Photometric Analysis of Phosphate Rock CHARLES J. BARTON Znternational Minerals & Chemical Corporation, Mulberry, Fla. Methods are described for determining four of the principal constituents of phosphate rock with a single sample. Acid-insoluble material is determined gravimetrically, while phosphate iron and aluminum are determined colorimetrically. The phosphovanadomolybdate method is used for the phosphate determination, 1,lO-phenantholine for iron, and alizarin for aluminum. Interference of iron in the alizarin method for aluminum is practically eliminated by making measurements at 370 millimicrons. The accuracy obtained using the methods described compares favorably with that of the commonly used gravimetric and volumetric methods and the colorimetric methods are considerably faster.
T
HE m6thods in general use for the determination of the
insolubles is required. Because nitric acid interferes with the colorimetric determination of iron and it is considered desirable to keep the chloride conccntration in the filtrate as low as possible, approximately 6 X hydrochloric acid was chosen to dissolve the acid-soluble portion of the rock.
most important constituents of phosphate rock-namely, acid-insoluble, calcium phosphate, iron, and aluminum-are volumetric and gravimetric. They have been in use for many years, with minor changes, and although they give satisfactory results in the hands of a good analyst, they have several disadvantages. Three samples and two additional weighings are required for the four determinations. The stirring and filtration equipment needed for large scale volumetric phosphate determinations occupies a considerable amount of desk space. The determination of combined iron and aluminum usually by precipitation as phosphates after removal of calcium, is very slow and requires a good deal of desk space for filtration. The aluminum result is calculated by subtracting the iron, determined volumetrically, from the combined iron and aluminum. The doubtful value of the indirect aluminum determination is reflected in the wide variations in results reported by different analysts on identical samples of phosphate rock. This paper presents methods for determining the abovementioned constituents of phosphate rock using only one sample, with a minimum of laboratory space and equipment. 411 the determinations are made by colorimetric methods, except for the gravimetric determination of acid-insolubles.
Procedure.
with a watch glass, and boil 2 to 3 minutes. Dilute to about 25 ml. with distilled water and wash the contents of the beaker into a fast filter paper. Wash the paper four times, catching the filtrate and washings in a 500-ml. volumetric flask. The funnel can be placed directly in the neck of the flask, eliminating the need for any fixed filtration equipment. Ignite the filter paper and its contents in a porcelain crucible a t 700" t o 800" C. for 15 minutes. Cool, brush the insoluble material onto a counterbalanced pan or watch glass, and weigh to the nearest 0.1 mg.
ACID-ISSOLUBLE MATERIAL
The method used by most Florida phosphate chemists (1) for determining the acid-insoluble content of phosphate rock calls for treating the sample with a mixture of 4 parts of concentrated nitric acid to 1 part of concentrated hydrochloric acid and boiling until all the nitrogen fumes are removed, as in the volumetric phosphate method, so that the filtrate can be used for the determination of phosphate. Concentrated hydrochloric acid is permitted as the solvent when only the determination of
Table I.
Sample
% Cas(PO4)z Volumetric
4v. Blin. Max. 75.27 75.53 76.02 747 69.26 69.71 70.24 847 71.86 72.40 73.50 947 78,97 80.09 80.67 1047 74.73 76.42 75.64 1147 68.24 69.08 69.79 1247 76.93 77.43 78.16 148 7.5 9.5 74.86 75.31 248 71.98 72.91 72.37 348 75.31 76.87 76.05 448 67.90 69.05 68.52 85120 76.82 77.57 ..... 77.20C ... Author's results. b Average corrected for obviously erroneous results. Includes both volumetric and gravimetric results. R47
Data and Discussion. Columns 2 and 3 of Table I give a comparison of results obtained on a number of samples, using the above procedure, with the average of the results obtained by a number of analysts using the method described in ( 1 ) . These samples, known as monthly check samples, are prepared and issued by H. H. Edwards, chief chemist, Florida Phosphate Division, International Minerals and Chemical Corporation. Identical samples are sent out once a month to a number of laboratories engaged in phosphate rock analysis. When the results are assembled, a minceographed sheet is sent to each analyst sho-ving all the results reported. These samples thus serve as "unknowns" to a large part of the phosphate industry Results are reported on the dry basis. The data in columns 2 and 3 show an average difference of 0.137, and a maximum difference of 0.28%. The maximum deviation is well within the range of figures usually reported for this determination and the results of the two method,q are CODsidered in essential agreement.
