ANALYTICAL CHEMISTRY
1818
to choose any known mixture as an unknown, analyzing it by using the caliSample NssPaOla I, NaSP301o 11, NasP801o.6HrO. KarPcO:. bration curves determined from the Sample Type No. % % 7% '70 other standard mixtures. Table I1 inRIechanical mixture 10 Actual . . , . 40 40 20 cludes such a treatment for two of the Found . . . . 42.5 37 21 9 Actual 50 50 ... ... known mixtures. The results found Found 48.5 51 ... are obtained by applying the calibration Commercial triphosphate A Found 6 5 . 5 29 Sone 5 3 B Found 9 . 5 83 None i 3 curves to the intensity ratios obtained C Found 2 7 . 5 58 3 None 1" experimentally for these samples, and Hydrated triphosphate D Found None 45 47 7 are typical of those to be expected on unknowns. Some analvses of tvDical commercial sodium tripolyphosphates The calibration curves for the hexahydrate of triphosphate and and a partially hydrated sample are also given in Table 11. sodium pyrophosphate are given in Figures 5 and 6. The 14" line There are several possible sources of error in this method of (see Figure 2) is measured for the hexahydrate; the stronger 8.7' analysis. The most serious one, because it is so subtle, is the line must be avoided, because it is very sensitive to orientation possible presence of amorphous material in a sample. When effects. I n the case of the pyrophosphate, the diffraction line a t such is the case, there cannot be good agreement between results '26.4" is measured. by this method and chemical methods. This method actually A glance through Table I, the tabulation of compositions of measures the weight fraction of crystalline components, Lack standard mixtures, shows that samples of all combinations of the of reproducibility in sample preparation, chiefly due to crystal components were not prepared, This is not necessary, however, orientation, is probably the next most serious cause of error, or for as Alexander and Klug ( 1 ) point out, the matrix (that part of lack of precision in results. For euample, while the average dethe sample other than internal standard and component being viation of results of repeat runs on a single sample ranges from 0.5 measured) does not affect the ratio of intensities of the component to 1%, the average deviation of runs on a number of preparations and internal standard, since the mass absorption coefficient for of the same sample is roughly 2%) nearly as great as the average the matrix, determined by its composition, affects the intensities deviation of the calibration data as a whole. Statistical errors of diffraction lines for both the internal standard and the compoin the determination of intensities may be appreciable if intensinent by the same factor. ties are low and sufficient time is not taken to measure them. I t is intended to extend the method to include other sodium Intensity measurements can probably be made more precise by phosphates. Sodium trimetaphosphate, phase I, has been inmeasuring diffraction line areas, rather than diffraction peak vestigated briefly, and its diffraction line a t 26.2' will probably heights. be chosen for measurement. Although this line is close to the LITERATURE CITED 26.4' pyrophosphate line, the two are adequately resolved to allow intensity measurement except when the pyro content is very high rllexander, Leroy, and Klug, H. P., ANAL.CHEM.,20,886-9(1948). and the trimeta content is very low. Beatty, S. V. D., Am. Mineralogist, 39, 74 (1949). Carl, H. F., Ibid., 32, 508-17 (1947). , Gross, S. T., and Martin, D. E., ISD. ENG.CHEM.,ANAL.ED.,16, ACCURACY O F R E S U L T S
Table 11. Analysis of Unknowns
I
An idea of the accuracy of these analyses may be obtained by looking a t the consistency of the calibration data. -4s mentioned above, the average deviation of points from the calibration line for phase I1 is 3% or less. The results for phase I and pyrophosphate are better than this, while those for the hexahydrate are not so good. Another way of looking at the consistency of results is
_
95-8 (1944). (5) Thilo, Erioh, and Seemann, H a n s , 2. anorg. und allgem. Chern., 267, 65-75 (1951).
RECEIVED for review December 30, 1952. Accepted September 12, 1953. Presented before the Division of Physical and Inorganic Chemistry, Symposium on Recent Developments in Phosphorus Chemistry, a t the l 2 l s t Meeting of the A h r E R I c i l r CKEWCAL SOCIETY, Buffalo, N. Y.
