21,

of "bottling up sunlight.". This observation, probably given out with cynical intent, contains the gist of most of the early theories, and even now there remains much to be said in favour of it. Phosphorescent substances or "light magnets" as they were often called were considered to act as veritable light-accumulators, rapidly absorbing light, and afterwards emitting it gradually when placed in the dark. Priestley, in his "History of Light and Colours," says: "The hypothesis of the materiality of light is peculiarly agreeable to the phenomena of the Bolognian Stone, which has the remarkable property of imbibing light, of retaining it for some time, and of emitting it again; and more especally by its emitting it more copiously according to the degree of heat applied to it.

Not all the early experimenters agreed with this type of theory. On the other hand, Kircher put forward a theory in which he supposed that the air "abounds with a subtle vapour which is very easily illuminated." The phosphorescent substance absorbed this vapour (and presumably the illumination also) and gradually evolved it in the dark. But the theory of Kircher and similar ones however, were regarded as fantasies, and only the speculations based on the material nature of light made any progress-that is, if chemical theories did progress at all, in those days.

It is not within the scope of this article to enter into the modern theories of luminescence. Suffice it to say that they are all highly problematical and delightfully at variance with one another. Probably, however, in view of the modern tendency towards a material or semi-material con ception of light and heat we may witness at some not long-distant date the absorption of all these varying and antagonistic hypotheses into one general, revived, and modified theory of The Storing of Light.

THE CHEMISTRY

AND CRYSTALLOGRAPHY OF SOME FLUORIDES OF COBALT, NICKEL, MANGANESE AND COPPER.*

By FLOYD H. EDMINSTER and HERMON C. COOPER.

Introduction.

THE fluorides of the bivalent heavy metals such as cobalt, nickel, and copper, have never been fully investigated, neither have their composition and crystalline forms, including isomorphous relations, been satisfactorily established. amination of the description of these compounds in the handbooks (e.g., “Gmelin-Kraut”) leaves one in doubt as to the facts and relationship.

An ex

In 1824 Berzelius (Berzelius, Pogg. Ann., I., 1824, p. 28; Ann. chim. phys., [2] 1823, xxiv., 61) prepared the fluorides of cobalt, nickel, and copper, and regarded them as so similar that he described them together. "If the carbonate (of the metal) is treated with hydrofluoric acid, it will dissolve with effervescence, but soon a salt is precipitated as a heavy powder. If more and

*A more detailed account of this work was submitted to the Faculty of the Graduate School of Syracuse University by Floyd H. Edminster in May, 1918, in partial fulfilment of the requirements for the degree of Doctor of Philosophy.

crust.

more of the carbonate be added, the effervescence continues, but the salt already formed decomposes, especially if warmed, and there results a pulverulent basic salt. If the addition and decomposition of the carbonate be stopped before this salt (powder) forms and the solution be evaporated, there separates out a crystalline In this process there is given off the excess of acid which the dissolved salt contains. If the crystallised salt be covered with a very small amount of water and the mixture be allowed to stand for a long time at room temperature, there results a saturated solution which will deposit these same crystals on evaporation. however, the mixture is heated to boiling with much water, decomposition occurs, part of the salt dissolving in the liberated acid and another part remaining undissolved as a basic salt."

If,

Berzelius selected the copper salt for anlysis, and found that on heating it with lead oxide there were given off two molecules of water, not in the least acid. The green pulverent salt resulting from the decomposition of the neutra Isalt by boiling water proved, by the same method of analysis, to be basic salt of the formula CuO.CuF. H2O.

The three metals, cobalt, nickel, and copper, were thus regarded by Berzelius as forming entirely analogous fluorides; a crystallised normal fluoride, MF2. H2O, and an amorphous basic fluoride.

About fifty years later, F. W. Clarke (Clarke, Am. Chem. Journ., 1887, xiii., 290) described the preparation of a supposedly new fluoride, made by evaporating a solution of nickel hydroxide in hydrofluoric acid, whereby a crystalline crust is formed. The analysis of this crust gave the formula, NiF2.3H2O.

