56 BERKELEY: ON THE MORE EXACT DETERMINATION OF V.-Oq$ the More Exact Determinfzation of the Densities of Crystals. By THE EARL OF BERKELEY. THIS communication mas read at the meeting of the British Associa- tion at Bristol in 1898. It has not hitherto been published as it was part of a somewhat lengthy research on the molecular volumes of crystals. The appearance of Messrs. Barlow and Pope's recent paper (Trans., 1906, 89, 1675) has so illuminated the subject, that it is not considered necessary to pursue the matter any further. A comparison of the several values obtained by different observers, for the density of one and the same salt, shows variations which in some cases amount to 10 per'cent. As the density is supposed to be independent of the may in which the crystals have been produced, these discordant results must be ascribed to errors of observation, - Of these errors the following seem to be the most important. (1) Errors in the operation of determining the density.(2) Errors caused by air adhering to the crystals. (3) Errors due to the hygroscopic nature of the salts. (4) Errors induced by the mother liquor occluded in the cryst'als. I n this paper I propose to describe the means whereby the amount of error due t o the first three causes have, I believe, been reduced.THE DENSITIES OF CRYSTALS. 57 ( 1) Opemtwn of Determining the Density.-Two similar conical pyknometers (Fig. l), 'of about 7 C.C. capacity, were made by Mr. Miiller from Jena thermometer glass. These are fitted with carefully ground thermometer stoppers and the thermometers are graduated in tenths of a degree such that the hundredth of a degree can be estimated.The thermometers were standardised a t Kew. The side capillaries are graduated in millimetres and were calibrated by running weighed mercury threads along the bores and reading their lengths. The capacity (7 c.c.) was chosen as being sufficiently great to keep down errors of weighing and of manipulation. capacity of the pyknometer the thicker the walls ; consequently it will take more than proportion- ately longer for the larger quantity of liquid to attain a constant temperature when in the balance case. Also a greater quantity of salt will be re- quired to keep the proportion of salt t o liquid the same, and hence the more difficult it will be t o dry and to free from adhering air.One of the pyknorneters was used throughout the weigh- ings as a counterpoise, and to obtain the most accurate results I found it essential that the surface of each pyknometer should be similarly treated ; if, for instance, the pyknometer itself had been filled with carbon tetrachloride and its surface wetted with that liquid, the surface of the counterpoise should also be wetted with the same liquid and both similarly dried before weighing. The pyknometers, after steaming for twenty minutes, were heated to 130' and cooled; this heating and cooling was repeat,ed about fifty The greater the FIG. 1. times so as to obtain, as far as possible, a constant state of molecular aggregation in the glass. It was filled with distilled water and placed in a small desiccator connected to a Sprengel pump.Vigorous boiling, which was continued for three- quarters of an hour, mas promoted by tapping the bottom of the desiccator. The pyknometer was then placed on the pan of an ordinary spring balance and the stopper inserted and pressed home, the reading of the balance index being noted so that in future the stopper could be pressed home with the same pressure ; the presumption is that with the same pressure and the same orientation of stopper to neck, the former will go home to the same extent. It is well, however, (a) The capacity of the pyknometer was then determined.58 BERKELEY: ON THE MORE EXACT DETERMINATION OF t o check this assumption by redetermining the capacity; for after much use the stopper is ground further into the neck, especially if powdered minerals have been used.After the pyknometer and its counterpoise have been wiped dry, they are placed on the pans of the balance, and when the level of the liquid in the capillary has, in consequence of evaporation round the neck, fallen below the highest graduation, the weight is determined, and at the same time the level in the capillary and the temperature of the liquid are noted; care, however, should be taken that the temperature is fairly constant. From these data the capacity of the pyknometer is calculated in the usual way, all weights being corrected for displaced air. The following are the values obtained for this capacity, reduced t o the zero mark OF the capillary : Temp. Cal'acity. 14-90" 7'15303 C.C.1 4 9 5 7-15322 ,, 14.95 7.15305 ,, 14-97 7'15293 ,, To1np. C:tlacit y. 15'27" 7.15315 C.C. 15.60 7.15208 ,, 15.52 7.15316 ,, .' 15'96 7'15311 ,, * The greatest difference between any two observations is 0.00029 C.C. and this corresponds to an error of 0.004 per cent. The mean of these numbers was taken as the capacity a t the mean temperature, and the capacity a t any other temperature was calculated from the known cubical expansion of the Jena glass. ( b ) The liquid used for determining the volume occupied by the crystals was carbon tetrachloride. It was purchased from Messrs. Kahlbaum and redistilled a t constant temperature after digesting with fused calcium chloride for several weeks. Both during the distillation and when drawing off the liquid for use, the access of moist air was prevented by means of calcium chloride tubes.The density of the carbon tetrachloride was found to be the same a t the end of a year as it was shortly after distillation. The pyknometer is 6lled with air-free carbon tetrachloride in a manner similar to that already described, and then wiped dry as in the case of water. Owing t o the high coefficient of expansion of the liquid, i t is im- portant that the pyknometer thermometer bhould register the true tem- perature of the liquid and glass. A double-walled glass case surrounding the balance case was found insufficient to secure a steady temperature, and eventually a zinc tank filled with water and placed over and round three sides of the balance, whilst the heat of the observer mas cut off on the fourth side by a glass trough filled with water, was substituted and found effectual.By this means, in about one and a half hours, These observations mere obtained after an interval of 5onie weeks.THE DENSITIES OF CRYSTALS. 