248 TRORPE: THE INTERDEPENDENCE OF THE PHYSICAL AND XXX1.- The Ch,ernicu I Interdependence of the Physical and Cderia in the Analysis of Butter- fat. By THOMAS EDWARD THORPE, C.B., F.R.S. IN July, 1901, a Departmental Committee was appointed, at the instance of the Board of Agriculture, to inquire and report as to what regulations, if any, might with advantage be made for determining what deficiency in any of the normal constituents of butter should raise a presumption that the butter was not genuine. To assist the Committee in arriving at a conclusion on this matter, it became necessary to obtain at first hand the values of various analytical constants of specimens of butter of known origin, and produced from milk given under varying conditions. Observation has shown that the chemical nature of butter-fat is dependent t o a certain extent on the climatic influences to which the cows are exposed, on the nature and amount of the food supplied, and on the breed, period of lactation, and idiosyncrasy of the individual cow.In order to give such weight as was practicable to the effect of these factors, the samples of butter were obtained from carefully selected districts, and often from cows set apart for the purpose of the inquiry ; whilst particulars of the breed, diet, stabling, and period of lactation were supplied with the samples in nearly all cases. For example, to illustrate the effects of more rigorous climatic conditions than obtain in the United Kingdom generally, farms and dairies in Caithness, Sutherland, the Orkneys, Shetlands, and the Hebrides were laid under contribution ; whilst a series of samples from Hollesley Bay, Suffolk, served to exemplify the influence of exposure of the cattle to the winds of the North Sea.From the arrangements made for the collection and transmission of the butter, it is believed that all the samples can be relied on as authentic. The analyses were made at the Government Laboratory. Altogether about 430 samples were received ; of these, for special reasons, 73 samples were only examined as regards water, salt, curd, and amount of fat, with which data we are not here concerned. The analytical figures obtained for the remainder, 357 in all, included, besides the foregoing particulars, determinations of the Reichert- Wollny number of the fat, its relative density and saponification value, and the Zeiss reading at 45'.I n nearly all cases, the proportions of soluble and insoluble fatty acids were also determined, together with the mean molecular weight of the insoluble acids. Details concerning the origin of the samples are published in theCHEMICAL CRITERIA IN THE ANALYSIS OF BUTTER-FAT. 249 Minutes of Evidence to the Report of the Committee (Cd. 1750, Appendix xxix, 505-58s). As relatively few systematic observations appear to have been made on authentic samples of British and Irish made butter having the representative character of those now dealt with, it may be useful to put on record a summary of the results obtained. The object of the present 'communication is to indicate the general relations between the analytically-important chemical and physical constants of butter-fat as disclosed by a study of the average values of these constants determined on a number of samples sufficiently large to ensure with reasonable certainty n fairly close approximation to the true values.Before attempting to make use of the observations with a view to general deductions, it is desirable to obtain some idea of the magnitude of the error to which the several determinations are liable. Taking first the Reichert-Wollny experiments, the results were obtained by working strictly under the conditions agreed upon by myself and a Committee representing the Council of the Society of Public Analysts, who in 1900 examined the matter in conjunction with analysts of the Goverment Laboratory in connection with the question of butter-fat in margarine (see Abstr., 1901, ii, 77).I n a large number of duplicate experiments the mean difference obtained between two determinations of the Reichert-Wollny number on the same sample of butter-fat was 0.20. So far, therefore, as what may be called fortuitous errors are concerned, the average values may be taken as known to within about k0.1 unit where the value is obtained from a fairly large number of samples, and probably to within +_O-2 unit where the number is relatively small. The relative densities of the fats were generally taken in a pycnometer of bottle form, adjusted to contain 50 grams of water a t 37.8'. . Duplicate weighings of the same fat with such a pycnometer rarely differ by more than 0-005 gram, corresponding with one unit in the fourth decimal place when expressed as specific gravity.I n a few cases a bottle of only half the above capacity was used ; the error here might be twice as great as with the larger pycnometer, or two units in the fourth decimal place. Consequently the average values may, as regards the variable errors of experiment, be looked upon as accurate to within & 0.0001. A small constant erroi- might conceivably arise thus : after the fat in the bottle has been brought to the desired temperature of 37.8' a little time is required to complete the filling and to insert the stopper. During this time, the temperature is falling, and the effect of this would be to slightlyincrease the weight of fat which the bottle would contain.The same tendency would, however, be found during250 THORPE: THE INTERDEPENDENCE OF THE PHYSICAL AND the adjustment of the instrument with water, and the actual error would therefore be only that arising from the difference of these effects. The average difference shown by duplicate determibations of the saponificatibn value of butter-fat, deduced from a large number of experiments made by several operators, was found to be 0.065, expressed as percentage of potassium hydroxide. This corresponds to about 0.7 of a unit when calculated into ‘‘ saponification-equivalents.’’ Sub- stantially the same difference was found in the case of the molecular weight determinations. The values may, therefore, be regarded as subject to an experimental error of about k0.4 of a unit.I n all cases the acid and alkali used were carefully standardised, and it is not thought that any appreciable error is to be attributed to this part of the experimental work. As regards the refractometer indications, it may be noted that there are two or three possible sources of inaccuracy in using the Zeiss butyro-refractometer. (1) Since the instrument is graduated into divisions representing whole units, the fractions of a unit have to be estimated by the eye. (2) The line of shadow is often not very sharply defined where i t cuts the scale of the instrument. (3) The temperature may fluctuate slightly during the course of the experi- ment. Probably a single observation might be liable t o a maximum error of & 0.4 unit, and the mean of a number of experiments‘ such as are used in the curves given below may be relied upon as accurate within about k 0 .2 of a unit. For the soluble and insoluble acids, the method adopted was sub- stantially the modification of Hehner’s process recommended by the American Association of Official Agricultural Chemists. The saponi- fication of the fat, however, was effected in silver flasks instead of in glass. For the soluble acids, the mean difference between duplicate analyses in a large number of cases was found to be 0.12 per cent., and for the insoluble acids 0.15 per cent. So far as manipulative error is concerned the average values given in the table may therefore be taken as accurate to within about f 0.06 and f 0.08 per cent. in the two cases respectively. The averages of the analytical data from which the curves given below are constructed are given on p.254. I n obtaining these figures -taking, for example, the first line-all the Reichert-Wollny values lying between 22-00 and 22.99 have been taken from the experimental results and averaged, giving the number 22.5. The corresponding values of the specific gravity, sapouification-equivalent, and Zeiss number were also abstracted and averaged, giving the figures shown in the table. For the next line, the Reichert-Wollny values lying between 23.0 and 23-99 were taken; and 80 on for the others. It wouId be small, and sensibly constant.CHEMICAL CRITERIA IN THE ANALYSIS OF BUTTER-FAT. 251252 THORPE: THE INTERDEPENDENCE OF THE PHYSICAL AND 0 0 I - c1CHEMICAL CRITERIA IN THE ANALYSIS OF BUTTER-FAT.253 Li VOL. LXXXV. S254 THORPE: THE INTERDEPENDEKCE OF THE PHYSICAL AND Specific gravity at 37.8" 37.2. I Saponifica- tion equivalent. No. of samples' Reichert- Wollny number. 22.5 23'5 24'5 25.5 26 5 27'5 28'8 29 -5 30.5 51 '3 32'6 450. Per cent. on fat. _____ 0'91C1 0'9104 0-9108 0.9110 0'9113 0.9114 0'9118 0'9120 0'9123 0'9125 0,9130 ~. 7 17 15 27 37 51 78 56 41 18 255.4 253'4 251.3 251.1 248.9 247'4 245'7 244'0 242 -4 241.5 241.2 _ _ I In- soluble acids. Per cent. on fat. I---- 42.0 I 4'3 90.1 41'5 1 4.5 89.7 41.5 4.7 ' 89.4 41'3 41'0 40.6 40'1 40.1 39.9 39 .7 39 '4 4 .8 4'9 5.2 5.4 5.6 5.8 5.7 6'0 89'3 88.9 88.7 88'4 88.3 87'9 87-9 87 -7 Mean molecular weight of insoluble acids. - 266.9 265-5 865.0 264'2 261'9 261.7 260-9 259% 260.1 258.0 257'8 * Calculated as butyric acid.Two or three samples, the data for which would fall beyond the extremities of the curves given, have been omitted on account of the number of observations being insufficient to afford trustworthy average values. As regards the probable errors of the observations, it may be remarked that any constant error pertaining to a process will not, when the results are presented in the form of curves, affect the shape of the curve; i t will only alter its position relative to the axes. Such errors are those which, for example, might arise from the thermo- meter of a Zeiss instrument having altered since the refractometer was standardised, or from the retention in the distilling-flask of a small constant quantity of volatile acids in determining the Reichert- Wollny figure.On the other hand, errors of manipulation and observation, the incidence of which is sometimes on one side of the truth, sometimes on the other, will, in accordance with statistical principles, be largely nullified by taking the averages of a sufficiently large number of experiments. Curves plotted, therefore, from the mean results of many observations should show with some degree of accuracy the general tendency of the relations between the several constants, and illustrate their interdependence. It will be seen that, in a general sense, the relative density of butter-fat increases as the Reichert-Wollny number is augmented. This would, of course, folIow from the well-known fact that the glycerides of low molecular weight have a greater density than the glycerides of the