when he alleged the disseveration of "plate diamonds" ("Ency. Brit.," 9th Ed., Art. "Mineralogy"), was confounding the mythical bursting of smoky stones with the lamination of brown ones.

(Addendum.-While the above paper was going through the press I had the good fortune to see a most important specimen of diamond embedded in calcite, found by Mr. J. T. Vigne in the working over of some old Kimberley lumps. The diamond appears to be of about four carats, and the enclosing calcite perhaps five carats. The dia

mond has evidently not been squeezed into the calcite, but has acted as a nucleus upon which the latter has crystallised. Mr. J. Stewart has, also, given me a portion of a beautiful shell of calcite taken from a Wesselton blue-ground diamond. The interior faces of this shell have acquired, from their intimate contact with the diamond, an almost perfect adamantine lustre).

FURTHER STUDIES CONCERNING
GALLIUM.

ITS ELECTROLYTIC BEHAVIOUR, PURIFICATION,
MELTING POINT, DENSITY, COEFFICIENT OF EX-
PANSION, COMPRESSIBILITY, SURFACE TENSION,
AND LATENT HEAT OF FUSION.*

By THEODORE W. RICHARDS and SYLVES ER BOYER.

a

1. Electrolytic Behaviour. INTRODUCTION.-The study of the electrolytic behaviour of gallium naturally had precedence in this investigation, because this behaviour had an important bearing on the preparation of the material needed for the rest of the work. The earlier experimenters upon gallium commonly used electrolysis rather as a means of precipitating the metal from purified salts than as a means of purification. In recent paper, Dennis and Bridgman have pointed out the value of electrolysis for the latter purpose (Dennis and Bridgman, Jour. Amer. Chem. Soc., 1918, xl., 1540; on p. 1537 references to earlier work are given). Independently, we also had simultaneously thus used electrolysis (Richards and Boyer, ibid., 1919, xli., 133). Whereas the earlier experimenters commonly used alkaline solutions, Dennis and Bridgman, as well as ourselves, worked with more or less acid ones, free from alkali salts, thus eliminating the danger of contamination with alkali metals.

Although in the main the verdicts of the recent researches agree, several obscure points need elucidation before the matter is entirely consistent and comprehensible. The points especially to be investigated were the following: first, the single electrode potential of gallium, and secondly, the order of precipitation of indium, zinc, and gal lium, with several different current densities from solutions of several different acidities.

The Single Electrode Potential of Gallium. — No adequate measurements of this potential appear in the literature, but the element is usually considered as coming between zinc (0.52) and aluminium (10), being nearer to that of zinc (see for example, Abeg's "Handbuch der anorg. Chem," 1906, iii., [1], p. 367; Compt. Rend., 1875, lxxxi., 493): This conclusion is based upon the earlier work of Lecoq and Boisbaudran.

For our measurements an accurate potentio meter, standardised by means of a Weston cell. was used. Cadmium and zinc were also measured against a calomel electrode in order to be certain that the apparatus was functioning properly. The single potential difference of cadmium in normal solution of its sulphate was found to be 0.176, and that of zinc in normal solution of its sulphate, 0'521. Hence the apparatus was adequate. Care * From the Journal of the American Chemical Society, February, 1921

was taken, in many of the trials, to have solutions of gallium sulphate neither basic nor with excess of acid, by converting weighed amounts of gallium (through the nitrate and weighed amounts of sulphuric acid) into sulphate. These precautions were, however, probably not necessary, because a small concentration of free acid does not usually much affect a single electrode potential. With amalgamated zinc, for example, 3 N sulphuric acid was found by rough measurements to affect the potential by less than o'05 volt. The potential of the decinormal electrode was assumed to be +0:56 volt. No allowance was made for the unknown solution-solution e, m. f.

In every case gallium showed at first a much smaller negative potential than the final value. With metal which had been exposed for a long time to the air the initial value was even positive (as much as o2 volt). In the course of two or three days the gallium potential gradually attained the maximum value 0 297 in o N solutions; somewhat less (-0.25) in N solutions, and even less when excess of sulphuric acid was present. 01 N gallium alum as electrolyte gave about the same potential (-0.294 volt) as pure o' N gallium sulphate.