Determination of Naphthalene in Wash Oil and Coke Oven Gas Infrared and Ultraviolet Spectrometry N. J. KLEIN AND G. W. STRUTHERS Polychemicals Department, E . I . du Pont de Nemours & Co., Inc., Charleston, W . Va.
I
S PROCESSING and purifying coke oven gas by wash oil scrubbing it is desirable to have rapid methods for the deter-
mination of naphthalene in both the gas and the wash oil. The method most generally applied to wash oil involves volatilization of the naphthalene with a stream of hot air into a solution of picric acid ( 2 ) , thereby precipitating the picrate. Another method ( 3 )requires distillation of the wash oil and isolation of the fraction boiling between 195" and 250' C. This fraction is then chilled and the solid naphthalene filtered and weighed. More recently ( 4 ) , Reichardt and White reported a modifica-
tion of the volatilization procedure in which the analytical manipulations were reduced to a minimum. Their method utilizes a titration of the excess picric acid after formation of the naphthalene picrate. Good accuracy and precision were obtained for the specific wash oil used. Schubert ( 5 ) reports an optical method, in which the freezing point of the sample ia compared with that of a naphthalene-wash oil standard. Kaphthalene in gas ( 1 ) is usually determined by passing the latter through a picric acid solution, filtering off the picrate, and estimating the naphthalene either gravimetrically or volu-
V O L U M E 25, NO, 1 2 , D E C E M B E R 1 9 5 3 The classical picrate methods for determining naphthalene in wash oil and coke oven gas were time-consuming and not wholly quantitative. Poor recovery was experienced in tests where known amounts of naphthalene were added to wash oil. Spectrophotometric methods were developed for determining naphthalene in wash oil by infrared, and in coke oven gas by ultraviolet, techniques. Naphthalene concentrations in coke oven gas as low as 0.001% can readily be measured, as can concentrations up to 45& in wash oil. Accuracy of the methods is 0.5Yo relative. The spectrophotometric methods are direct, simple in application, and specific, and offer a definite improvement in sensith ity, accuracy, and analysis time. No significant interferences are encountered from the more common compounds associated with naphthalene. The methods have been applied successfully to routine plant samples.
metrically. In general these methods are time-consuming and not readily adaptable to routine plant control. In this laboratory rapid methods were developed and successfully applied for determining naphthalene in wash oil by infrared absorption and in coke oven gas hj- ultraviolet techniques. The method utilizes the intensitv of the specific naphthalene absorption band a t 12.77 microns as a quantitative measure of the naphthalene content of the wash oil. Compensation is made for the slight absorbance exhibited by wash oil itself at this wave length. 2,2,4-Trimethylpentane (iso-octane) is used as the solvent for the wash oil samples, and in this medium the background absorbance of the oil is fairly linear with its concentration The amount of naphthalene in an unknown oil sample can be determined by comparing it to a standard solution of naphthalene-wash oil in iso-octane. The standard naphthalene solution is prepared to contain a weight of wash oil approximately equal t o the weight of the sample being tested. This makes it unnecessary to use an accurately calibrated cell in the analysie. The procedure is designed to cover a concentration range of 0 50 to 4.00% naphthalene in wash oil. A sample weight is selected which will place the naphthalene content of the isoortane solution between 0.005 and 0.10%. In this range the best linearity is obtained between absorbance and concentration. Apparatus. Spectrometer. Perkin-Elmer model 12-C, with sodium chloride optics. One 0.5-mm. sodium chloride cell. Reagents. Iso-octane, Phillips 66 pure grade, or comparable, free of absorption peaks between 12.50 and 13.25 microns. Sodium chloride, U.S.P. Naphthalene, C.P. Fresh wash oil, free of naphthalene absorption peaks.