In 1884, Balbaino (Balbaino, J. Chem. Soc., 1884, ii., 1264) stated that he prepared the hydrated cupric fluoride (CuF2.2H2O) of Berzelius by dissolving copper carbonate in hydrofluoric acid and adding 95 per cent alcohol, which precipitated a pale blue crystalline powder.

Poulenc (Poulenc, Compt. Rend., 1892, cxiv., 1426; Ann. chim. phys., [7] 1894, ii., 47; Ber., 1892, xxv., R. 662) is credited with the first description of the anhydrous fluorides. They are only of incidental interest to us here, but it may be noted that, according to Poulenc, the anhydrous nickel fluoride is formed as an amorphous yellow powder or as green crystallised prisms. The anhydrous cobalt fluoride is either a rose-red amorphous powder or red prismatic_crystal. In 1905, Böhm (Böhm, Z. anorg. Chem., 1905, xliii., 330) reviewed the work on the fluorides of the heavy metals, called attention to the lack of accurate inevstigations of well-crystallised material, and prepared, along with several other complex fluorides, the acid fluorides of cobalt, nickel, and copper. He stated merely that he obtained these by dissolving the freshly precipitated hydroxides or carbonates of the metals in hydrofluoric acid and concentrating until crystals appeared. His cobalt and nickel fluorides are describe as, respectively, red and green prisms, having similar form and composition. They are stable in the air and easily soluble in water and dilute acids. The metal was determined by electrolysis from a sulphate solution, the fluorine by heating the salt with pure Iceland spar and the

28

ments.

Notwithstanding the slight experimental foundation on which the existence of the normal fluorides is based, one is inclined to prefer a normal salt formula to an acid salt formula for a halide obtained from the hydroxide or carbonate and acid, because it is in this way that the most familiar normal chlorides and bromides are prepared.

We have undertaken a combined chemical and crystallographic study of these fluorides, not merely to ascertain the facts in question, but to see whether an isomorphous series is involved.

Preparation of Materials.

The following general methods were investigated. (1) Treating the hydroxide of the metal with an excess of hydrofluoric acid. The excess is necessary for avoiding the formation of the basic salt. (2) Decomposing the carbonate with hydrofluoric acid. In either of these cases the resulting solution is evaporated until crystals appear. Both methods yield satisfactory results.

(3) "Dissolving" the metal in hydrofluoric acid. In most cases the action was so slow that only a very small concentration could be obtained. (4) Treating a salt of the metal (the acetate) with hydrofluoric acid. (5) Double decomposition of an alkali fluoride and a salt of the metal. A complex fluoride resulted.

The first general method was ordinarily employed. The hydroxides were precipitated from the acetate solution with sodium hydroxide, the precipitate being washed repeatedly by decantation with cold water and finally with hot water, until the filtrate gave no colour with phenolphthalein. The washing must be very thorough, since the gelatinous hydroxide adsorbs the alkali greedily. An excess of alkali was avoided and the precipitate was heated in an excess of water to boiling repeatedly before bringing it upon the filter. The moist hydroxide was then dissolved in C. P. hydrofluoric acid in a platinum basin and the solution was filtered through a wax funnel into a wax bottle, from which it was taken as needed.

Böhm specifies that the solution should be evaporated in a vacuum dessicator over sulphuric acid until crystals appear. Attempts to follow

this procedure resulted in the formation, not of crystals, but of a crystalline crust on the sides and bottom of the container and an arborescent creeping growth extending often over the sides of the vessel. The walls of the container and the dessicator were bedewed with a strongly acid solution, which was not absorbed by the sulphuric acid. If the evaporation took place in the open air, the crust was drier and harder and more difficultly soluble.

Under the microscope this crust looked different according to the conditions under which the evaporation took place. In some cases the crust was only a mass of nodules, while in others it appeared to be a mixture of amorphous material and of small but fairly definite crystals. In the case of copper the crust was, fortunately, almost wholly crystalline.

