Determination. of Phosphorus in Aluminum and Aluminum Oxide by Radioactivation Analysis L. M. FOSTER and C. D. GAITANIS Aluminum Research Laboratories, N e w Kenrington, Pa.
A method is described for determination of trace amounts of phosphorus in aluminum and aluminum
The usual way to circumvent this difficulty is to irradiate along m-ith the u n k n o m , a small quantity of a pure compound of the element to serve as a standard. After irradiation, a known quantity of carrier of the element is added and a conventional chemical separation is carried out in both the standard and the unknown. The activities of the separated material are then compared under identical counting conditions. The ultimate sensitivity is determined bv the degree to which the element can be separated radiochemically pure; or, if impure, bv the degree to which the desired activity can be measured in the presence of the interfering activities. Phosphorus lends itself particularly well to determination by radioactivation analysis. It has a fair activation cross section of 0.23 barn. Its intermediate half life of 14.3 days is long enough to permit the decay of many possible interfering activities, but short enough to permit the ascertainment of radiochemical purity by half-life measurement. Moreover, it is a pure beta emitter and can be readily distinguished by absorption measurements from most contaminants that have associated gamma emission.
oxide by neutron activation analysis. After a sample is irradiated in a slow neutron reactor, a large, known amount of carrier phosphorus is added, then a conventional chemical separation for phosphorus is made. The radioactivity of the separated phosphorus is compared with that of a standard treated in the same manner to give the phosphorus content of the unknown. Phosphorus contents in the range of 0.001 to 0.0001% were determined to 5% of the absolute value. A practical sensitivity of 0 . 0 0 0 0 0 0 5 ~phosphorus is possible.
P
RACTIChLLY all methods for the separation of phosphorus are based on precipitation of ammonium phosphomolybdate. Numerous procedures for the subsequent treatment of this precipitate have been proposed; it may be redissolved and precipitated as magnesium pyrophosphate or the phosphorus may be estimated directly from measurement of the yellow precipitate by gravimetric, volumetric, or colorimetric means. The determination as magnesium pyrophosphate is not applicable in the case of commercial aluminum alloys and refined aluminum oxides where the phosphorus contents are very low, and the sample size is limited by solution difficulties. The other procedures depend upon a precipitate of the formula (NH4)sPO4. 12hIo03, which, however, generally contains slightly more molybdic oxide than is indicated ( 2 ) . Also, small amounts of silica, always encountered in the oxide or flux mixture, cause an increase in the mole ratio of molybdenum to phosphorus ( 1 ) . The presence of silica precludes the use of a colorimetric method, since silicon forms a compound with molybdenum that is reduced to a blue color. Also, the presence of aluminum interferes in the molybdenum blue reaction ( 3 ) . The technique of radioactivation analysis circumvents many of the difficulties usually encountered in the determination of trace elements. There are a number of descriptions of this technique in the recent literature (6-7). The technique can be simplj- stated. When any material is irradiated in a chain-reacting, neutron reactor, the activity induced in any given element can be calculated from the expression:
A ( t ) = (1.63fu+ X l O I 3 ) ( l -
,-At)
where
A ( t ) = activity per gram atom
f u
+ A t
= isotopic abundance
of the parent element cross section in barns = neutron flux in neutrons/sq. cm. sec. = decay constant = time of irradiation =
The reliability of the parameters in this relation is generally well known with the exception of the neutron flux, 4. This is uncertain, and without detailed knodedge of the operation of the pile throughout the peiiod of irradiation there is no assurance of its magnitude or constancy.
EXPERIMEYTAL
Activation. For irradiation, the samples n-ere sealed in the customary Oak Ridge-type of container. This consists of a commercial purity ( 1100 alloy) aluminum, impact-extruded can, inch in diamkter by 33/8 inches long. The aluminum alloy samples were in the form of turnings or strips of thin sheet. The aluminum oxide samples and the standards were sealed in individual capsules made by closing the ends of thin-walled, 1100aluminum tubing inch in diameter.
Table I.