1819
INFRARED MEASUREMENT. Set the slit schedule to obtain a slit width of 0.222 mm. at 12.50 microns. Adjust the gain to place the base line in the high transmission region and balance the instrument (0.0 a t 12.62 microns) by means of the lithium fluoride shutter. Make all measurements with a speed setting of 4, and a response setting of 3. Fill a 0.5-mm. cell with the dry wash oil-iso-octane solution and scan over the range of 12.50 to 13.02 microns. Remove the cell, empty, and clean with iso-octane. Refill the cell with a portion of the wash oil-naphthalene-iso-octane standard containing a weight of wash oil nearest to the sample weight taken. Scan the standard over the same spectral range of 12.50 to 13.02 microns. CALCULATIONS. Draw a straight base h e across the 12.77micron absorption band on both the sample and standard scans. Measure and record the absorbance of the band from base line to peak for each scan and calculate the naphthalene content of the sample as follows:
where A = absorbance of sample A , = absorbance of standard Urn = weight of naphthalene, grams in standard W , = weight of sample, grams Interferences. The interfering compounds which were studied were divided into tnFo classes (Table 11): compounds whose prefience might be expected in wash oil and those related structurally to naphthalene but not likely t o be present.
Table I. Approximate Size of Wash Oil Sample for Naphthalene Determination Concentration Level,
Approximate Sample Weight, Gramsa
0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00
Approximate Sample Volume, $11. 18.4 9.2 6.9 4.6 4.0 3.5 2.9 2.3
16.00 8.000 6.000 4.000 3,500 3.000 2.500 2.000
a These weights will give absorbance values of approximately 0.300 a t their respective concentration levels.
Table 11. Interferences" Interfering Compound Benzene Phenol Toluene Xylene
Maximum Canon. for Relative Weight Error, 2 % Relative Error, h l g Added, Mg. Compounds Likely to Be Present 879 536 866 43
1 .o 1 .o 0.6 1
.o
879 536b 866 43 b
Compounds S o t Likely to Be Present
...
lob
497 b 108
...
All solutions contained, in addition, about 2.800 grams of wash oil and 0.0850 gram of naphthalene diluted to a final volume of 100 ml. in iso-octane. b Any additional amount of the interfering compound would be likely t o give an error greater than 2%. C Only slightly soluble in iso-octane. a
Y4PHTHALENE IN WVASH OIL BY INFRARED ABSORPTION
Procedure. PREPAR.4TIOiV O F ST.4XDARDS. Weigh accurately 0.0800 f 0.0002 gram of pure naphthalene into each of eight 100-ml. volumetric flasks. Add to the flasks 2.000, 2.500, 3.000, 3.500, 4.000, 6.000, 8.000, and 16.000 f 0.001 grams, respectively, of naphthalene-free wash oil. Dilute each flask to volume with iso-octane and mix thoroughly. Transfer and store these standards in well-stoppered glass bottles, labeling each with the exact weight of naphthalene and wash oil added. PREP4R.4TION O F SAMPLE. Weigh 4.000 & 0.001 grams Of wash oil sample into a clean 100-ml. volumetric flask. This weight of sample is usually sufficient for wash oil containing from 1.50 to 2.0070 naphthalene. (See Table I for sample sizes of wash oil containing more or less naphthalene. The absorbance should fall between 0.090 and 0.360.) Dilute the sample to volume with iso-octane and mix thoroughly. Dry a portion of this solution by shaking with 0.5 gram of sodium chloride in a 25-ml. volumetric flask. This step is taken to protect the sodium chloride cell from any water present in the sample.
The interfering compound was added to a 100-ml. flask (containing wash oil and naphthalene) in an amount about 10 times the weight of naphthalene in the sample. The flask was diluted to volume with iso-octane and the absorbance of the solution measured in the manner already described. If more than a 2y0 relative error in the calculated absorbance was observed, another solution was prepared containing half as much of the interfering compound. This process was repeated until an absorbance error not exceeding 2% relative was obtained. I n some cases-Le., phenol, xylene, and the two alpha-substituted naphthalenes tested-it was found that when the amount listed in Table I1 under maximum concentration for 2% relative error was doubled, a three- to sevenfold increase in relative error was observed. The data in Table I1 show the magnitude of the interference of these compounds. Of the interfering compounds that might be present in the wash
ANALYTICAL CHEMISTRY
1820 Table 111. Accuracy and Precision Data for Naphthalene in Wash Oil Samples Accuracy Naphthalene, % Testa
Added
Precision Test 1 2 3 4 5 6 7 8 9 10
Average Standard deviation
95% confidence
limits
Found
Naphthalene, % 2.54 2.45 2.44 2.53 2.48 2.49 2.50 2.48 2.43 2.44 2.48 0.04
2 . 4 8 3z 0 . 0 8
Test solutions were prepared b y adding known weights of C.P. naphthalene t o a definite weight of wash oil.
