I N D U S T R I A L A N D ENGINEERING CHEMISTRY
All of the acetone, however, was not volatilized from the one sample of bread which was baked, for the characteristic odor of acetone was strongly evident. Stability
Several preliminary experiments have been conducted to determine the stability of acetonedicarboxylic acid baking powder a t various temperatures. According to the literature acetonedicarboxylic acid decomposes in a few hours a t room temperatures. However, in the course of a study (3) of the kinetics of the decomposition of acetonedicarboxylic acid, weighed quantities of the acid that were allowed to stand in a desiccator over phosphorus pentoxide from March to October showed no decomposition. Several samples of acetonedicarboxylic acid baking powder were kept in stoppered glass bottles a t various temperatures for some time. The available carbon dioxide was then determined by boiling a 1.5- to 2.0-gram sample with water and absorbing the gas in potash solution. The apparatus used was essentially the same as that used in determining the available carbon dioxide in a baking powder. The data obtained are given in Table I. The carbon dioxide a t the beginning of the experiment was calculated from the percentage of acid mixed with the starch, except in samples 1 and 2, in which the carbon dioxide was actually determined. Table I-Available
Carbon Dioxide i n Acetonedicarboxylic Acid Baking Powders
1 SAMPLB TIME TEMPER.4TURE
C. 32.2“ 23.96
2i.9 23.9 0
AVAILABLE CARBON DIOXIDE Calcd,
Calcd. to 155; availatAe COz powder
F. 90 75
Per cent 14.00
Per cent 13.69
Per cent 11.36
Per cent 12.17
I ;: I 75 32
a refrigerator for more than 2 months, no carbon dioxide being obtained. The results have been calculated to a 15 per cent available carbon dioxide powder as shown in the last column in Table I. The data indicate that mixtures made up to give 15 per cent carbon dioxide would contain more than 12 per cent (the legs1 minimum) after standing a t room temperatures for 1 year. Samples 3 and 4 have been kept a t room temperature for 17 and 142/3 months, respectively. The percentage of available carbon dioxide was then determined by titrating the undecomposed acetonedicarboxylic acid with tenth-normal alkali. Sample 3, which originslly contained 13 per cent available carbon dioxide, was found to have 9.2 per cent a t the end of 17 months, while sample 4, which originally had 15 per cent carbon dioxide, still contained 11.3 per cent a t the end of 142/3 months. Calculations from both samples indicate that a powder containing 15 per cent available carbon dioxide a t the start would contain 12 per cent a t the end of 12 months. Conclusion
Most baking powders are “double-acting;” that is, a part of the carbon dioxide is liberated on mixing the batter and the remainder on baking. Acetonedicarboxylic acid decomposes only slightly in the batter unless some of the ingredients are heated. This can be overcome by adding a small amount of sodium bicarbonate, sodium carbonate, ammonium carbonate, or other slightly alkaline salt which, in the kinetic study previously referred* to, were found to catalyze the decomposition of acetonedicarboxylic acid. However, this would leave the catalyst in the baked product, which it is desired to avoid. Furthermore, the acetonedicarboxylic acid baking powder raises cake sufficiently without a catalyst. An acetonedicarboxylic acid baking powder might be manufactured to compete with the more expensive baking powders on the market, especially if cheaper citric acid is made available. Such a powder would have the advantage of leaving nothing in the baked product. The study of the suitability of acetonedicarboxylic acid as a leavening agent is being continued with the aid of the Wisconsin Alumni Research Foundation,
Vol. 21, No. 11
In a refrigerator the acid does not decompose a t all or a t least only very slowly. This was shown by blowing nitrogen through bottles of the baking powder that had remained in
communication. Loevenhart, (2) Marvel, “Organic Syntheses,” Vol. V, p. 6 , John Wiley & Sons, Inc., New York, 1925. (3) wiig, J. phys. Chem., 32, 961 (1928).
Sampling Cleaned Apples for Determination of Arsenical Spray Residue’ J. W. Barnes and C. W. hlurray IjUREAU OF
A N D SOILS,
u. s. DEPARTMENTOF AGRICULTURE, WASHINGTON, D. c.