Analysis of Standard Samples
7% Acid-
Insoluble Av. reported CJBa 5.40 5.42 5.74 5.64 8.85 8.94 4.03 4.15 5.77 5.76 8.33 8.16 6.51 6.65 7.09 7.22 9.06 9.28 5.68 5.96 10.08 10.19
Weigh a 0.5-gram sample of dry, ground rock
( - lOO-mesh) to the nearest 0.5 mg. and transfer to a 100-ml. beaker. Add 10 ml. of 1 to 1 hydrochloric acid, cover the beaker
% Caa(P,Or)? Photometric C J B Av. Min. Max. 75.40 76.4 74.7 70.0 69.85 69.2 71.7 71.64 71.6 80.5 79.87 79.3 75.65 75.95 75.4 69.2 68.87 68.4 77.0 76.86 76.7 75.5 75.24 74.5 73.1 72.13 71 .O 76.3 76.09 74.8 68.8 68.50 68.2 78.2 77.2 76.1
1068
% Fez08 Av. reported 1.28 1.1Ob 0.87 0.61 0.85 0.99 0.83 0.98 1.08 1.06 1 .oo 0.89
CJB 1.23 1.10 0.88 0.61 0.86 1.02 0.88 0.93 1.03 1.05 0.97 0.85
% AlnOa AT. reported 1.03 0.85 0.97 0.71 1.06 1.06 0.92 0.89 0.79 0.83 0.94 0.80
CJB 1 .OB
....
1 .oo 0.68 1.06 1.17
0.77
0.86 0.71 0.86 0.87 0.82
V O L U M E 20, N O . 11, N O V E M B E R 1 9 4 8
360
I
380
WAVELENGTH IN MiLLlMlCRONS I I I 400
420
440
460
480
Figure 1. Transmittance of Phosphovanadomolybdate Solution and Reagent Blank PHOSPHATE
The determination of phosphorus pentoxide or tricalcium phosphate (usually referred to in the phosphate trade as B.P.L., bone phosphate of lime) is by far the most important determination in the laboratories concerned with phosphate rock analysis. Although colorimetric phosphorus methods have been in use for many years,, fen- efforts have been made to apply these methods to phosphate rock, probably because of the general belief that colorinietiic methods are applicable only to the determination of low concentrations of materials I d i o v and Icazaiineva ( 7 ) used visual comparison of molybdenum blue solutions for field b s t s on phosphate rock and claimed a maximum error of 2cc, presumably 2C;. phosphorus pentoxide. Eddy and De Cds (4) used a photoelectric colorimeter for the analvsis of bone phosphate, which is similar to phosphate rock in phosphate content. llehlig (10) and others have shonn that the w e of a sensitive photoelectric colorimeter or spectrophotometer n ith a reliable colorimetric method can give result3 that compare favorably with those of volumetric or gravimetric methods for high concentrations. The phoqphovanadomolybdate (I’VAI) method originally proposed by Misson (11) has been used principally in steel analysis (6, 8, 12, 14) but has also been applied to the analysis of iron ore (16) and biological materials (9). .Is pointed out by Kitson and Wellon (S), who made a thorough spectrophotometric atudy of the method, it has several advantages over the mole commonly used molybdenum blue methods. One advantage for phosphate rock analysis is the lower sensitivity that permits larger samples to be used. These authors also state that sulfuric, perchloric, and hydrochloric acids behave much like nitric acid in the phosphovanadomolybdate determination. Transmittance measurements were made m-ith a Beckman Model DU spectrophotometer on a phosphovanadomolybdate
1069 solution and a solution containing only the reagents. The data shown graphically in Figure 1 indicate a maximum spread bet n een the two solutions a t about 400 millimicrons and this Tyave length has been used for all the spectrophotometric determinations reported in this paper. Higher wave lengths (430 to 480 millimicrons) have been recommended in the literature for this determination (8, 9, I C ) , probably because most work with the method has been with steel and iron samples where the iron coloration would interfere a t lower wave lengths. Calibration data obtained n4th the Beckman spectrophotometer, using 1 .O-em. cells, shoned a linear relationship betueen density reading and concentration of phosphorus pentoxide up t o 4.0 mg. of phosphorus pentoxide in 100 ml. of solution. With the Klett-Summerson photoelectric colorimeter equipped 15 ith a blue filter (niaximum transmittance a t 425 millimicrons) and test tubes 12 nim. in inside diameter the linear relationship between concentration and scale reading only extended to 2.0 mg. of phosphorus pmtouide per 100 ml. The variations in reagent concentrations considered permi-sible are: nitric acid, * 1 nil. (1 t o 1 solution); ammonium molybdate solution, *0.2 nil. : ammonium vanadate solution, *0.2 ml.; mixed rragent * 0.5nil. These reagents are conveniently added from burets. Temperature variations were found to have a small but important effect on the color intensity of the phosphovanadomolybdate solutions, as noted by Center and Willard (3, 16). Increasing the temperature 10” C. increases the density reading about 0.010 unit. This is too great an effect to be neglected for accuinte determinations ;\lost of the data reported in this paper i5ere obtained by adding the reagents separately to the sample solution, as described in the -?parate reagent procedure below. Later in the investigation it iyas found that the procedure could be simplified by adding the reagents together, and also that the color development could be speeded by reducing the nitric acid concentration from 1.6 V t o 0 8 S nithout affecting the accuracy of the determinations. h n earlier attempt t o use the acid concentration