In view of the fact that the anlyses of Berzelius and Clarke were made on material prepared in this way, special attention was given to preparing a product which could be regarded as suitable for analysis, but examination with the petrographic microscope showed the preparations-except in the case of the copper salt-to be either amorphous and of indefinite water content or mixtures of amorphous and crystalline material. Analses of these products were made and indicated roughly a ratio of cobalt or nickel to fluorine of 1: 2, but the character of the crystals was too indefinite to warrant careful investigation of their composition. Recrystallisation.-If this crust is placed in about an equal weight of water, the crystalline portion can be seen to dissolve out, while the amorphous portion remains. If the amount of water is increased and the temperature raised, all of the material dissolves. Except in the case of cobalt no well-defined crystals, capable of measurement, were obtained from the original solution of the hydroxide in hydrofluoric acid. All of the crystals used for analysis and measurement have been obtained from the water extract of the crystalline crust, the extract being allowed to evaporate slowly in the air. Even then some of the crust is nearly always formed around the crystals and a second recrystallisation is necessary to free them entirely from the crust. Each recrystallisation reduces the acid concentration until in the third is practically negligible.

We hesitated to recrystallise from pure water because of the possibility of hydrolysis and, particularly, because of the possible loss of acid in case the salts should be acid fluorides, as our investigation as well as Böhm's disclosed. Contrary to expectations, the proportion of amorphous material lessened, rather than increased, with decreasing acidity, indicating that it is not a basic salt. The acid concentration, however, appears to control the equilibrium involving the crystal and amorphous phases.

Preparation of Cobalt Fluoride.-Our first preparation of cobalt fluoride was by treating the hydroxide with hydrofluoric acid. Because of the tendency of the cobalt (ous) hydroxide to oxidise in the air during the long washing process to cobaltic hydroxide, wihch is insoluble in hydr. fluoric acid, we generally used the carbonate instead. It is permanent in the air, filters more quickly and is more easily washed. It is, however, less soluble than the cobaltous hydroxide. To be sure, the carbonate, when added to the

hydrofluoric acid, is decomposed as long as there is any acid present, but, as Berzelius stated, there is precipitated a fine rose-red powder, which is dissolved only upon the addition of more acid and water.

The original solution was evaporated on a water-bath until solid appeared, whereupon it was removed to a heat-insulated box of the Swedish box, or fireless cooker type, and allowed to cool slowly. No crystals formed. Even after standing in the box until half of it had evaporated, the solution continued to yield a brittle crystalline crust, the last portion being the same as the first.

The next trial was made in the open air, using the mother liquor from this crust, but no real crystals were obtained. The same method was tried, starting with a dilute solution and allowing it to evaporate slowly at an even temperature, with no better results. Böhm's method of letting stand in an evacuated dessicator over sulphuric acid was then tried. The evaporation proceeded slowly because, as stated above, the escaping vapour was not absorbed by the sulphuric acid but condensed upon the inner wall of the dessicator. Solid fluoride formed upon the sides of the basin and then crept over the edge. Under the microscope this crust appeared to have no crysta! form and no definite composition.

The cobalt fluoride crust, however formed, is slowly soluble in very dilute hydrofluoric acid. For this reason a few drops of hydrofluoric acid were added to the water on dissolving. Upon evaporation the same results were obtained as in the first evaporation. Since the crust appeared to consist of more than one form of material, differing in solubility, a comparatively large amount was treated with a small quantity of water and warmed. The filtered solution upon standing for six hous began to deposit crystals which, though accompanied by crust, were sufficiently distinct from it to be separated mechanically. When these crystals were recrystallised from slightly acid water, very clean, well-defined crystals were obtained.

Preparation of Nickel Fluoride. For the preparation of nickel fluoride either the hydroxide or the carbonate may be used, since the hydroxide, unlike that of cobalt, is stable in the air. The first product was a crystalline crust, as for cobalt. The crust forming on the surface of the evaporating solution is identical in appearance under the microscope with that deposited on the bottom. Satisfactory crystals were obtained in the same way as for cobalt, the cleanest ones resulting from a second recrystallisation from water.

Preparation of Manganese Fluoride.-Of all the fluorides investigated that of manganese is the most difficult to prepare. The hydroxide is less stable in the air than that of cobalt; so that, before it can be thoroughly washed, much of it has become insoluble in hydrofluoric acid; but a more serious difficulty is the comparatively slight solubility of the manganous hydroxide in hydrofluoric acid. The carbonate likewise dissolves only to about 10 per cent; further addition of the carbonate produces a marked effervescence and the precipitation of a fine powder, which does not dissolve upon the addition of more acid and water. The solution of the fluoride is consequently very dilute.