Adequate comparison of the effects of ion concentration, corresponding to the Nernst equation, can hardly be made, because the ion-concentration of gallium sulphate has never been determined. The substance is much hydrolysed, and probably hydrolysis diminishes the concentration of the gallium ion in dilute solution, making a colloidal hydroxide.

Liquid gallium decreased in potential with the time of immersion, to a lower value (-018) than the solid. Resolidification reversed this effect, but the gallium thus liquefied never assumed the value it would have had if its potential had been first determined as a crystalline solid. The decrease was hardly consistent enough to substantiate Rudorf's conclusions (Rudorf, Abegg's "Handbuch," loc. cit., p. 366) concerning the free energy of the liquid based on Regnauld's measurements (J. Regnauld, Compt. Rend., 1878, lxxxvi., 1457 (not Regnault, as Abegg gives the name).

Whichever final value is chosen from the abovementioned results, the single electrode potential of gallium thus measured stands between that of indium (010) and that of zinc (-0.52). This was one reason for the statement to that effect in our earlier paper, because preliminary measurements of the electrode potential had been made before that paper was published. It will be seen, however, that this outcome gives an incomplete picture of the situation.

De

Another method of attacking qualitatively the problem of relative electrode-potential is by placing one metal in the solution of the salt of another, in order to determine if the second metal is precipitated at the expense of the first. Boisbaudron's description of his result in this direction is not easily interpreted, but the simple experiment seemed worth repeating. Accordingly small amounts of fresh electrolytically precipitated zinc were placed in two portions of 5 cc. each of N gallium-sulphate solution. After 15 hours, most of the gallium was precipitated in the form of a voluminous semigelatinous, white basic precipitate. In one case the slight metallic residue (after thorough washing, and the solution of

the zinc in very dilute acid), gave spectroscopic evidence of gallium-but the amount was only a very small fraction of the whole, and may have been due to inclusion. Evidently zinc does not precipitate metallic gallium in the definite fashion in which, for example, it precipitates cadmium.

The contrasting experiment of placing gallium in a nearly neutral zinc-sulphate solution was likewise made. After standing for 24 hours, the nugget of gallium showed no sign of a "tree," even in the microscope, and its surface appeared almost, if not quite, as bright as at first. Evidently gallium does not precipitate zinc.

When a solid alloy of gallium and zinc is treated with a little diluted sulphuric acid, both zinc and gallium are dissolved until the acid is exhausted, and no "tree" of either metal forms on the etched and pitted button.

These qualitative experiments leave one still largely at a loss with regard to the electrode potential of gallium. They seem to indicate that this quantity may be nearly the same as that of zinc. A clue to the answer to the puzzle was afforded by the following experiments. Small fragments of the purest gallium, freshly cut, were placed in solutions of cadmium sulphate and of copper sulphate. Neither fragment showed more than a very slight tendency to precipitate these metals, each of which (especially copper), must have a lower solution tension than gallium. After six hours no evidence of precipitation at all was manifest, and even after several days, only a few minute spots of copper could be seen. (Afterward, following a suggestion by A. B. Lamb, copper chloride was tried. Here also the precipitation of copper was very slow in starting, but when once started it proceeded more rapidly with the chloride than with the sulphate). Clearly gallium, like its analogue aluminium (see for example v. Deventer and v. Lumnel, Z. physik. Chem., 1909, lxix., 136), has a distinct tendency to become passive (or "ennobled"), which must vitiate all attempts to determine its true active electrode potential by the methods heretofore recounted. Accordingly, another line of attack was begun.

The Electrolytic Precipitation of Gallium, Indium, and Zinc from Solutions of Varying Acidity and Concentration. The outcome of electrolytic precipitation depends, at least in part, of course, upon the electrode potentials of the substances concerned; but it depends also on at least three other conditions of electrolysis, namely, on the relative concentrations of the ions, on the current density at the cathode, and on "over-voltage" phenomena.

The effect of the concentration on the cation, as expressed by the Nernst equation, is well known. The solution tension of the solid metal is assumed to remain constant, but the effect of the opposing osmotic pressure of the cation increases with its logarithm; hence the single electrode potential decreases when the cation becomes more concentrated. For a trivalent ion such as indium or gallium, the theoretical increase in electromotive force is about o'02 volt for each tenfold dilution at room temperatures. In order to change the electrode potential by as much as o' volt, the solution must obviously be diluted to at least one hundred thousand times its original volume (or possess a concentration of 0'00001 of its original concentration). Hence this effect can

not play a very important rôle in causing two metals o2 or o'3 volt apart in the electrochemical series to appear together.