recommended bv Kitqon and IIellon ( 8 ) ,0.5 S,nas abandoned becauqe of erratic results. Reagents. Ammonium molybdate, lOyc solution (weight by volume) was prepared by boiling the salt in mater. “Acid molybdic S5y0’,”may be substituted for the ammonium molybdate. ‘ Ammonium vanadate, 0 5 % solution. Five grams of the salt were dissolved in hot Tvater, and 20 ml. of concentrated nitric acid added and diluted to 1 liter after cooling t o room temperature. Metavanadic acid (red) sometimes precipitates when the nitric acid is added, but this usually redissolves on stirring. Sitric acid, c.P.,1 to 1 solution. Mixed Reagent. Forty grams of ammonium molybdate were dissolved in about 400 ml. of distilled water, 1.0 gram of ammonium vanadate was diseolved in about 300 ml. of distilled water, 200 ml. of concentrated nitric acid n-ere added, and the two solutions m-ere mixed and diluted t o 1 liter. Standard Phosphorus Pentoxide Solutions. Stock solutions containing 1.0 nig. of phosphorus pentoxide per ml. were prepared from reagent grade potassium dihydrogen phosphate and sodium phosphate dodecahgdrate. These solutions were standardized gravimetrically and diluted to give 0.10 mg. of phosphorus pentoxide per nil. Separate Reagent Procedure. Dilute the filtrate and nxshings from the determination of acid-insoluble to the mark on the 500ml. volumetric flask and mix thoroughly. Pipet a 5-ml. aliquot oortion into a 100-ml. volumetric flask, and add in the order Lamed: 20 ml. of 1 t o 1 nitric acid, 5 ml. of ammonium vanadate solution, and 10 ml. of ammonium molybdate solution. These reagent proportions are essentially the same as those given bv Snell (16). Rotate the flask to mix its contents during the addition of ammonium molybdate solution. Dilute t o the 100-ml mark, mix, and xllo~vto stand for 12 to 15 minutes to permit full development of color. lleasure the transmittance or optical density of the solution a t 400 millimicrons or take the reading with a photoelectric colorimeter equipped with a blue filtcr Use a reagent blank solution in the reference cell. With each group of samples, take a portion of standard phosphorus pentoxide containing approximately the same amount of phosphoruc
1070
ANALYTICAL CHEMISTRY
pentoxide as the sample aliquots or, preferably, take an aliquot portion of a phosphate rock sample solution of known phosphorus pentoxide content, and treat in the manner previously described for sample aliquots. Calculate results as follows for measurements with the spectrophotometer, using standard phosphorus pentoxide solutions for calibration: Calibration factor =
comparison of volumetric and photometric results obtained on a number of monthly check samples and the Bureau of Standards No. 120 sample (Florida phosphate rock). With only two exceptions, the average photometric result is well within the range of the volumetric results reported, and the variation in photometric results is no greater in general than the variations in volumetric results from the different laboratories. Most of photometric results were obtained with the spectrophotometer. Table I1 gives a comparison of photometric results with routine volumetric determinations on a group of plant control samples, as well as an indication of the variations in volumetric results between laboratories. The difference between volumetric and photometric results is about the same order as the differences between the two laboratories. The average difference in optical density at 400 millimicrons for 11 sets of duplicate determinations was 0.0034, equivalent to 0.4y0 tricalcium phosphate. The maximum difference was 0.007, equivalent to 0.8%. Duplicate determinations by the volumetric method are considered to be in good agreement if they check to 0.2% and differences of 0.5% tricalcium phosphate are not unusual. Obviously, the second decimal figure commonly reported in volumetric tricalcium phosphate results cannot be considered significant. Most of the work has been carried out with Florida phosphate samples, but a limited number of comparisons of volumetric and photometric phosphorus pentoxide determinations on Tennessee and Montana rock samples showed no indication of interfering elements in these samples. Table I11 shows the results of duplicate photometric determinations of a group of Tennessee phosphate rock samples and also the volumetric data. The photometric phosphorus pentoxide method is significantly faster than the volumetric method for a small number of samples but for larger groups of samples the difference is not so great. A single photometric determination can be completed in 14 minutes, using the mixed reagent procedure, as compared to approximately 50 minutes for a single volumetric determination. For batches of 18 samples the time saving amounts to about 15 minutes. The reagent cost for the phosphovanadomolybdate method is 0.9 cent per determination as compared to approximately 3.5 cents per determination for the volumetric method. The only desk space required for this method is that occupied by the colorimeter or spectrophotometer plus temporary use of space for the filtering and aliquoting operations.