Attenuation of Neutron Flux in Standard
Weight of
?;H4n2Poa, Gram 0.0012 0.0136 0.0586
Relative Specific Activity 1 . O O h 0.05 weighing error 1.03 0 01 probable counting error 1 . 0 3 + 0 . 0 1 probable counting error
*
I n some cases several irradiation cans n-ere required t o contain all of the samples of one series. I n these cases, a separate standard was contained in each can. The samples were activated for one week in the thermal neutron reactor operated by the Union Carbide and Carbon Corp. at Oak Ridge, Tenn. They were placed in the 14-2 positions of the pile. These positions are in the outer “cool” part of the pile where the fast neutron population is very low. Attenuation Effects in Standard. An initial study was made to determine the most suitable phosphorus standard to accompany the samples. I n the general case, a compound made up of elements whose absorption for neutrons is low in comparison to the standard element is desirable to minimize self-attenuation effects. The ammonium phosphates aptly fulfill this requirement, as nitrogen, hydrogen, and oxygen do not have large cross sections. Diammonium monohvdrogen phosphate decomposed slightly a t 130” C. overnight and was not suitable for that reason. Monoammonium dihydropen phosphate was stable and was used as the standard. If self-attenuation occurs, it is made apparent when the specific activities of samples of considerably different size are
1342
V O L U M E 27, NO. 8, A U G U S T 1 9 5 5 compared, and the effect can be corrected for by extrapolating the activity of a range of sample sizes down to zero weight. I n a separate experiment, approximately 0.001-, 0.01-, and 0.05-gram samples of the standard were irradiated and the relative specific activities were determined bj- counting equalweight portions of their ammonium phosphomolybdate precipitates. The results are given in Table I. It is seen that self-attenuation effects in the st,andard were absent. For the subsequent activations, a standard of approximately 0.02 gram was employed as a conveniently weighable amount. Attenuation in Samples. Errors could also result if the samples were sufficiently opaque to neutrons to result in an appreciable attenuation of the flux t,hroughout the thickness of the material. T h e comparison of the measured activity with that of the standard would then no longer be valid, as the activity would not be strictly proportional to the content of the element in question. I n the case of the metal samples, the composition was sufficiently well known from the analysis and history of the material so that there could be assurance that self-attenuation would be absent. Such information was not as complete in the ca;se of the aluminum oxides, however. Boron was the most likely high cross section contaminant in the oxides: but its concentration was established spectroscopically to be below 0.017,. S o information was available, however, on the concentration of the rare earth element? and, since some of these have extremely high cross sections, very minute amounts could he deleterious. A supplementary experiment TI-as carried out with one representative oxide. Approximately 0.1-, 0 . 5 , and 1-gram samples of this material were irradiated, and the relat,ive specific activities of their phosphomolyhdate precipitates were compared. The results are shown in Table 11.
Table 11. Weight of Sample, Gram 0.1 0.5 1 .o
Attenuation of Neutron Flux in Sample Relative Specific .4ctivity 1.00 =!= 0.01 probable counting error 0 . 9 8 f 0 . 0 1 probable counting error 0 . 9 8 zt 0.01 probable counting error
1343 instance was only 4.9%, and this was taken into account in evaluation of the final results. Substantial differences were found, hoffever, between t h e average activity of the foils in different irradiation cans. I n one series, the samples were contained in four different cans located in the graphite stringer in positions 14-2-19 to 2 2 . The relative activities induced in the foils in these four positions were 1.00, 1.07, 1.20, and 1.24. A 24% difference in flux between the outermost and the innermost cans in the stringer was seen. Since these two cans (19 and 2 2 ) were separated by approximately 8 inches, a flus gradient of approximately 3% per inch existed. This introduces a possible maximum error of 2.5’% over the diameter of the can. This maximum error could only he realized, however. if the standard were entirely against one vial1 of the can and the sample were entirely against the wall diametrically opposite in the direction of increasing or decreasing flux. A similar exploration was made in other series where more than one irradiation can m-as employed. A similar flux distribution \vas observed, so a 2% possible error from this variable was considered reasonable throughout the investigation. Counting Procedure. The counting planchet consisted of an inch thick with a conicalaluminum disk 1 inch in diameter b y hottom hole 3 / * inch in diameter by 3/16 inch deep drilled in the renter of the top face. The conical bottom of the hole served to ($enter small precipitates, vhereas large precipitates could be packed uniformly over the entire 3/8-inchdiameter. The samples were contained in a 2-inch-thick lead shield for counting. The Geiger tube \?