As an example of the combined effect of changing current density and "over-voltage," the follow. ing fact may be cited. Zinc, although commonly possessing a far greater solution potential than hydrogen, can be precipitated in part from normal solution of the sulphate in the presence of 2N sulphuric acid by a current density of 0'4 amp./cm2., whereas this precipitate will redissolve while the current is running if the current density is reduced to 0.2 amp./cm2. In brief, there is a fairly definite relation between the current density, the concentration of the acid and the concentration of the metal which remains in solution after longcontinued electrolysis. These considerations are of importance in interpreting the separation of metals electrolytically, especially if conclusions concerning relative electrode potentials are sought.

As

The following experiments upon the electrolytic deposition of gallium, indium, and zinc, separately and together, were comprehensive, but will be summarised as briefly as possible. In all of these experiments sulphuric acid was present in definite, but not always identical, excess over the amount corresponding to the bases present. the electrolysis proceeded, sodium hydroxide was added in amounts needed to neutralise the acid formed and keep the hydrogen-ion concentration about constant. The nature of the electrolytic precipitate was determined partly by its melting point, but usually by spectrum analysis. The three metals chiefly concerned (gallium, indium. and zinc) give clear spark- and arc- spectra, alone or together; our experience agrees with that of Dennis and Bridgman in showing that very small traces of each metal may be detected in the presence of the others. The characteristic lines, being in the blue and violet, are all easily photographed. Our comparisons were made almost entirely in this way with a good glass-prism spectrometer. The metals were volatilised by an arc spark obtained from a half kilowatt transformer from the 110 volt, 60 cycle, alternating current of the laboratory. The spark was intensified by a suitable condenser in parallel with the transformer, and the air lines were cut out by adequate inductance. Platinum points about 2 mm. in diameter were used as terminals of the spark gap, and the metals to be analysed were precipitated electrolytically directly upon the terminals, although these were not the electrodes of the fractional precipitation. Blank tests were frequent.

To recount first experiments upon the metals taken separately:

The

Comparable results were obtained with platinum foil electrodes of 8 sq. cm. area and 03 ampere current (that is, an approximate current density of 0:04 amp. per sq. cm.). The volume of the solution in each case was about 100 cc. or less, and during the electrolysis each was continuously and thoroughly stirred by a mechanical stirrer. amounts of indium, zinc, and gallium originally present were respectively about o'4g., 0'2 g., and O'I g. The precipitated metal was completely transferred electrolytically from time to time to the spark gap terminals. With solutions of 0.2 N acidity, about o'6 mg. of indium remained per 100 cc. solution after 12 hours' electrolysis, four

in the electrolysis conducted in this way, most of it appeared in the final deposit obtained after neutralising the free acid. This fact is in accord with the results obtained with zinc alone, in which it was shown that this metal cannot be precipitated completely by current density o'04 from a solution o N in acid.

The results with more concentrated acid pointed in the same direction and need not be recounted in detail. No metal at all was precipitated by this low current density from a solution containing 04 g. per 100 cc, and 17 N in sulphuric acid; very little indium, less zinc and no gallium from such a solution normal in sulphuric acid; and nearly all the indium, most of the zinc, and very little gallium even from a solution o5 N in sulphuric acid.

These results are all consistent; they point indubitably to the following order of precipitation : indium, zinc, gallium, and indicate that, whatever the cause, gallium has a considerably higher deposition-potential than zinc in these acid solutions. But this result is inconsistent with the direct potential-measurements already recounted -for these placed gallium between indium and zinc.

Turning now to the electrolysis of mixtures, the following methods were used and results obtained. The volume was in each case about o'5 litre; the same current density as before (namely, o'04 amperes per square cm.) was maintained; and the acidity was initially o'15 N. When precipitation was almost finished, the acidity was reduced in order to recover all of the gallium. Six successive electrolyses of mixtures were conducted. In every case zinc was detected in the first precipitate (which was chiefly indium) and it was found at every stage of the electrolysis, successive portions of the precipitate being taken out and examined spectroscopically. Evidently indium and zinc are too near together to allow easy electrolytic separation, although theoretically their electromotive ranges should not seriously overlap; and the same is true of gallium and zinc. But in every case gallium began gradually to appear only after most of the indium had been precipitated. Hence these two metals are far enough apart to be easily separated.