mg. of PzOs optical density
Calibration factor X sample optical density X 100 = % p*os mg. of sample in aliquot When an aliquot portion of a standard sample containing the same weight of sample as the unknown solution is used for calibrating the spectrophotometer, the calculation is simplified: Optical dewity of unknown solution .Optical denhitv of standard solution X % PzOb in standard = % P205in unknown The same methods of calculation can be used for the KlettSummerson photoelectric colorimeter or other colorimeters equipped with logarithmic scales, but in this case the calibration factors will not give accurate results over a wide concentration range. This necessitates the use of standards closer to the unknown in order to obtain the best accuracy. Mixed Reagent Procedure. Follow the yeparate reagent procedure to the point where the reagents are added. Dilute the aliquot portion of the sample solution to about 50 ml. with distilled water. .4dd 25 ml. of mixed reagent, dilute t o the 100-ml. mark, and mix. The transmittance can be determined 2 minutes after mixing. Calculations are the same as for the separate reagent procedure. Data and Discussion. Early results with the phosphovanadomolybdate on phosphate rock samples, calculated by means of calibration data obtained with a single series of standard phosphorus pentoxide solutions, showed rather wide variations. Consequently, the standard sample method, which eliminates the effect of a number of minor errors such as temperature variations and pipet errors, was adopted. The minimum instrumental error with the Beckman spectrophotometer is 0.12% tricalcium phosphate (B.P.L.). Using the Klett-Summerson colorimeter with 20-mm. cells and the No. 42 filter, the error is 0.24%. The instrumental error is one of the ,principal sources of error in this determination. Obviously, care must be used in making the photometric reading in order to obtain good results. Table I, columns 4 to 9 inclusive, gives a
IRON
Table 11. Analysis of Plant Control Samples sample NO.
1 2 3 4 5 6 7 8 9 l1 o1 12
% B.P.L. Volumetric Lab. 1 31.36 53.47 19.20 66.62 17.02 13.94 9.56 11.06 7.87 71.50 79.65 78.75
Lab. 230.68 54.48 16.57 66.13 16.97 14.07 9.73 10.25 7.53 71.39 80.19 78.19
% B.P.L., Photometric Klett 31.6 53.2 18.9 66.9 17.2 12.9 9.4 11.0 8.4 71.2 79.4 77.6
% B.P,L.
Spectrophotometer 31.2 55.5 18.5 66.4 18.0 13.2 9.5 10.7 8.3 71.5 80.6 79.0
Sample Volumetric, No. Lab. 1 13 36.22 14 64.98 15 20.95 16 73.76 17 34.81 18 14.13 19 44.40 20 20.98 21 10.79 22 68.94 23 80.40 24 74.77
% B.P.L., Photometric Klett 36.7 65.4 21.0 72.5 34.8 13.5 42.8 21.5 9.2 69.6 82.0 76.6
SpectroDhotometer 36.7 63.0 21.2 71.9 34.6 13.5 41.7
...
10.1 68.1 80.9 75.0
Table 111. Analysis of Tennessee Phosphate Rock Sample % B.P.L,, No. Volumetric 1 73.42 2 71.92 3 66.57 4 65.64 5 63.67 6 65.07 7 70.92
% B.P.L., P h o t o m e t r i L 1
72.6 71.0 66.6 65.3 64.3 64.3 71.0
2 74.0 71.0 67.0 66.0 63.8 63.6 70.5
A\,. 73.3 71.0 66.8 65.7 64.1 64.0 70.8
sample
KO. 8 9 10 11 12 13 14
% B.P.L. Volumetric 71.42 71.02 68.47 68.17 70.69 68.27 72.92
% B.P.L., Photometric 1 70.1 71.1 69.6 66.8 73.6 68.7
73.4
2 70.8
...
