None of the methods for estimating higher alcohols gives results which are strictly and scientifically accurate. That is to say, none determines the exact quantity of each separate alcohol - or even of the total alcohols with certainty.

In the colorimetric processes, for example, the results are obtained in terms either of a single constituent (isobutyl or amyl alcohol) or of a standard mixture. Since the higher alcohols severally produce different intensities of colour with sulphuric acid, it follows that the results in either case are of the nature of an approximation. This objection applies, however, with less force when a mixture of higher alcohols is taken for the standard than when a single alcohol is used as the criterion.

Furthermore, some doubts have been thrown upon the trustworthiness of the standard alcohols themselves - namely, as to whether the higher alcohols, if absolutely pure, do in fact yield colorations with sulphuric acid. It has been suggested that the colorations obtained may really be due to associated and difficultly-removable impurities such as terpenes. Definite proof of this, however, is not forthcoming.

On the whole, perhaps the most precise method for the determination of higher alcohols in spirits is the Dupre-Marquardt process as improved by Allen, Schidrowitz, and others, and now generally known as the "Allen-Marquardt " method. It is, however, somewhat lengthy and troublesome, and, moreover, it cannot claim to be scientifically accurate in the sense defined above, since no determination of the separate alcohols is obtained. All are expressed in terms of amyl alcohol. Further, there is at the outset the objection that if secondary alcohols such as isopropyl alcohol or secondary butyl alcohol are present they are not, in theory, properly taken account of. Isopropyl alcohol on oxidation is converted into acetone, and would not be shown at all in the result. Secondary butyl alcohol, if oxidised beyond the ketone stage, yields eventually two molecules of acetic acid for each molecule of the alcohol, and thus would give an unduly high result. True, if both these alcohols occur together the errors tend to neutralise each other, but there is yet a further objection. It has been shown by direct experiment upon known quantities of the individual higher alcohols that whilst good results are obtainable in the case of amyl alcohol, the butyl and propyl alcohols are much under-estimated by the process. Thus Bedford and Jenks,1 in testing the Allen-Marquardt process, obtained results of which the following is a summary: -

Alcohol.

Found. Percentage of quantity added.

Amyl ..........................................................

93

to

110

n-Butyl .........................................................

30

"

70

Isobutyl ........................................................

26

"

52

n-Propyl .......................................................

13

"

31

A part of the losses was found to be due to the fact that when the carbon tetrachloride extract of the higher alcohols is washed with solution of sodium chloride and sodium sulphate in order to free it from ethyl alcohol, some portion of the intermediate alcohols is removed at the same time. This, indeed, is only what would be expected. If amyl alcohol were practically the only higher alcohol present, the Allen-Marquardt procedure would give excellent results.

Much the same remarks apply to the Beckmann method as to the Allen-Marquardt process, in respect of the loss in washing and the expression of the results in terms of one unit.

The nearest approach in principle to an accurate method of analysis is that due to Bardy, inasmuch as the propyl alcohol is determined separately, and it would be possible to estimate the respective proportions of amyl and butyl alcohols, if these were the only others present, by converting them into the corresponding mixed acids and finding the combining weight. The process appears, however, to be devised for the estimation of larger quantities of higher alcohols than are usually present in potable spirits, and it is undeniably lengthy and troublesome.

As regards the Rose process, it suffers in principle from the fact that the various alcohols produce different amounts of increase in the volume of the chloroform, one part of amyl alcohol being equivalent to nearly six parts of propyl alcohol in this respect. It requires very careful attention to details of temperature and manipulation. According to W. L. Dudley,2 the results are much higher than those obtained by the Allen-Marquardt process, though a marked relationship is observable between the results given by the two methods. This author gives an account of the precautions necessary in using the Rose process. He considers the Allen-Marquardt method the better, as it is more rapid, has fewer sources of error, and gives the amyl alcohol content more accurately.

Whilst in English-speaking countries the Allen-Marquardt method has become the standard process, it has not been adopted universally. On the Continent, colorimetric and other methods are used, and they offer so much saving of time that they will probably remain long in favour, especially as data based on them have been accumulated which serve as standards - as, for example, in the analysis of brandy. Hence it has been necessary to describe the chief of these at some length. There is all the more reason for this since the Allen-Marquardt process, as already shown, is open to criticism in some respects, even though it may be accepted as on the whole the nearest approach to a satisfactory method. The colorimetric processes, however, should be used with discretion, and the results should not be regarded as strictly comparable with those given by the Allen-Marquardt process. A number of comparative figures obtained by this process and by colorimetric methods have been published, some of which show relatively enormous differences in the results, the colorimetric values being generally the higher. Without inquiring into the probable causes of these larger discrepancies, it may be remarked that the colorimetric method as used in the Government Laboratory has been found to give fair approximations to those obtained on the same samples by the Allen-Marquardt process, and in only a fraction of the time required for the latter (see the figures given below). Probably this is due to the facts that in the Government Laboratory process, unlike some other modifications of the colorimetric method, the action of the sulphuric acid is effected in the cold, and the standard of reference is a mixture of higher alcohols, not a single alcohol.

1 J. Soc. Chem. Ind., 1907, 26, 123.

2 J. Amer. Chem. Soc, 1908, 30, 1271.

Where a large number of samples have to be examined, the analyst will probably find it advantageous to use one of the colorimetric processes in the first instance, employing the Allen-Marquardt method for confirmatory purposes or for special samples if desired. Wherever necessary, the method of estimation should be stated in giving the results. For brandy, if the French data are to be used for comparison, the process due to Girard and Cuniasse is obviously desirable, either with isobutyl alcohol as standard, or with amyl alcohol, and the results calculated in terms of isobutyl alcohol.

The kind of agreement which may be expected between results obtained by three of the chief methods will be seen from the following table.1

1 Appendix Q, Royal Commission on Whisky.

Higher alcohols, estimated by different methods. (Parts per 100,000 of absolute alcohol.)

Allen. Marquardt.

Beckmann.

Govt. Laboratory.

250.9

320.3

228.0

221.3

304.3

245.6

310.6

278.7

350.8

313.1

276.3

266.6

260.0

288.9

317.5

291.4

288.6

317.5

It is scarcely necessary to say that very close agreement is not to be expected, in view of the admitted difficulties attending such determinations.