Similar results were obtained with platinum points giving much greater current density, in strongly acid solutions. When a platinum point is used for deposition instead of a foil, the current density is so great that the ionised hydrogen in the immediate vicinity of the electrode is not enough to carry the current even when the solution is strongly acid. Hence other cations of greater solution potential may come down with the hydrogen. By preference, of course, those with smallest negative solution potential come out first. Thus, with a platinum point in 5N sulphuric acid, we found that almost pure indium with melting point 155° came first out of a mixture of the three metals under consideration, if sufficient current density were used to precipitate anything besides hydrogen.

Thus there is no real inconsistency between our results and those of Dennis and Bridgman. Different conditions produced different results. The amount of zinc in the specimen involved in our early trials was so small that little of it could be deposited by the current density and with the acidity then employed. Hence, most of the zinc remained in the electrolyte to the end and the order of deposition in this case was gallium, indium, zinc, as stated. The fact that most of the zinc was deposited in the early part of the electrolysis of Dennis and Bridgman is easily explained if their solution (as is probable from their description) was decidedly less strongly acid than

ours.

The explanation of the apparent inconsistency between the relative magnitude of the electrodepotentials and the order of actual deposition seems to indicate that gallium possesses a larger negative electrode-potential in the act of depositing from a solution than it possesses when actually deposited. In other words, it probably possesses some degree of "passivity" when in a metallic state, as already indicated by other tests. This conclusion is supported by the fact that gallium is markedly passive toward dilute nitric acid, although it dissolves in concentrated nitric acid It seems to occupy a place, in this respect, between indium, which dissolves easily in nitric acid of any concentration, on the one hand, and aluminium, which is passive towards dilute sulphuric and concentrated nitric acid, on the other hand. Our experiments on electrode-potential indicated that liquid gallium is even more passive than the solid. This agrees with the fact that liquid gallium is less easily attacked by acids than solid gallium, although this fact may be due, as Rudorf has pointed out (Rudorf, Abegg's "Handbuch," loc. cit.), merely to the possession of more points of attack in the solid. All these questions will receive more detailed examination here in the near future. An answer to the vexed question as to the cause of passivity in general, or of this case in particular, is left in abeyance for the present.

II. Purification of Gallium.

The foregoing statements make clear the advantage of electrolysis as a first step in purification. The elimination of indium and all metals of lower solution tension is easy by this means. Accordingly this method was used as an initial step in the purification of the large quantity of gallium needed later for determining the properties of this substance.

The choice of electrodes for electrolysis was made with care. The liquid metal alloys with silver, making it brittle; and the deposited gallium contains dissolved silver. Hence this metal is entirely unsuitable. Iron seems to show in less degree a similar effect, except that the solid alloy is soft instead of being brittle. Gallium alloys slightly even with platinum foil, which is blackened afterwards by acid. Moreover, the

liquid gallium removed from this platinum foil may contain traces of platinum.

A point of wire is evidently safer than a large piece of foil, since the action of one metal upon the other can take place only on the surfaces where they adjoin. Thus gallium from an iron point was found to contain no appreciable iron. That from a platinum point is even safer, since platinum is attacked less than iron is. Finally, in preparing our purest material we used as the electrode a gallium cathode contained in a cup and making contact beneath its surface with a platinum point of minimum dimensions.

During the electrolysis of a solution of the raw material (a crude gallium indium alloy) the hydrogen evolved was found to give a mirror in the Berzelius-Marsh apparatus. Part of this sublimate was shown to be arsenic, and germanium was indicated by the line 4033, although the two more prominent lines, 4227 and 4180, were often absent. A line very close to 4058 was likewise found. This was most probably due to lead, although this metal does not ordinarily form a volatile compound with hydrogen on electrolysis. It is true that Panett and Fürth (Panett and Fürth, Ber., 1919, lii., 2020) state that they have found such a compound (the particulars to be given later) but we have not been able to reproduce it. Whether the apparent presence of lead and also trace of zinc in the Marsh-Berzelius mirror were due merely to the spattering (which, however, had been carefully guarded against) or to traces of volatile compounds we do not pretend to decide. That arsenic was present there could be no doubt. The spectrum was compared with the spectrum of pure arsenic in a Plücker tube at 220° to 230°, taking the helpful precautions of Herpertz (Herpertz, Dissertation, Bonn, 1906). The suitable vacuum was obtained by a Langmuir

pump.