If the manganous fluoride solution be heated on the water-bath, the fluoride precipitates as the powder, so that all evaporation must be done at room temperatures, requiring on the average six weeks to produce crystals. The crust also forms as in the case of cobalt and nickel, but decomposes when digested with water. We attempted to obtain manganese fluoride by dissolving the metal in hydrofluoric acid, but obtained only manganese dioxide, which was formed by an unusually vigorous reaction between the metal and acid.

Preparation of Copper Fluoride.-For the preparation of copper fluoride the hydroxide is to be preferred to the carbonate, since the hydroxide reacts faster with hydrofluoric acid. Although it forms a basic salt upon standing in the air, the product is soluble in more acid. If the copper fluoride solution be evaporated slowly, small irregular blue crystals form on the bottom of the basin; while, if the evaporation takes place on the water-bath, the fluoride appears as a clear crystalline crust of constant composition. Both the small crystals and the crystalline crust were free from the granular, amorphous material so persistent in the deposition of cobalt and nickel fluorides. They were regarded as sufficiently well defined for analysis. The analyses showed the two products to be identical in composition, viz. to be a normal fluoride.

These normal fluoride crystals are difficultly soluble in water, a white coat forming upon the surface. If an excess of water is added and the solution is heated, the filtrate yields upon evaporation well-defined three- or sixsided prismatic crystals. These are very soluble in water and, strange to say, have the composition of an acid fluoride. On exposure to the air they effloresce rapidly. If they are heated, water and hydrogen fluoride are evolved, leaving copper oxide in the tube. Presumably the hydrogen fluoride is driven off first, whereupon the water of crystallisation hydrolyses the copper fluoride. Observations on Dissociation.

The crystal fluorides of cobalt, nickel, manganese, and copper all evolve hydrogen fluoride when exposed to the air. After a portion of one of these substances is washed and dried and set aside on the watch glass the latter is found to be etched in three minutes. The gradual loss of hydrogen fluoride is a very important property of these fluorides. It means that the compound can not be preserved with certainty of its retaining its original composition and that analyses must be made with the freshest possible produc: This phenomenon will be referred to the discussion of the analysis. It would be interesting to deter mine the rate or decomposition, but no work has been done upon it thus far.

The cystals are soluble in water and dilute acids. When heated, they give off water and hydrofluoric acid and are themselves converted to the oxides, as stated by Böhm, for the copper salt.

Analysis of the Fluorides.

For the metal the following general methods are applicable: (1) precipitation as the hydroxide or sulphide with subsequent reduction in a stream of hydrogen; (2) evaporation with sulphuric acid

and weighing as the anhydrous sulphate; (3) heating at red heat to convert the fluoride into the oxide; (4) electrolytic deposition.

The total fluorine can be determined volumetrically by adding an excess of standard alkali and titrating back with standard acid. In this case the precipitated hydroxide is filtered off before the titration, else it will obscure the endpoint. With nickel and copper, using phenolphthalein as an indicator, a very good end-point can be obtained if the alkali be added slowly to the warm solution until the free hydrofluoric acid is neutralised and the hydroxide of the metal is precipitated.

The gravimetric determination of fluorine is based upon its precipitation as an insoluble fluoride. A number of metals form insoluble fluorides, calcium being the one most commonly used. Lead chloride precipitates the fluorine as the double halide, PbFC (G. Starck, Z. anorg. Chem., 1911, lxx., 173; J. Chem. Soc. Abs., 1911, c., ii., 436), and lithium chloride is said to give accurate results (Deladrier, J. Chem. Soc., 1904, lxxxvi., 440). In our work with the fluorides lead chloride gave results uniformly low and lithium chloride failed in some cases to give a precipitate. The difficulties of precipitation are well known. Calcium fluoride is an extremely finely divided, slimy precipitate, which passes through nearly all filters the first time. Refiltering through the same filter proved to be exceedingly tedious, and washing was impracticable because of the time consumed in filtration of the solution. However, a number of determinations for fluorine were made by precipitating it with calcium chloride or cal-. cium acetate. While the results were not accurate, they afforded independent confirmation of those obtained by other methods so far as the interpretation was concerned.