59 after placing the pyknometer on the balance, the thermometer does not change by more than O.0lo in fifteen minutes. The following table gives the densities of the carbon tetrachloride used, and for comparison, those calculated from Prof, Thorpe's results (Trans., 1880, 37, 199) : Density. Temp. Self. Thorpe. 16-44' 1.60133 1.59980 16.33 1.60157 1'60003 16-17 1.60189 1.60033 15'55 1-60310 1'60155 Diff. 0.00153 0'00154 0.00156 0.00155 Density. Teinp. Self. Thorpe. 14.88" 1.60438 1.60286 14.68 1.60475 1.60324 13'94 1.60620 1.60467 Diff.0.00152 0 '001 51 0.00154 Thinking that the somewhat large differences between our values might be caused by dissolved air, I determined the density of my unboiled carbon tetrachloride with the following results : Temp. Self. Thorpe. Di K. 15.47" 1.60291 1.60171 0*00121 15.35 1 -60320 1.60195 0.00125 To show the effect of the rate of change of temperature when the pyknometers are on the balance pans, I append the following : Density. Temp. Rate of change. Self. Thorpe. Diff. 15-67' 0'04" in 20 minutes (fallii~g) 1.60274 1.60132 0'00142 14.34 0'02 ,, 10 ,, 7 , 1.60534 1~30390 0.00144 17.67 0.03 ,, 20 ,, ,> 159904 159741 0.00163 14.04 0.02 ,, 15 ,, (rising) 1.60584 1'60448 0.00136 16.92 0'02 ,, 20 ,, Y 9 1'60052 1-59887 0.00167 16-42 0.04 ,, 20 ,, $ 7 1.60144 1'59984 0'00160 (c) To determine the density of the crystals, the pyknometer, con- taining a known weight of salt, is placed in the " bulb desiccator " (shown in Fig.2), the bulb of which has previously been half-filled with carbon tetrachloride. by exhausting through A, and warming and tapping the bottom of the bulb. Tap A is then closed, and the exhaustion continued through B for a t least three-quarters of an hour, the bottom of the desiccator being vigorously tapped a t intervals so that the vapour of the liquid may penetrate among the powdered crystals and thus sweep out the air. The apparatus is then disconnected from the pump and while still vacuous is tilted so that the liquid in the bulb flows down the fine capillary, C, into the pyknometer.The pyknometer is then weighed as already detailed. The liquid is then caused to boil60 BERKELEY: ON THE MORE EXACT DETERMINATION OF The following results were obtained for the density of powdered and ignited quartz : Observed in water. - Temp. Density. 21.95" 2.6484 21 '75 2,6487 21.52 2.6487 21'48 2,6487 21 '24 2'6486 Density of Quartz. Observed in carbon tetracliloriile. r A I Density. Temp. 21-14" 2.6486 2.6486 21.07 2.6480 2.6480 20.18 2'6483 2%482 19.99 2'6485 2.6484 18-85 2.6483 2.6480 18'24 2,6489 2.6486 I n the last column, the density observed in carbon tetrachloride is corrected ti5 21' by the use of Fizeau's coefficient of ex- A pansion of quartz. The greatest diff eren'ce between any two of these is 0.0007, which corresponds to a maxi- mum error of 0.02 per cent.It was considered possible that there might be some difference in the densities of large and small fragments of the same substance. A clear specimen of barytes was selected, coarsely pow- dered, a n d t h e n s i f t e d through sieves of different mesh. The density of the fragments retained by the finest mesh, the openings of which average 0.36 by 0.33 mm., was compared with that of the fragments re- tained by the medium mesh The barytes fragments were heated to 200' and The results are as FIG. 2. )1 (0.59 by 0.54 mm.). cooled over sulphuric acid before each observation. follows : Density of Bmytes. Small fragments. Large fragments. Temp. Density. Tcmp. Density . / \ T h > * 15.76" 4'4700 * 16.46" 4'4702 * 15-96 4.4698 * 15.05 4.4707 16.67 4.4696 17.35 4'4701 16.75 4.4702 17-06 4.4697 16.07 4'4703THE DENSITIES OF CRYSTALS.61 The observations marked (*) were obtained with fragments immersed in water, the remainder in carbon tetrachloride. The greatest difference between any two observations is 0.0011, which corresponds to a maximum error of 0.025 per cent. It will be noticed that there seems to be no difference between the two sets of densities. (2) Ail- Ileld by the C?ystals.--In section (c) I have already FIG. 3. described the method whereby it was hoped that the air difficulty had been overcome. It is evident that the shape of the pyknometer lends itself to this method, for the salt, while conserving the same relative proportion of salt to liquid, can lie in a thinner stratum than in either a cylindrical or globular form of pyknometer.The ordinary method of covering the salt with the liquid and boiling in a vacuum leads to loss of salt through spirting, and also to the adhesion of small particles to the inside of the neck which prevents62 MORE EXACT DETERMINATION OF THE DENSITIES OF CRYSTALS. Tenip. Density. 16.20" 2.3320 16'14 2.3322 Tenil). Density. 16'13" 2.3314 17.00 2'3312 The maximum error is 0.04 per cent. (4) Bother Liquor in the CrPstaZs.--No general method of over- coming this source of error was found. Some substances give the same density when obtained under conditions which differ fairly widely, but others, such as potassium chloride, even when crystallised from a solution which was kept at a constant temperature and con- tinuously s tirred, give successive crops which differ markedly in their density. The greater part of this investigation was carried out a t the Christ Church Laboratory, Oxford; I am glad to have this opportunity of thanking Mr. A. Vernon Harcourt for his kind permission to use the resources of this laboratory. * A useful precaution to take is to note the rate at which the carbon tetrachloride evaporates normally (loss of weight) when on the balance ; the presence of particles in the neck will be indicated by an increase i n the rate.