curve. It is therefore immaterial whether the time and temperature intervals are the same for each distillate fraction, and this is borne out by the excellent repeatability of our experiments. In Dr. Hickman’s opinion it is essential to maintain the distillation rate constant, and he proposes that, if it were practicable, this should be done by progressively reducing the distilland pumping rate after the removal of each distillate fraction so &s to compensate for the decreasing quantity of distilland remaining in the still, and thus keep the time taken to collect each fraction constant. This procedure, however, is not to be recommended, even if it were practicable, for it assumes that the measured distillation temperature is independent of the distilland pumping rate, whereas our experimental evidence shows that, for a given distillate yield, increase in feed rate is a,cconipanied by a slight increase in temperature. For this reason, in our standard procedure we use a feed rate of approximately 500 grams per hour which is kept unchanged throughout t5e dktillation. Because of the impracticability of changing the distilland pumping rate during a disti!lation, Dr. Hickman suggwts that rule of thumb adjustments should be made t o “bring the data to constant time and increment values.” No details are given concerning these adjustments, but they presumably involve the application of some empirical
rule such as, for instance, increase in temperature of 10’ C. doubles distillation rate. The data presented in Table I11 of the paper are then recalculated on the basis that the distillation rate is constant throughout the distillation and that it is equal to the rate corresponding to the middle fraction, The temperature a t the beginning of the heavy lubricating oil distillation is thereby raised by 20’ C., while the temperature for the last fraction is lowered by a similar amount. Adjustment of the data in this way is purely arbitrary, however, and it is equally possible, for example, to recalculate the data on the basis that the distillation rate is the same as for the first fraction. In this event the temperature corresponding to the first fraction is the same, but with increase in distillate yield there is increasing discrepancy between the two sets of data until, a t the end of the distillation, the recalculated temperature is about 40” C. lower than the experimental value. With regard to our comments on the elimination curve technique, it was not our intention to belittle the importance of this work. Indeed, as we ourselves said in the paper, these curves have proved to be extremely valuable for comparatke purposes. Nevertheless, it is equally true that the presentation of molecular distillation data in terms of true boiling points represents a great advance on the elimination curve technique.
Finally, we find it difficult to understand the logic of Dr. Hickman’s concluding remarks. We cannot perceive any merit in adopting a procedure which allocates to the heaviest fractions boiling points which, from the molecular weights of the fractions, are known to be low. In fact the opposite might almost be true-that the value of the method increases as the boiling range covered extends upwards. Also Dr. Hickman does not mention that while the modified procedure leads to the heaviest fractions having lower boiling points than were indicated in the paper, by the same token, the lowest fractions have higher boiling points. One of the main objectives of our work was to produce a boiling point correlation for lubricating oils from experimental molecular distillation data, and this aim was achieved using the procedure described. Modifying the procedure in the manner prescribed could only lead to poorer correlation of the data. LITERATURE CITED
(1) Jones, W. C., American Vacuum
Society, Third National Symposium on Vacuum Technology, Transactions, Pergamon Press, p. 161 (19.56). (2) Langmuir, I.,Phys. Rev. 2, 329 (1913).
W. LL. THOMAS Analytical Branch, BP Research Centre The British Petroleum Co., Ltd. Chertsey Rd., Sunbury-on-Thames Middlesex, England
Determination of Trace Quantities of Alcohols by Ultraviolet Spectrophotometry of AI kyl Nitrobe nzoates SIR: The rapidly increasing production and use of petrochemicals have resulted in a need for sensitive methods of analysis for impurities that will, for example, poison catalysts, inhibit polymerizations, initiate undesirable polymerizations, or cause chain termination. Alcohols constitute a class of organic compounds which fall in this category even when present in trace concentrations. The procedure of Johnson and Critchfield (1) based on the esterification of alcohols with 3,5-dinitrobenzoyl chloride followed by a color development in the presence of base is an improvement over previously published methods. Preliminary experiments with this procedure indicated that the ester formation step was very effective. However, separation of the reactants from the products and the formation of colored products proved difficult. These experiments suggested that perhaps the esterification step might be used to 1152
ANALYTICAL CHEMISTRY
determine alcohols by measurement of the ultraviolet absorption of the alkyl nitrobenzoate products. This would avoid the unstable color formation in the ,Johnson-Critchfield method ( I ) . The new procedure is based on the esterification of alcohols with p-nitrobenzoyl chloride. The esters are separated from the excess reagent and measured by ultraviolet absorption. Primary and secondary alcohols through Clz are determined by this procedure. Higher alcohols were not tested. EXPERIMENTAL
Reagents. Pyridine, Fisher Spectro Grade, freshly distilled. p-Nitrobenzoyl chloride, Eastman Kodak P 499, solution prepared by dissolving 1 gram of reagent in 25 ml. of pyridine. Prepare fresh daily. Cyclohexane. Spectro Grade. Procedure. Transfer 1 ml. of liquid hydrocarbon sample to a 125-
ml. separatory funnel (Teflon stopcock) and add 2 ml. of pyridine. (Scrub gaseous hydrocarbons through a known volume of pyridine in a small gas washing bottle a t a rate of approximately 0.1 cubic foot per 15 minutes. Transfer a 2-ml. aliquot of the pyridine solution to a separatory funnel.) Then add 1 ml. of p-nitrobenzoyl chloride solution to the separatory funnel and mix by shaking. Allow the mixture to react at room temperature for 30 minutes. Add 25 ml. of cyclohexane to the separatory funnel, mix well, and wash the mixture with 10 ml. of 2M potassium hydroxide. Allow the phases to separate and discard the lower aqueous phase. Wash the cyclohexane with two 10-ml. portions of ‘LJI hydrochloric acid and follow this with two 10-ml. alkali washes. Finally wash the material with 10 ml. of 251 hydrochloric acid and allow the phases t o separate. hfeasure the absorbance of the cyclohexane phase in a 1-cm. cell a t 253 mp us. a blank treated in the same manner. Determine the alcohol concentration
b Figure 2.