The purpose of adding the carbonate to the fluoride in the calcium fluoride method (Berzelius) is to assist in the filtration. A modification of this procedure has given good results in our work. Instead of adding sodium carbonate we made the solution slightly acid with acetic acid. An amount of standard ammonium oxalate was then added so that the calcium oxalate precipitated would be approximately the same as the weight of calcium fluoride. The two precipitates were then filtered off together in a Gooch crucible, dried and weighed. The difference between the weight of the combined precipitate and the weight of the calcium oxalate, corresponding to the ammonium oxalate added, gave the calcium fluoride. The decided advantage lies in the elimination of the decomposition, and furthermore, it does not necessitate a second filtration. The calcium oxalate serves the same purpose as the calcium carbonate and does it equally well. None of the absorption methods was used in the analysis for fluorine.

The determination of crystal hydrate water is the most difficult, since a temperature sufficiently high to expel the last traces will cause a loss of at least a part of the hydrogen fluoride. The method used by Böhm-heating the fluoride with lead oxide-is the only one described as being satisfactory. In the analyses that follow the water determinations were so variable that they were used only as rough indications of the amount present. For calculating formulas use was made of the percentage of water by difference. (To be continued.)

THE MOLECULAR STATE OF WATER

VAPOUR.

By JAMES KENDALL.

UP to a few years ago it was universally considered that water vapour at ordinary temperatures was, so far as could be deduced from vapour density determinations, entirely monomolecular H2O and obeyed the gas laws within the limits of experimental error. Regnault's measurements gave 180 for the molecular weight, Gay-Lussac's 1801 and Leduc's 181, while the formula H2O requires the value 18:016 (Abegg, "Handbuch der anorganischen Chemic," [1] 1908, ii., 67). In the derivation of equations for osmotic pressure, vapour pressure lowerings, &c., in dilute aqueous solutions, the assumption was accordingly made, without any question, that water vapour could be treated as a perfect gas. Some measurements by Winkelmann (Winkelmann, Wied. Ann., 1880, ix., 208) appeared to indicate a tendency towards association at temperatures slightly above normal, but even as late as 1907, in a discussion at the Faraday Society (Wilsmore, Trans. Faraday Soc., 1907, iii., 85) the different results quoted for saturated vapour at 15° varied more among themselves than from the theoretical value, and the conclusion that the vapour behaved as a perfect gas was undisputed.

In 1908, however, some calculations were published by Bose (Bose, Z. Elektrochem., 1908, xiv., 209), based upon new vapour density determinations by Kornatz (Kornatz, Inaug-Diss., Königsberg, 1908), claiming that association in the saturated vapour was considerable even at ordinary temperatures. The equilibrium (H,O), 2H,O was assumed to exist in the vapour phase and an equation for the variation in the equilibrium conThis equastant with temperature was derived. tion indicated for the saturated vapour at o°, 6·6 per cent association; at 50°, 8.2 per cent association; at 100°, 8.9 per cent association.

In 1915, Oddo (Oddo, Gazz. chim. ital., [1] 1915, xlv., 319), evidently ignorant of the work of Bose, calculated from the tables of LandoltBörnstein ("Tabellen," 1912, p. 369) (Regnault's data, reproduced from Zeuner's "Technische Thermodynamik") the molecular weight of saturated water vapour between -20° and 200°, obtaining values steadily increasing from 17.03 at -20° to 19'92 at 200°. Only at 32° did the experimental value agree with that req.red by monomolecular H,O. The conclusion was drawn that at temperatures below 32° a dissociation equilibrium, H,OH++OH-, existed, and above 32° an association equilibrium: (H,O), 2H,O. At -20° the degree of "ionisation" was calculated to be 5.79 per cent; at 200° the "association” reached 1911 per cent. Great emphasis was laid upon the former phenomenon-the spontaneous ionisation of water vapour-as opening up a new field in science, and in a second, article (Oddo, Gazz, chim, ital., [1] 1915, xlv., 395) remarkable deductions were drawn therefrom upon such diverse topics as atmospheric electricity, the influence of temperature on vegetation, the respiratory processes of plants and animals, the electrolytic reactions in a monocellular organism, and

* Presented at the Buffalo Meeting of the American Chemical
Society, April 10, 1919.

