Anhydrous potassium fluoride dehydrates alcohol somewhat more effectively than potassium carbonate, and like the latter, when added in certain proportions it causes an aqueous solution of the alcohol to separate into two layers. If more water is added, separation no longer occurs, and at the point where this just happens the liquid contains, for a given temperature, definite proportions of fluoride and alcohol. Frank-forter and Frary1 have constructed a solubility curve of potassium fluoride in alcohol-water mixtures, and by means of it have devised a method of estimating alcohol based on the foregoing principle.

A weighed quantity of freshly-ignited potassium fluoride is introduced into a weighed stoppered flask, then a weighed quantity of the alcoholic solution is added, and the fluoride dissolved. The proportions of solution and fluoride are so chosen that the resulting liquid separates into two layers. Water is then added drop by drop from a burette, with frequent shaking, until the solution is just homogeneous again at 25°; the whole is then weighed again. Thus the total quantity of liquid and of fluoride is known; and by reference to the following table (or a curve plotted from it) the quantity of alcohol in the homogeneous liquid, and therefore in the original solution, can be calculated.

The method is not applicable to solutions, such as beer, containing a large quantity of dissolved solids. Moreover it will, of course, be understood that the process is not given as one for use in the ordinary estimations of alcohol, but as one based upon a different principle from those of the usual methods, and of theoretical interest. A modification of the method, whereby it can be used for estimating alcohol in tinctures, etc., has been described by Haines and Marden.2 Weighing is dispensed with, the volume of the separated alcohol being read off in centrifuge tubes. These are of 15 c.c. capacity, graduated to 0.1 c.c, and carefully calibrated so that readings can be estimated to 0.01 c.c. Ten c.c. of the sample are taken, unless the alcoholic strength is 50 per cent. or more, retained in the fluoride layer, an addition of 015 c.c. should be made to the observed reading, taken from the bottom of the meniscus. With careful readings fairly accurate results can be obtained, namely, to within about ±04 per cent. A solution containing 1 per cent. of alcohol just shows the presence of the latter by this method. Liquids such as acetone, essential oils, etc., must be absent; and if the addition of the fluoride produces a precipitate, the sample should be distilled before applying the test. Whisky, however, can be analysed without distillation if it is first clarified with alumina cream, in which case 5 c.c. are diluted with water to 10 c.c. in the tube. Dry potassium fluoride is then added until the volume in the tube reads about 13 c.c, and a small crystal of malachite green is dropped in to colour the alcohol layer. The tube is then closed with a well.fitting stopper, shaken vigorously for two minutes, and centri. fuged for two or three minutes. The alcohol separates in an upper layer. Its volume is then read off. If its temperature is not 15.6°, a correction can be applied, based on the consideration that the variation in volume per 1° is approximately 0.001 c.c. for each c.c. of alcohol. Further, since traces of alcohol are still

1 Eighth Int. Cong. App. Chem., 1912; J. Phys. Chem., 1913, 17, 402.

2 J. Ind. Eng. Chem., 1917, 9, 1126.

Ternary system water, alcohol, and potassium fluoride. (frank. Grams per 100 c.c. of solvent.

MM. Duboux and Dutoit1 consider that the determination of the critical temperature of solution as described below affords a more rapid and sensitive method of estimating alcohol than the determination of the specific gravity. The liquids found to give the best results as solvents were: (A), a mixture of 5 volumes of aniline with 3 volumes of 95 per cent. alcohol; and (B), a mixture of 1 volume of nitrobenzene with 9 volumes of 95 per cent. alcohol. On mixing 15 c.c. of liquid A with 10 c.c. of an aqueous solution of alcohol, and heating the mixture, the critical temperature of solution varied proportionately with the amount of alcohol in the aqueous solution, at the rate of 2.35° for each 1 per cent. Similarly, in the case of liquid B, each 1 per cent. of alcohol caused the critical temperature of solution to vary by 12°.

The apparatus required consists of a test-tube 3 5 cm. in diameter and about 15 cm. long. closed by a cork, through which is passed a thermometer graduated in tenths of a degree, and a glass stirring-rod (bent to a circle at the end) which can move freely through another opening. Fifteen c.c. of the liquid A and 10 c.c. of the alcoholic distillate (e.g., from wine) are gently heated and stirred in the tube until the turbidity suddenly disappears. On now cooling the liquid the turbidity should reappear at the same temperature, and this is noted as the critical temperature of solution. The result may then be checked with liquid B in the place of A.

To establish the relationship between the critical temperature of solution and the proportion of alcohol, three solutions are prepared, containing approximately 8, 10, and 12 per cent. of alcohol, the exact quantities being found by determinations of the specific gravities. Then, by plotting a curve in which the percentages of alcohol form the ordinates and the respective critical temperatures of solution the abscissae, the percentage of alcohol corresponding with a given critical temperature of solution may be found.

Applied to the distillates from wines, the method gave results rather higher (0 1 per cent.) than those obtained from the specific gravity. This is attributed to the presence of traces of substances other than alcohol and water, and affecting the critical temperature more than the specific gravity.