-as x halogen-filled, end-vindow type obtained from the Suclenr Instrument and Chemical Corp. A 45 mg. per sq. em. aluminum filter \vas generally interposed hetx-een the sample and tube to miriimize the error from small variations in sample thickness. The filter was not used, however, phosphorus because with samples that contained lew than 0.001 7, of the lo^ counting rate. A total of 10,000 counts or greater was taken on all samples, with the exception of the very low phosphorus alloys where 3000 counts or greater vats taken. Counts n-ere made over a 1- t o 2neek period, and values were taken from the best 14.3-day line through the points. Chemical Procedure. For one series of aluminum oxides, 1 gram of activated sample, 0.013 gram of nonradioactive monoammonium dihydrogen phosphate carrier, and 6 grams of boraxcarbonate flux mixture were added to a platinum crucible. T h e mixture was heated for a t least 30 minutes a t the highest temperature of a Meker blast burner, then the fusion was cooled and dissolved in dilute nitric acid. A standard was prepared by fusing a like mixture conbining 1 gram of inert alumina instead of the activated sample. T o the solution of this melt was added a known amount of the radioactive monoammonium dihydrogen phosphate standard. The acidity of the solutions was adjusted, and all the phosphorus was oxidized t o phosphate with potassium permanganate. The phosphorus was precipitated as phosphomolybdate, using ammonium molybdatecitrate solution. This precipitate was dried and weighed, then 0.1 gram was transferred t o a planchet and counted. After counting, the precipitate was dissolved and the phosphorus precipitated as magnesium ammonium phosphate, which was ignited t o magnesium pyrophosphate and counted in that form. For the remainder of the aluminum oxides, the phosphomolybdate precipitate was dissolved without drying and reprecipitated as a purification step. For the aluminum alloys, 2 grams of the metal, 0.25 gram of mercurous nitrate, and 0.013 gram of the carrier were dissolved in 100 nil. of 1 to 1 nitric acid. After removal of most of the acid tiy boiling, the solutions \yere gassed with hydrogen sulfide, filtered, and boiled. After adjusting the acidity, part were doubly precipitated as phosphomolybdate and the remainder only singly precipitated.
Self-attenuation effects are seen to be absent within the error of the measurement and, since the chemical history of the oxide samples was similar in all cases. this result, would be applicable to the others. Variation of Flux over Cans. Since the thermal flux distribution in the pile extremities was unknown, it was necessary, by determining the maximum variation of the flux over the dimension of a can, to establish the maximum error that could result if the flux over the stand:trd ivere different from that over the sample. From the construction of the pile, flux gradients of two sorts would be expected. Since the graphite stringers that contain the cans are inserted a t right angles betm-een the fuel rods, a variation in flux over the length of the can might be expected. Also, since the total flus falls off rapidly toward the edge of the structure! a variation from one can to another, or across the diameter of a single can, might be expected. The flux distrubution over the length of the can was established by employing a neutron-monitoring foil as a liner in each irradiation can. This consisted of 0.003-inch-thick, 3003 aluminum The several procedures above were employed with different foil that contained about, 1% manganese. The flux distribution was determined by measuring the activity of small, n-eighed series of samples to see which variation gave precipitates most squares cut from the foil a t the top, center, and bottom. radiochemically pure. The purity was established by decay Throughout the investigation there n-as a very slight indication rate determinations taken over one to two half lives. All prethat the neutron flus at, the top and the bottom of the can ~ a s cipitates, regardless of the procedure employed, appeared to he greater than that in the center. The maximum spread in any pure, with the except,ion of the phosphomolybdstes from one
ANALYTICAL CHEMISTRY
1344 aluminum oxide series. These contained traces of tungst.en as demonstrated by analysis of the composite decay curves. The 21.1-hour component of tungsten-187 was readily distinguished and an accurate difference curve could be plot,ted. The 73.2-day component of tungsten-185 was not perceptible so was disregarded. The tungsten contamination was not surprising, since tungsten coprecipitates with phosphorus as the insoluble ammonium phosphotungstate. In comparing t,he specific activity of an unknown precipitate with that of a standard, it is imperative that the counting geonietry be identical. It is particularly necessary in the case of a pure beta emitter that the thickness (on a weight basis) of the unknown and standard be identical. This requirement was more nearly met with the phosphomoll-bdate precipitate than the phosphate. The phosphomolybdate was very voluminous. It powderedreadily and could be compacted firmly into the counting planchet 11y gentle shaking. The pyrophosphate, on the other hand, WM very dense and refractory after ignition, 2nd was difficult to spread uniformly over the count.ing planchet. The double precipitation as the ammonium phosphoniolybdate evolved as the preferred procedure with the reservation that hugsten, if present, would follow the phosphorus. but could be corrected for with a high degree of accuracy. R E S U L T S Ah-D DISCUSSION
Some representative phosphorus determinations are shown in Table 111. These are given t o illustrate the precision and reproducibility of the method, and should not be taken as necessarily typical of the particular materials. The sets of values are duplicate or trjplicate determinations of the same sample with the exception of the values for 1100 sheet that were from two different lots of ~ I c J I - .