69.5 67.0 72.7 69.1 73.0
Av. 70.5 71.1 69.6 66.9 73.2 68.9 73.2
P r e l i m i n a r y W o r k . -4s Fortune and hfellon (61have reported a thorough spectrophotometric study of the 1,lOphenanthroline method for iron, it was only necessary to calibrate the spectrophotometer and colorimeter with standard iron solutions prepared from pure iron wire, and test for possible interfering substances in the phosphate rock samples. This was done by analyzing samples of known iron content. S o difficultywas experienced in checking the results of volumetric iron determinations and it was concluded that the ions normally present in solutions of phosphate rock do not interfere in the 1.10-phenan-
1071
V O L U M E 20, N O . 1 1 , N O V E M B E R 1948
20
-
0.05 MG. FE
WAVELENGTH IN MILLIMICRONS
1 Figure 2.
1
1
1
1
1
-
1
1
~
Transmittance Curves for Aluminum and Iron with Sodium .4lizarinsulfonate
throline method for iron. The reagents are prepared, with minor changes, according to the directions given by Fortune and Mellon (6). The Corning S o . 5543 filter is suitable for use with photoelectric colorimeters.
Procedure. Pipet an aliquot portion of thc filtrate from the determination of acid-insoluble, containing 0.05 to 0.5 mg. of iron, iiito a 100-mi. volumetric flask. Dilute to approximately 50 ml. and add in the order named: 1.0 ml. of 1 to 1 hydrochloric acid, 2.0 ml. of 107, hydroxylamine hydrochloride, 5.0 ml. of 20% ammonium acetate, and 5.0 ml. of 0.2y0 1,lO-phenanthroline (all percentages on a weight by volume basis.) Allow 2 minutes between the addition of the hydroxylamine and ammonium acetate for complete reduction of the iron. Dilute to the 100-ml. mark, mix, and allow the solution to stand at least I5 minutes to prrmit full development of the color. Determine the density with a spectrophotometer a t 510 or 520 millimicrons or the scale reading with a photoelectric colorimetric and convert the instrument reading into milligrams of iron or ferric oxide by means of the calibration data. Tt is not necessary to run a standard with each group of samples, although an occasional check on a standard sample is considered desirable. Data and Discussion. Calibration data for both the Beckman spectrophotometer and the Klett-Summerson photoelectric colorimeber showed a linear concentration-scale reading relationship over a wide range of concentrations, eliminating any need for calibration curves or charts. For the spectrophotometer a t ,520 millimicrons, the smallest division in the density scale that can be estimated (0,001) is equivalent to 0.00075 mg. of ferric oxide or 0.003% on a 25-mg. sample, normal size for Florida phosphate samples. The sensitivity with the KlettSummerson colorimeter using test tubes 12 mm. in inside diametrr and Corning S o . 5543 filter is 0.002 mg. of ferric oxide per scale division or O.OOSC7, on a 25-mg. sample. Duplicate aliquot8 frequently check as closely as the instruments can be read, and the maximum difference is seldom more than 0 . 0 2 5 for samples in the neighborhood of 1% ferric oxide. h comparison of the average volumetric iron results on the monthly check samples with the results of photometric iron determinations is shown in columns 9 and 10 of Table I. The average difference is 0.03V0 and the maximum difference is
0.05% ferric oxide. As the usual iange of volumetric iron results reported on the check samples is about 0.3 to 0.4% ferric oxide, the photometric iron method is probably more accurate than the volumetric method. Table IV shows a comparison of volumetric and photometric iron determinations on a group of 11 samples of Tennessee phosphate rock covering the range from 2 to llyOferric oxide, using both the Klett-Summerson colorimeter and Beckman spectrophotometer. The average difference betneen the results obtained by the two methods is O.lOVo for the colorimeter and 0.094, for the spectrophotometer. TKO-and 5.0-mg. aliquot. aere used with these samples. -1single photometric determination of iron can be completed in 20 minutes from the time the sample is dissolved and diluted, or about 27 minutes in all. A large part of this time is accounted for by the standing period, requiring no attention from the arialvst. A volumetric iron determination requires approximately 45 minutes. The main source of error in the photometric determination of iion seems to be failure to reduce the iron completely. A fresh solution of hydroxylamine hydrochloride readily reduces iron, hut the solutions appear to lose strength on standing. Conyequently, only a few days’ supply of this reagent should be Inepared a t one time. Another source of error is a small amount of iron that remains with the insoluble residue. It usually amounts to only 0.01 or 0.0274 and can be neglected for most of iron in the residue can be readily ~ routine ~ work. ~ The ~ amount . determined by extracting with a little hot concentrated hydrochloric acid and measuring it by the regular photometric procedure.
Table IV.
Comparison of Volumetric and Photometric Results for Iron
Sample
% Fe?Ol
NO.