oil, xylene can be tolerated least. Its absorption bands lie close to either side of the 12.77-micron naphthalene band, thus distorting the normal base line of the latter. Benzene, phenol, and toluene cause much less trouble and can be tolerated in greater concentrations. The compounds that are not likely to be present, but which are of interest, interfere considerably more. The 1-substituted naphthalenes, in particular, depress the naphthalene absorption band about equally, and consequently neither one can be present if the parent compound is to be measured. The 2-substituted isomers cause less difficulty and may occur in concentrations approximately five times that of the naphthalene. Anthracene and 2-naphthol are only slightly soluble in iso-octane and no toleration limits are given for them in Table 11. The errors noted, therefore, are for iso-octane solutions saturated with respect to those two compounds. Accuracy and Precision. The results from testing in duplicate a series of eight samples, each containing a known weight of naphthalene in wash oil, show the method to have 95% confidence limits of &0.04%. These samples included a naphthalene range from 0.50 t? 4.007,. By testing 10 separate portions of a standard containlng 2.49% naphthalene the precision of the method was shown to be 3.2% relative (2 sigma or 95% confidence). The results of the analyses for accuracy and precision are tabulated in Table 111. NAPHTHALENE IN COKE OVEN GAS BY ULTRAVIOLET ABSORPTIOX
In determining naphthalene in coke oven gas advantage is taken of the greater sensitivity of ultraviolet measurements as compared to infrared. The naphthalene content of coke oven gas is determined by scrubbing the gas with cyclohexane and measuring the ultraviolet absorption of the cyclohexane-naphthalene solution a t 309, 311.5, and 318.5 mp. Xaphthalene exhibits an absorption maximum at 311.5 mp and minima a t 309 and 318.5 mp in cyclohexane. The true absorbance for naphthalene a t 311.5 mp is obtained by subtracting from the observed reading the corrected base line absorbance calculated from the absorbances observed a t 309 and 318.5 mp. This correction is calculated according to the theorem which states that the coordinates of a point, Pa, on a straight line, PIP?, where P, divides PIP, in the ratio mlmt may mlX2 mnXl FIYZ mnY1 be represented as
.
-
+
+
absorptivity, a, for the system and the volume of gas sampled, the naphthalene content can be calculated. Apparatus and Reagents. Beckman model DU photoelectric quartz spectrophotometer, with ultraviolet light source.