21,

the first experimental mechanism for the origin of life.

An adequate discussion of so many important questions cannot be entered into here, but since the molecular weight of water vapour is a matter of fundamental interest in a few minor fields (such as the modern theory of solutions and steam engineering practice) it has been thought profitable to subject the mutually contradictory conclusions of Bose and Oddo to a critical examination, particularly since these conclusions are now quoted without reserve in standard monographs (Turner "Molecular Association," 1915, p. 89). The results below 32° may first be considered.

According to Oddo, saturated water vapour below 32° is partly dissociated into hydrogen and hydroxyl ions. At o° the calculated degree of ionisation is 2.6 per cent, the vapour pressure being 46 mm. From these figures we can determine the concentration of hydrogen and hydroxyl ions in saturated vapour at o°. The value obtained is 7× 10-6 g. ions per litre, in other words, just 200 times as great as the corresponding concentration in pure liquid water at o° (Kohlrausch and Heydweiller, Z. physik, Chem., 1894, xiv., 317). Since the mobility of the ions in the vapour phase would certainly be enormously greater than in the liquid, owing to the diminished viscosity of the medium, it follows that the conductivity of saturated water vapour (or air saturated with water vapour) at o° should be comparable with that of a fairly concentrated salt solution. Now this is admittedly not the case, hence Oddo's whole argument must be quite invalid.

The explanation is not far to seek. Regnault's data at temperatures below the normal (although smoothed off on a curve to render them consistent) are far too inaccurate to be employed as a basis for determinations of molecular weights to the second place of decimals. In fact, the density of the saturated vapour at pressures so low as 4'6 mm. becomes so small that it is impossible to measure it with anything approaching the above order of accuracy. Even the exceedingly careful determinations of Young (Young, Proc. Roy. Soc. Dublin, 1910, xii., 374) on the specific volumes of the saturated vapour of 30 organic liquids are conceded to be uncertain at the lower temperatures investigated, for this same reason. Consequently, until more definite proof of this spontaneous ionisation of water vapour is brought forward, we cannot consider the molecular weight of water vapour to be appreciably diverted from the normal value as a consequence of such ionisation. (The dielectric constant of water vapour is so near to unity as to appear to preclude more than the merest trace of ionisation--Bädeker, Z. physik. Chem., 1901, xxxvi., 305).

It remains to examine the abnormally high vapour densities for temperatures above normal, upon which both Bose and Oddo postulate the equilibrium (H,O), 2H,O. Bose's calculations are dependent entirely upon a series of 19 determinations at 13 different temperatures (7 ranging from 50° to 182°, p ranging from 52.6 to 790 mm.) by Kornatz. It is true that a few isolated measurements of previous investigators quoted, which are in fair agreement with these results, but of the many determinations which do not agree no mention is made.

are

It will be evident from the above table that the observed variation is approximately only 50 per cent of the calculated. Bose remarks, in a passing comment on this fact, that it is possible that the experimental results are affected by systematic errors, which may be partially explained by →>>> assuming that the equilibrium (H,O), 2H,O in the vapour phase is only slowly established. It is extremely improbable that this assumption is correct, since even in liquid water no indication of anything but instantaneous equilibria between the different molecular species present has ever been obtained. If Kornatz's data are trustworthy, Bose's interpretation of them is not justified; if they are not trustworthy, no such interpretation should have been attempted.

We are consequently left to face the fact that, while vapour density measurements for temperatures above the ordinary give abnormally high results, these high results cannot be adequately explained by assuming association in the vapour phase. Indeed, other evidence renders it certain that such association is negligible even at 100°. Let us compare the abnormality in the density of saturated water vapour at this temperature with the abnormalities shown by the saturated vapours of other liquids at their boiling points. The data in the following table are taken from Washburn (Washburn, "Principles of Physical Chemistry," 1915, p. 31).