Titration 1.94 2.74 3.50 4.32 3 , I9 7.00 8.24
1 2
3 4
5 6 7 8 9 10 11
8.70 9.84 10.18 11.30
% ’ FenOz, Photometric Klett 1.97 2.76 3.53 4.12
...
7.17 8.38 8.55 9.70 10.15 11.10
Spectrophotometer 1.96 2.74 3.50 4.29 5.10 7.20 8.40 8.51 9.80 10.30 11.17
ALUMINUM
Preliminary Work. The colorimetric determination of aluminum with sodium alizarinsulfonate (Alizarin Red S) originally proposed by Atack (8) has been studied more recently by Yoe and Hill ( I T ) , Musakin (Is),and others. According to Yoe and Hill ( I T ) , iron must be removed if more than a trace is present. They state that chloride, calcium, and phosphate ions do not interfere. The effect of fluoride ion is not mentioned, but it seriously interferes with the aluminum determination. Therefore, it was necessary to find methods of eliminating the influence of iron and fluorine in order to determine aluminum in phosphate rock with sodium alizarinsulfonate. Fluorine is readily removed from the solutions by evaporating to fumes with sulfuric acid. Small aliquots of the sample solution are used (normally 10 ml.), so that only a few minutes are required for the evaporation. The iron interference wm a more difficult problem. Transmittance curves on the reagent, iron reagent, and aluminumreagent solutions (Figure 2) indicated a possible solution of the problem. They shorn that, although the iron-reagent complex absorbs very strongly throughout most of the visible spectrum, the curve approaches that of the reagent in the near ultraviolet, while the aluminum-reagent complex has a higher transmittance than the reagent. A more detailed study of the violet and near
ANALYTICAL CHEMISTRY
1072
6 5 t
i
‘
WAVELENGTH IN
MILLIMICRONS
Figure 3. Transmittance Curves for Aluminum and Iron in Near Ultraviolet Region
ultraviolet region of the spectrum (Figure 3) revealed that the transmittance of the reagent and iron-reagent solutions is the same a t 370 millimicrons. Consequently, it appears feasible to determine aluminum a t this wave length without interference from iron. Later tests with mixtures of iron and aluminum indicate that iron may interfere, to some extent, by using up part of the reagent. With a mixture containing 50 micrograms of iron and the same amount of alumina, the iron had no effect, while with a mixture of 100 micrograms of iron and alumina, the apparent amount of alumina was decreased slightly. Seventy-five micrograms of iron (107 micrograms of ferric oxidej did not change the optical density rcading of a solution containing 100 micrograms of alumina, Inasmuch as the sum of the ferric oxide and alumina in Florida phosphate rock, to which this method has been so far confined, does not usually go much higher than 2% or 200 micrograms on a 10-mg. scniple, the iron interference has not been investigated with iron-alumina ratios higher than 1.5 to 1, where iron has a significant effect. The effect of iron a t higher concentrations of iron and aluminum can be minimized by using higher concentrations of reagent than is recommended in the following procedure. An attempt was made to determine aluminum with a photoelectric colorimeter using a Corning KO.5874 filter, but the sensitivity of the instrument was too low to permit accurate determinations. Solutions of the reagent, sodium alizarinsulfonate, are strongly colored and give a low transmittance reading a t 370 millicrons. Therefore, it is important to use the same amount of reagent in the blank and sample solutions. Variations in the volume of ammonium hydroxide and acetic acid solutions added had an effect on the transmittance and must be controlled to *0.2 ml. Increasing the amount of sulfuric acid present, used to remove fluorine, from 0.1 to 1.0 ml. of 1 to 1 solution had no significant effect on the transmittance reading of solutions containing 100 micrograms of alumina. Temperature variations from 20 O to 30” C. also did not affect the reading appreciably. Changing the volume of solution a t the time of lake formation from 25 to 45 ml. likewise had no effect. Atack ( 2 ) allowed his solutions to stand 5 minutes after adding ammonium hydroxide before acidifying with acetic acid. KO difference was observed between solutions allowed to stand 1 minute and solutions allowed to