H E results of an investigation of the variation of arsenical residue on individual apples as taken from the orchard conducted to find the size of sample necessary for an accurate determination of arsenical residue were published in the February, 1929, issue of INDUSTRIAL AND ENGINEERIXG CHEMISTRY.The results of analyses of apples which had been passed through apparatus designed to remove this residue, or a t least to reduce it to less than 0.01 grain of arsenic trioxide per pound of fruit, the official British toler1
Received August 17, 1929.
ance, are reported in the present paper. For apples direct from the orchard a sample of approximately 50 was found necessary, but for cleaned apples it seemed probable that a much smaller number would be adequate. Three lots of apples-one from Virginia, one from Washington, and one from Oregon-were obtained. These apples had been cleaned according to regular commercial methods in standard apparatus. Those in two lots had been washed in acid solution and those in the third lot had been wiped in a scrubber of the revolving brush type. Sixty apples were
INDUSTRIAL A N D ENGINEERING CHEMISTRY
drawn a t random from each lot and analyzed separately for arsenical residue. The fruit was peeled, the peel was oxidized with nitric acid in the presence of sulfuric acid, and the arsenic was determined by the more recent modifications of the Gutzeit method for-minute quantities of arsenic. The results are expressed in grains of arsenic trioxide per pound of fruit, in order to conform to commercial usage, Although each set constituted a separate sample, the means of these samples were nearly the same and the distribution curve's were very much alike. Figure 1 shows the distribution curves for the three lots. The numerical data for the three groups follow: LOT
P R O B A B LEER R O R OF A SINGLE OF FOURAPPLES SAMPLE t0.00098 t o , 00093 *0.00075
These values are distinctly less than the experimental error of the method. Hence the sample is adequate for the quantity of residue indicated.
Grain per pound I1 111
Grain p e r pound
0,0042 0.0032 0.0044
0.003‘ 0.002‘ 0,0034
The probable error of the mean of a series of observations bears the following relation to the number of observations and the deviations of these observations from the mean: PE = PE ‘\’
= = =
0 674.5 u
probable error of mean n.iml:er of apples in sample the roct-mean-square, or the square root of the mean of the squares of the residuals, or
B @ ..
= residua! = single observation minus
mean of entire group used X u 2 = sum of squares of these individual
When the original sample (A’) is large enough to give a determination of u really representative of the dispersion in the whole “population” from which the sample was taken and when the individual samples give a normal or symmetrical distribution curve this formula may be used to calculate the probable error occurring when a smaller sample (AT)is used. Application of the formula to these data, accepting an accuracy of *0.0015 grain per pound as the best to be expected from the Gutzeit method under the circumstances, indicated that four-apple samples would be sufficient. \Then the distribution curve is decidedly skew, as in this case, it is necessary to supplement the direct application of this formula to the observed values by “artificial sampling.” The analyses of these individuals are recorded on cards, pellets, or other convenient objects of uniform size and shape. These numbers are then shuffled and one number is drawn. This number is recorded and replaced. After mother mixing, a second number is drawn, and so on. This results in a random series of the original analyses, but of any length desired. This series is divided into groups of equal size, say of four individuals each, this being the size indicated by the original series of analyses, and the mean of each of these groups of four is calculated. This gives a second series of observations consisting of the means of a number of samples of the size to be tested. When this series is treated in the usual manner by calculating the arithmetical mean and the sum of the squares of the residuals and the standard least squares formula for the probable error of a single observation is applied, the following values result:
0, per pound Figure 1
G r a i n s Cis,
o f fruif
Although these apples are very clean, the mean being a little less than half the tolerance of 0.01 grain per pound, it is probable that the results may be safely extrapolated up to this tolerance. The standard deviation in each of these three samples was less than eight-tenths the value of the mean of the sample. Application of Chauvenet’s criterion for discarding indiGidual determinations removes the highest one in Lot I and the highest three in Lot 111. It also reduces the standard deviation to 50 per cent of the mean. The standard deviation for apples taken direct from the orchard was also found to be of the order of magnitude of one-half the mean of the sample. It seems probable that for a mean of 0.01 the standard deviation would still approximate this 50 per cent. Substituting in the formula this 0.005 gives 6 as the number sufficient to yield the required accuracy. -4sample of ten apples would allow a standard deviation of 70 per cent of the mean with equal accuracy. Conclusion
Study of the variations in arsenical residue on individual apples that have been cleaned indicates that for apples carrying not more than 0.01 grain of arsenic trioxide per pound a sample of six apples, taken a t random from a lot all parts of which have had the same treatment, will give an accuracy of *0.0015 grain per pound.