higher fatty acids which occur in butter.Very approximately, the increase is A few comments may be made on these curves.CHEMICAL CRITERIA IN THE ANALYSIS OF BUTTER-FAT. 255 directly proportional to the increase in the Reichert-Wolhy values, as is shown by the curve approaching more or less closely to a straight line, Speaking broadly, the variations of the saponification numbers are in inverse relation to those of the Reichert-Wollny values and the relative densities; but they show, as from the nature of the case might be expected, some notable departures from proportionality in the amount of variation. The aberrations aro most marked towards the extremities of the curve. Here the observations are fewar in number than those from which the middle of the curve is constructed, but it is not likely that any presumable experimental error would account for the variations shown.When it is considered that the saponification-equivalents of the glycerides of the volatile fatty acids range from 100.6 for tributyrin to 212-6 for trilaurin,* whilst those of the glycerides of the non-volatile acids vary between 240.7 for trimyristin and 296.6 for tristearin, it will be obvious that even small variations in the relative proportions of the glycerides may affect rather considerably the mean saponification-equivalont of the whole fat. Somewhat similar remarks apply to the Zeiss numbers These generally decrease in magnitude as the Reichert-Wollny values in- crease, but the rate of diminution is not regular, and this points, like the variations of the saponification-equivalent, to changes in the proportions of the individual acids which, in butter-fat, constitute the ‘‘ volatile ” and ‘‘ non-volatile ” groups respectively.The curve showing the mean molecular weights of the insoluble acids is of some interest, from the physiological problem whichit suggests. From it we infer that, as the acids of low molecular weight increase in amount, the mean molecular weight of the remaining acids decreasep, and, broadly speaking, decreases proportionally. This may indicate one of several things. Remembering that the molecular weights of the insoluble acids which are believed to occur in butter are as follows : lauric (200),? myristic (228), palmitic (256), oleic (282), stearic (284),f and that the mean molecular weights dealt with in the table decrease from 266.9 to 257.8, it will be seen that this decrease may be due to one of the following causes.In the first place, i t may be due to a diminished proportion of oleic acid. Or, the oleic acid remaining the same, the decreased molecular weight may be due to a displacement of palmitic acid by myristic and lauric acids, one or both ; or, possibly, of myristic by lauric acid. Finally, if the oleic acid increases, this can * Or to 184.7 for tricaprin. t Lauric acid is said to be slightly soluble in water and appreciably volatile when $ Apparently stearic acid exists in butter to a small extent only. If this is so, it distilled with steam.may be neglected in the present connection. s 2256 ANALYSIS OF BUTTER-FAT. only be compensated for, and the decreased mean molecular weight explained, by a n increase in the amount of lauric acid, with a simul- taneous decrease in the myr istic and palmitic acids. The physiological question which presents itself is whether the metabolic changes which, in the bovine organism, result in the production of an increased proportion of the lower saturated fatty acids, bring about this produc- tion a t the expense of the oleic acid. Judging from the values obtained for the iodine-absorption of a number of the specimens, it would seem that, speaking generally, a low proportion of volatile acids in butter-fat is associated with a high percentage of oleic acid, and vice ves-sd.The figures are as follows : Mean molecular Reichert- weight of Wollny Iodine - Oleic Insolnlde in so 1 uble number. value. I - acid. acids. aceicls. (1) Average of 20 samples 24.2 40.0 44.4 per cent. 89% per cent. 264'6 (2) 30 ,, 30'8 32.4 36.0 ,, 88-1 ,, 259.8 I n the first series, the oleic avid cor~stitutes 4'3.6 per cent. of the insoluble acids; in the second series, 40.9 per cent. This decrease in the proportion of oleic acid would diminish the mean molecular weight of the insoluble acids from 264.6 to 261.5. The actual figure obtained for the molecular weight in the second series being 259.S, it would appear that the decrease is l;irgely, but not eutirely, brought about by the diminution in the proportion of oleic acid. Since the whole of the decrease is not accounted for by the oleiu acid, it follows t h a t the mean molecular weight of the insoluble saturated acids is lower in the second series than in the first. I n fact the values, calculated from the foregoing table, are 247.4 and 244.4 respectively. It should be mentioned that the iodine-values were determined after the fats had been kept some time, and they are not given as repre- senting precisely the values for the fresh fat. It is well known t h a t the oxidation changes which may take place during the keeping of butter-fat tend to affect the iodine number. Neverthelesp, since the two groups cover practically the same dates and were kept side by side until examined, it may be assumed that, on the whole, they would be about equally affected, and the relative values of the averages would remain substantially unchanged. THE GOVERNMENT LABORATORY, LONDON.