Two matched 10-cm. silica cells. Calibrated wet test meter. Dewar flask, pint or quart capacity. Scrubber tube, borosilicate glass, 20- to 25-mm. inside diameter, 7 to 8 inches long, and of approximately 100-ml. capacity, fitted with a cork through which are passed a gas outlet tube several inches long, and a gas inlet tube on the bottom of which is a medium grade, fritted-glass sparger. Cyclohexane, any grade, free of absorption peaks between 305 and 320 mp when referred to distilled water. Procedure. CALIBRATION OF SYsrEhf. The naphthalenecyclohexane system conforms to Beer's law over the concentration range of 0.001 to 0005% naphthalene. The system was calibrated by measuring the absorbances of a series of naphthalene standards a t 309, 311.5, and 318.5 mp and calculating the absorptivity, a. These standards were prepared from a 0.01% stock solution of pure naphthalene in cyclohexane by diluting lo-, 20-, 30-, 40-, and 50-ml. aliquots to 100 ml. with cyclohexane in separate 100-ml. volumetric flasks. The true absorbance, A , for naphthalene a t 311.5 mp was determined by the procedure outlined below for sample analysis. The data are tabulated in Table IV. An a value of 9.15 was obtained ( a = A/bc where b is the cell length in centimeters, and c is the concentration in grams of naphthalene per 100 ml. of cyclohexane). COLLECTION AND PREPAR.4TION O F SBYPLE. Place 100 ml. of cyclohexane in the scrubber tube and close with the stopper carrying the gas inlet sparger tube and the gas outlet tube. Place the assembled scrubber tube in a Dewar flask containing an ice-water slurry, and maintain the temperature of the cyclohexane a t about 10' C. (The freezing point of cyclohexane is approximately 6.5' C.) Open the valve of the gas sampling line and purge sufficiently to ensure collecting a representative sample. Close the valve. Attach the inlet sparger tube to the sample line valve with a suitable length of rubber tubing and similarly connect the gas outlet tube to a calibrated wet test meter. Open the sample line valve cautiously and adjust the flow of gas through the sparger tube to give a rate of about 1 cubi: foot per hour. Pass about 2 cubic feet of gas through the cyclohexane scrubber, close the sample line valve, and record the exact volume of gas measured by the wet test meter. Disconnect the sparger tube. Carefully transfer the cycloCLTRAVIOLET MEASUREMENT. hexane from the scrubber tube into a clean 100-ml. volumetric flask and dilute to volume with cycloheyane. Fill the reference
Table IV.
Absorptivity, a, from Naphthalene Standards"
Kaphthalene in Cyclohexane, Mg.per100hfl. 1. o 2.0
3.0 4.0
5.0
a
309mp 0,102 0,196 0.292 0.371 0.462
Absorbance 311.5mp 318.5m~ 0.174 0,017 0,025 0,335 0,493 0.040 0 652 0.050 0.062 0.813 Average
A
a
0.095
9.50 9.15 8.90 9.10 9.08 9.15
0.183 0.267 0.364 0,454
0.40-mm. slit; IO-cm. silica cells; 0.002 cell correction.
Table Y. Accuracy and Precision Data for Naphthalene in Coke Oven Gas" A4ccuracy Test
Precision Test
i
9
10 Average Standard deviation 9 5 R confidence limits
Naphthalene. hIg. ______ Added Found
Xaphthalene, I I g . 10.6 10.7 10.7 11.1 10.4 10.8 10.9
10.5 10.8 10.5 10.7
0.21 1 0 . 7 + 0.42
a Test solutions were prepared b y adding known weights of lene t o cyclohexane.
C.P.
naphtha-
V O L U M E 25, N O , 1 2 , D E C E M B E R 1 9 5 3 and sample cells with cyclohexane and determine the cell absorbance corrections a t 309, 311.5, and 318.5 mp with the spectrophotometer using a slit width of 0.40 mm. Empty the sample cell and fill it with the sample solution. Measure the absorbance a t 311.5 mp and 0.40-mm. slit width. If the absorbance is greater than 0.600, dilute 50 ml. of the solution to 100 ml. with cyclohexane. Using the same slit width, measure the absorbance of the final sample solution a t 309, 311.5, and 318.5 mp. Correct the observed sample absorbances for CALCULATIONS. the cell absorbance corrections previously made. Using the corrected sample absorbances, calculate the base line absorbance of the naphthalene a t 311.5 mp by means of the following formula:
+
A B = (0.74 X A ~ O B ) (0.26 X
A318.j)
Subtract this absorbance from that read a t 311.5 mp to give the absorbance from naphthalene only:
-4 = A
3,I.b-
A B
G = -1543 A 9.15 V where G = grains of naphthalene per 100 cubic feet of gas and TT = volume of gas sampled, cubic feet If dilution of the initial 100-ml. sample volume was necessary, the proper dilution factor must be incorporated into the above calculation. Interferences. No adverse effects on the determination of
1821
naphthalene by this method were observed from the presence of light oils (benzene, xylene, and toluene) when added to naphthalene-cyclohexane solutions. These light oils are the most likely interfering compounds present in coke oven gas. Accuracy and Precision. The accuracy of measuring naphthalene in cyclohexane by this method was determined by analyzing five samples, each containing a known weight of naphthalene in cyclohexane. The 95% confidence limits as determined from this dat,a are =k0.42%. Precision was determined by analyzing 10 separate portions of one of the standard samples which contained about 10 mg. of naphthalene per 100 ml. of cyclohexane. The results show a relative error for precision of 1.67,. The data for the accuracy and precision tests are hbulated in Table V. LITERATURE CITED
Altieri, V. J., “Gas Analysis and Testing of Gaseous Lfaterials,” pp. 400-3, New York, American Gas Association, 1945. (2) Altieri, V. J., “Gas Chemists’ Book of Standards for Light Oils and Light Oil Products,” pp. 232-4, Kew York, dmerican Gas ,issociation, 1943.