Removal of pyridine by HCI wash
A.
Original ethyl-p-nitrobenzaate solution B. Pyridine added to original solution C. 1st HCI wash D. 3rd HCI wash E. Cyclohexane-pyridinesolution after 3rd HCI wash (blank) I
I
Figure 1. A. B.
I
I
4
,
I
I
I
I
1
1
,
Ultrav,ioletspectra
Ethyl-3,5-dinitrol>enzoate Ethyl-p-nitrobenzoate
from a previously prepared calibration curve. Calibration. The calibration curve is prepared by treating 1-ml. aliquots of cyclohexane-alcohol blends containing from 25 to 300 pg. of alcohol per ml. of solution in the same manner as given above. The total absorbance values are plotted us pg. of alcohol. Beer’s Law is followed for concentrations up to 300 pg. of a,lcohol per 25-ml. of solution. RESULTS
Table I shows thc results of the analysis of synthetic blends of ethanol in cyclohexane. In the concentration range of 25 to 200 p.3.m. the average Samples conrecovery is 100.4’%. taining down to 5 p.p.m. alcohol have been analyzed with good accuracy. The applicability of the method for higher alcohols is shown in Table 11. The blends were prerlared from commercial alcohols without further purification. Results given for these blends are based on the ethanol calibration. Therefore, errors are magnified by the alcohol-ethanol molecular weight ratio. The applicability of the method for gaseous hydrocarbon samples was determined with two blench of ethanol in propane and propylene. Analyses obtained on the blends immediately after preparation gave an average recovery of 103%. After aging for several days, both blends gave valuw which differed greatly from the original. Because of this difficulty, data on the recovery of alcohol from gaseoL s hydrocarbon samples were obtained by adding known amounts of alcohol to the pyridine scrubbers and then passing gaseous hydrocarbon through the scrubbers. For each case two scrubbers in series were used and 0.3 cubic foot of hydrocarbon (propane or propylene) was scrubbed. Aliquots of pyridine from each scrubber were analyzed for alcohol.
These experiments indicate quantitative retention of alcohol by pyridine scrubbing. An average recovery of 97% was obtained. In every case, more than 95% of the recovered alcohol was found in the first scrubbers. The experiments suggest that gaseous hydrocarbon streams should be analyzed at the source and not collected in a metal pressure vessel for subsequent analysis. DISCUSSION
Although alcohols and p-nitrobenzoyl chloride react rapidly in pyridine solution to form alkyl p-nitrobenzoates, it was necessary to study a number of reaction variables during the development of this method. Selection of Reactants. Initial studies were carried out using 3,5dinitrobenzoyl chloride as the esterification reagent. Ultraviolet absorbing species interfere a t the wave-
Table 1.
Analysis of Ethanol-Cyclohexane Blends
Ethanol, p.p.m. Recovered
Added 34.0 78.0 131.8 196.2 206.1 212.2 25.2
35.8, 34.5, 36.4, 36.0, 39.5, 39.8 73.5. 77.5. 74.0. 78.5. 75.0. 78.0. 75.0 124.2, i28‘.0, 131.8, 126.8, i26.8,’124.2 191.0, 196.7, 192.8, 197.0, 190.3, 197.0 210.0, 199.6, 205.3, 205.3, 205.3, 211.3 213.5, 215.8, 212.5, 210.0, 211.2, 211.2 26.3, 26.0, 2 5 . 5
Table II.
Alcohol Butanol- 1 3-Methylbutanol-1 Hexanol-1 Cyclohexanol Butanol-2 Propanol-2 Heptanol-1 Octanol-1 Dodecanol a
length of maximum absorption for ethyl 3,5-dinitrobenzoate, 220 mp. For this reason other nitro derivatives of benzoyl chloride were studied. pNitrobenzoyl chloride was selected because of its ready availability, high reactivity, and longer wavelength of maximum absorption of its esters, 253 mp (Figure 1). Removal of Excess Pyridine and p-Nitrobenzoyl Chloride. Pyridine and p - nitrobenzoyl chloride have ultraviolet absorption bands in the same region as alkyl p-nitrobenzoates and must be removed from the cyclohexane phase prior t o measurement. Blends of pyridine-cyclohexane-ethylp-nitrobenzoate and p-nitrobenzoyl chloride-cyclohexane-ethyl - p - nitrobenzoate were prepared to determine if pyridine and p-nitrobenzoyl chloride could be preferentially extracted from the mixtures. Three extractions with 10-ml. portions of 2M HC1 completely
Analysis
Av.
Std. De