stand 15 minutes. The solutions are fairly stable, and show no change in transmittance for a t least 30 minutes after acidifying. The lakes frequently coagulate but the precipitate can be redispersed by shaking vigorously and the transmittance reading is apparently not affected. Even after the solutions have stood overnight and are thoroughly coagulated, they can be redispersed with little change in the transmittance reading from the freshly prepared solutions. Procedure. Pipet a 10-ml. aliquot of sample solution, for samples containing 2.0% R2O3 or less, into a small beaker, add 0.5 ml. of 1 to 1 sulfuric acid, and evaporate to strong fumes of sulfur trioxide. Dilute with a little distilled rrater and wash the contents of the beaker into a 100-ml. volumetric flask with about 30 ml. of distilled water. Add 5.0 ml. of 0.1% sodium alizarinsulfonate and 10 ml. of 3 S ammonium hydroxide, mix, and then add 5.0 mi. of 5 S acetic: acid. Dilute to the 100-ml. mark, mix thoroughly, and read the optical density a t 370 millimicrons. Run a blank along with each group of samples. Data and Discussion. The calibration data indicate that Beer’s law holds up to 7 50 micrograms of alumina, using 5.0 ml. of 0.1% reagent solution. Above 250 micrograms the sensitivity drops off rapidly and erratic results are obtained. The optical density reading used for plotting and calculations is the difference betv,-een the reagent blank density and the sample solution density. The minimum instrumental error is equal to 1.4 micrograms of alumina, or 0.014% on a 10-mg. aliquot. The aluminum method was developed more recently than the other methods described; consequently only the figures on the last six check samples were determined before the results of other analysts were known. The comparison of photometric results for alumina with the average of indirect gravimetric determinations is given in the last two columns of Table I. The maximum difference noted is o.15yO and the average difference is 0.05% I n 31 determinations made on five of these samples, the mean deviation from the average reported value was 0.065% with a maximum error of 0.19% alumina. The mean deviation from the author’s averages was 0.060% alumina, with a maximum deviation of 0.14%. Khile these deviations are undesirably large, the accuracy conipares favorably with that of the indirect method now in use. The alumina results reported on the monthly check samples seldom var>-by less than 40y0 betxeen the maximum and minimum values, and occasionally vary by as much as 200%. One of the principal sources of error in this method has been the variation in density readings on the reagent blank. Density readings on standard alumina solutions and phosphate rock samples, in general, have been more reproducible than the readings on the blank solution over a period of weeks, possibly because of the greater effect of minor variations in reagent concentrations on the blank reading. ACKNOWLEDGMENT
The author is indebted to present and former assistants, Charles Colter, Ernest Vlk, and Doyle Lamb, for many of the data used in this paper. Thanks are also due to I. hlilton LeBaron, research supervisor, Phosphate Division, and Paul D. v. Manning, vice president in charge of research, International Minerals and Chemical Corporation, for their encouragement in this work. LITERATURE CITED (1) Assoc. Florida Phosphate Mining Chemists, “Methods Used and
Adopted by the Association of Florida Phosphate Minine Chemists,” 1948. J . Soc Chem Ind., 34, 936 (1915). (2) iltack, F. W., (3) Center, E. J . , and TT*illard, H. H., IND. EXG.CHEY.,ASAL.ED.. 14, 287 (1942). and De Eds, F., ISD.EXG.CHEM.,ANAL.ED., 9 (4) Eddy, C. W., 12 (1937). (5) Fortune, W.B., and Mellon, hl. G., Ibid., 10, 60 (1938). (6) Getzov, €3. B., Zaeodskaya Lab., 4, 349 (1935). (7) Isakov, E. N., and Kaaarineva, T’. A., Udobrenie i Urorhai 2. 416 (1930).
V O L U M E 20, NO, 1 1 , N O V E M B E R 1 9 4 8 (8) Kitson, R. E., and Mellon, M ,G., IND.ENG.CHEM.,ANAL.ED., 16,379 (1944). (9) Koenig, R. A., and Johnson, C. R., Ibid., 14, 155 (1942). (10) Mehlig, J. P., Ibid., 7, 27, 387 (1935); 9, 162 (1937). (11) Misson, G., Chem. Ztg., 32, 633 (1908). (12) Murray, W.M.,Jr., and Ashley, S. E. Q., IND.ENG.CHEM., ANAL.ED.,10,1 (1938). (13) Musakin, A. P., 2. anal. Chem., 105,351 (1936). (14) Schroder, R . , Stahl u. Eisen, 38, 316 (1918).
1073 (15) Snell, F. D., and Snell, C. T., “Colorimetric Methods of Analysis,” Vol. I, p. 485, New York, D. Van Nostrand Co., 1936. (16) Willard, H. H., and Center, E. J., IND.ENG.CHEM., AXAL.ED., 13,816 (1941). (17) Yoe, J. H., and Hill, W.L., J . Am. Chem. Soc., 50,745 (1928). RECEIVED May 8, 1948. Presented before the Division of Analytical and Microchemistry at the 113th Meeting of the AXERICAKCHmrIcAL SOCIETY Chicago, Ill.