Incidentally, mention may be made of a black precipitate consisting chiefly of gallium, and of a spongy aggregation which appears at the cathode under some conditions, the latter especially when the solution is almost neutral. It was easy to prove that both consisted of metallic gallium mixed with a trace of basic salt and bubbles of hydrogen which prevent its complete coherence.

Gallium deposited from the solution of the crude liquid alloy, when most of the indium had been separated electrolytically, contained considerable zinc. This could be, in part, eliminated by fractional electrolysis, but the process is slow, as may be inferred from the foregoing account.

Two methods were therefore employed which were found to be effectual in removing the last trace of this metal, namely, ignition at a red heat in a vacuum (when zinc distils by virtue of its lower boiling point) and fractional crystallisation of the gallium with centrifugal draining (when all heteromorphous materials remain in the liquid).

The purification by heating in vacuum is most conveniently carried out by placing the electrolytic metal in a silica boat enclosed in a silica tube, and heating for several hours in a good vacuum until the boat-load remains almost constant in weight, Complete constancy is not attained, since gallium itself is slightly volatile at 800°. The sublimate was shown by suitable

qualitative tests to consist chiefly of zinc. The residue from this treatment must still have contained a trace of some unknown impurity since its melting point was still not quite constant.

a

Fractional crystallisation of the metal was the method used to complete the purification. Liquid gallium, supercooled to a temperature slightly below its true freezing point, was inoculated with trace of the solid phase, introduced on a platinum point. Crystals of gallium 4 to 6 mm. in length were allowed to form slowly. Consisting of the purer gallium, these crystals were removed, but upon their surface they carried some of the less pure liquid. This was removed by means of a hand centrifuge, the inner vessel of which consisted essentially of a test-tube with a much constricted place near the bottom. The crystals of gallium rested upon the constriction, and the liquid was driven by centrifugal force through the narrow opening, while the test-tube, resting upon a pad of cotton wool in a hollow wooden cylinder attached to a stout leather strap, was whirled in a two-metre circle. Before the slightly warm glass vessel was placed in the wooden receptacle, the latter had been warmed to 32 in order to prevent solidification of the adhering liquid before the centrifuging was complete. Two or three successive recrystallisations were enough to bring the purest fraction to a constant melting point, but in order to economise the small amount of fairly pure electrolytic material at our disposal as much as possible, many systematic recrystallisations from the mother liquors were conducted, thus purifying more of the gallium

In all 104 g. of pure gallium of constant melting point (that is, showing no change in melting point from the beginning to the end of liquefaction) was separated from about 15g. of the less pure electrolytic metal. This specimen was called Sample E.

III. The Melting Point of Gallium,

De Boisbaudran, in successive trials, at first found the melting point of gallium to be 29.5°, and later 3015, with very small amounts of material (Compt. Rend., 1876, lxxxii., 1036; 1876, lxxxiii., 61). More recently, Browning and Uhler found the value 297° by reading the temperature at which globules of solid gallium fell through a slightly smaller platinum ring (Am. ]. Sci., 1916, [4] xlii., 389). Neither experimenter mentions the precautions taken with regard to standardising the thermometers; and judging from our experience, probably neither had perfectly pure gallium.

Our own preliminary experiments, made by watching a crystal of the metal (either at the bottom of a small test-tube or suspended in a ring immersed beneath a slowly changing bath) usually gave too high results because of convection currents, and the lag in temperature of the metal. Methods of this sort led to the value 30·8° (which we found subsequently to be about a degree too high) for the melting point of our best gallium. An accurate determination of a melting point can hardly be made unless enough of the substance is at hand wholly to surround the registering instrument with a mixture of the two phases whose equilibrium is concerned. Therefore we did not succeed in obtaining satisfactory values until we used a much larger quantity of the metal.