(1)
(3) Ibid., p. 235. (4) Reichardt, P. E., and White, D. L., IND. ENG.CHEM.,h . 4 1 . . ED., 18,286 (1946). ( 5 ) Schubert, S.,Gas-u. Tt‘ussrrfuch, 92, 277-8 (1951).
RECEIVED for review
April 20, 1953. ilccepted September 17, 1953.
Flame Spectrophotometric Determination of Sodium and Potassium In Viscous Solutions or Plant Extracts HOWARD M. BAUSERMAN AND ROB ROY CERNEY, JR.’ Research Laboratory, American Crystal Sugar Co., Rocky Ford, Colo.
A new method was needed for the routine determination of sodium and potassium in plant extracts, using the Beckman Model DU spectrophotometer and its flame attachment. The principle of operation of this instrument is such that viscosity and other characteristics of the fluid under test affect the flame color intensity. When properties of the fluid carrier alter the flame color intensity, this instrument is ideally suited to the application of a lithium internal standard method. Synthetic organic mixtures in solutions containing known quantities of the elements in question were tested. The apparent concentrations of sodium and potassium found spectrophotometrically could be corrected by use of concentration correction factors. Based on this work, a controlled analytical method is developed for solutions having characteristics differing from water.
T
HE instrument used in all the following experimental work was the Beckman Model DU spectrophotometer and the Model 10300 flame attachment, described by Gilbert et al. (6). It is set to measure the intensity of one of the characteristic wave lengths of light produced by exciting the element in question in an oxygen-natural gas flame. In this instrument the rate of introduction of the solution containing the element is controlled by causing it to pass through a glass capillary tube under a constant pressure differential. When the unknown solution has the same physical characteristics as those of a known standard solution, the rate of introduction vi11 be the same, and the line intensities can be directly related to concentrations (3). Often the material to be tested will have various kinds of organic matter as contaminants. The resultant changes in viscosity and surface tension may cause the unknown sample to be introduced a t a rate different from that of the known standard solution (4,IO). Various authors (1, 2, 7 , 9. 11-13) have proposed wet or dry aahing of the sample to remove organic matter 1
Present address, Ides1 Cement Research Laboratory, Boettcher, Colo.
or have proposed to ignore its presence. To avoid ashing, other Rorkers (4, 8, 10) have proposed that the same amounts of organic matter be added to the standard solutions as were contained in the unknown solution. If this method is used, the samples must contain a relatively constant concentration of the contaminants. When the unknown solutions have variable physical characteristics, particularly in regard to viscosity, the rate of introduction varies and a problem presents itself, as the concentrations are not directly correlated to line intensities. Such a problem, x-hich becomes acute in the case of certain plant extracts, can be solved by the introduction into the solution of a rarely occurring element, in this case lithium. Lithium has been used as an internal standard in other t \ pes of flame photometers for a number of years (4, IO). Its use, however, with the Beckman instrument for this purpose has been discouraged (6) because of uncompensated errors that may arise. As these errors may be significant in precise work, they should be appropriately compensated by the means of the conditions stipulated below. Errors due to interference caused by the presence of ions other