Determination of Rare Earth Elements Yttrium in Uranium Compounds E L G. SHORT
AND
W. L. DUTTON, Imperial Chemical Industries, Ltd., Widnes, England
A method has been developed for the determination of rare earth elements together with yttrium in uranium and its compounds. The greater portion of the uranium is separated from the rare earths by taking advantage of the solubility of uranium nitrate in ether. The rare earths are then precipitated as fluorides and subsequently purified as hydroxides. The final determination is carried out spectrographically. On 50 grams of sample, 0.2 part per lo6of gadolinium can be detected by this method and for some of the other rare earths the sensitivity is even higher. Gadolinium can be detected with greater sensitivity by increasing the quantity of sample used.
N
0 CHEIIIC.iL method of determining rare earth elements
and yttLium in uranium compounds which in any way approaches the required limit of sensitivity, has been published and it was decided that the spectrograph offered the only practicable solution to the problem. For the quantitative determination of minimum quantities of any element, an ideal spectrographic technique requires that the solution of the element to be examined be quite free from foreign elements, which affect the evcitation of the element under investigation in an unpredictable manner. These requirements immediately raiied a chemical problem of some difficulty-viz., the separation of microgram amounts of rare earths from 20 or more grams of uranium and their subsequent isolation as chlorides in a solution containing as few other elements as possible. The investigation of the chemical qeparation and of the spectrographic technique proceeded to a certain evtent simultaneously as described below. SPECTROGR4PHIC DETERMIN4TION OF R 4 R E EARTH ELEMENTS
The spectrographic work vias carried out with a Hilger Automatic Littrow Model spectrograph furnished with quartz and glass trains, The quqrtz train was used with the wave-length setting 6000 to 2975 A., in preference to the greater dispersion given by changing over to the glass train, a t the higher wave lengths, since the former enabled the characteristic lines of all the rare earths to be photographed with only one setting. Spectroscopically pure samples of all the rare earth oxides except thulium were obtained from Adam Hilger, Ltd., and standard solutions were prepared to contain 10 micrograms of element per ml. Suitable amounts of these solutions were evaporated off in depressions drilled in Hilger H.S. copper rods 7 mm. in diameter and their spectra photographed using a 5-ampere direct current arc, the arc gap being 4 mm. The arc was fed from a 220-volt line, sufficient resistance being introduced to limit the current to the above amount; the voltage drop across the arc was 45 t o 50 volts. The optical arrangements used are described in detail below. In general, the minimum amount detectable was found to be approsimately 20 micrograms for each element. By the use of copper electrodes 5 mm. in diameter and the same exposure, the sensitivity was increased to 5 micrograms, but further decrease in electrode size or increase in current strength gave no improvement on this figure, owing to melting of the electrodes and heavy increase in background. A standard plate was then prepared for
each rare earth element in steps of 10 up to 50 micrograms. Above this latter amount the intensity of the characteristic lines was too great for quantitative work. The amount of rare earth element in an actual test XTas determined by direct visual comparison of the intensities of the characteristic lines a i t h these standard plates. The lines used for the identification and estimation of each element are detailed in Table I, together with “coincident” lines which are limited to strong lines of other elements. Wave lengths are quoted from the N.I.T. tables ( 2 ) . The effect of foreign elements in the solution to be evaporated on the electrode had been noted in previous work. In general, foreign elements diminish considerably the intensity of the rare earth lines, but the quantitative effect varies from element t o element. Of cations likely to be present, 1 mg. of aluminum was found to affect the rare earth results very little, whereas a similar amount of thorium was much more objectionable. Of the acid radicals, fluoride and sulfate were particularly undesirable. The tests of Table I1 were made by addition of other elements to standard rare earth solution, followed by evaporation on the electrode and arcing, etc. The results shoa clearly the depressing effect of extraneous ions. Calcium had a depressing effect and was also objectionable, because two of its main lines almost coincided with the two gadolinium lines usually used for analysis. EXTRACTlON OF RARE EARTH COMPLEXES WITH ORGANIC SOLVENTS
The technique of solvent extraction of metallo-organic complexes has given such favorable results in many cases that this seemed the most desirable method of chemical separation, if a suitable reagent could be found. h large number of organic compounds were tried, but although in certain cases the rare earths were extracted from pure solutions, the addition of the reagents necessary to keep the large amount of uranium present in the aqueous layer prevented the extraction of the rare earths themselves. The following reagents were thoroughly investigated: cupferron a t various pH values, sodium diethyldithiocarbamate, 0-benzoin oxime, sebacic and other organic acids, 0-naphthol, and a-nitroso @-naphthol.