Table 111. Phosphorus Determinations b? Hadioactivation Analysis Material Description % P Material Description AI?&
Alcoa 1
2 3
0.000?0 0.00021 0.00013 0.00013 0.00023 0.0002:3
.\I
1100 extrusion I100 s!1eet
High-phosphoruq pig
Hi~ti-pho..ptiorus sheet 3
Refined phosphate rock
“,P 0,000053 0.00005B 0.000069 0.0000O7 0.002u 0.0029 0 002i
1
0 0012 0.0013
0.001:3 0.00 13
2
0.0019
0.0034 0.0034 0.0035
3
0.0019 0.0019 0.0021
was about 1 n1.e.v. Ttie neutron population of this energy in the extremities of the pile where the samples were irradiated is low, so this possible source of error was disregarded. Since a large, constant amount of carrier was added to each sample, weighing errors were constant regardless of the phosphorus content. The smallest quantity that v a s weighed was the 0.01 gram of carrier added to each sample, so weighing error:! were considered negligible. The maximum error from variation of the neutron flus over the dimensions of the can was 2%. -1possible error mas the coritamination of the precipitates by other radioactive elements with radiological properties similar to t,hose of phosphorus. The only such isotopes with half lives between 12 and 17 days are barium-140, cesium-136, osmium-191, vanadium-48. palladium-103, and tellurium-121. Only osmium need be considered as a possible contaminant. The remainder are excluded either because they do not result from neutrongamma reactions, or have small cross sections and are produced from isotopes of very low abundance. Osmium, with a 15.O-day half life, could be present to an appreciable exteiit and not be distinguished by. half-life determinations done. I t s cross section of 0.6 barn would give about threefold higher yield than phosphorus if both were present to the same extent. This would require, however, that osmium follow phosphorus exactly through the chemical separation. Though the details of osmium cheniistry are not \yell known, this latter is unlikely, since osnlium tetroxide would probably be lost from the boiling nitric acid solutions. SIoreover, osmium has a very weak beta (0.14 n1.e.v. compared to 1.7 for phosphorus) and has associated gamma emission. Such radiation vould be detected by absorption measurements, though admittedly this technique is not sufficiently sensitive to detect t r x e s of such contamillants. Considering also that osmium is % very unlike1~-contnminant in the materials under study, its contribution has been disregarded. The largest error in the results of the loxv-phosphorus s:mplcs undoubtedly was :E statistical error because of the low counting rate. This appeared to be about 5 % in the m r s t case. Everything considered, it nppenrs that the maximurn probable error in the determinations irae about 10% in the low-phosphorus samples and possibly 5% in the high-phosphorus samples. Some consideration has been given t o the ultimate practical sensitivity of the method with the materials described here. h s much as 10 grams of sample could be handled TTithout difficulty to give another order of magnitude in counting rate. Irradiation times of 2 half lives would iiicrease the activit’y by about 50%. Coiinting rates of about 25 counts per minute above background could be measured to perhaps 3% probable error if measurements were repeated over several days. This would give a “practical” sensitivity with the counting equipment described of about 0.0000005% phosphorus, determined with an error of about lo%, assuming no impurities came in a t this concentration that could not, be recognized and corrected for. LITERATURE CITED
A variable whose magnitude was not determined esperimentally was the production of radioactive phosphorus from sulfur by the (n,p) reaction. This is a common pile reaction for producing carrier-free phosphorus. The sulfur content of the alloy samples was negligible. I n the oxide samples, however, sulfur contamination was possible, and particularly in the material that was refined from phosphate rock by a sulfate process the sulfur content might have been appreciable. It is known that the reaction probahility of the sulfur-32 (n,p)phosphorus-32 reaction with low energy neutrons is extremely low. Klema and Hanson (4)found that the cross section leveled off a t about 0.3 barn a t about 4 m.e.v. It fell 08 very rapidly a t lower neutron energies. The threshold energy
(1)
Birnbaum, S . , and Walden, G. H., ,Jr., J . I t n . Chem. SOC.,60,
6G-70 (1938). (2) Hillebrand, W. F., Lundell, G. F., Bright, H. h.,and Hoffman, J. I., “.4pplied Inorganic Analysis.” 2nd ed., Wiley. New Y o r k , 1953. ( 3 ) Kitson, R. E . , and lIellon, 11. G . , .%SAL. CHEM.,16, 46&9 (1944). (4) Klema, E. D., and Hanson, A . O., Phys. Rev., 73, 106 (1948). (5) Leddicotte, G. W., and Reynolds, S. -I.,Sucieonics, 8 , S o . 3, 62-5 (1951). (6) RIuelhause, C. D., arid Thomas, G. E., I b i d . , 7, Xo. 1, 9-17, 59 (1950). (7) Taylor. T. I . . and Havens. W. K . , Jr., I b i d . . 6 , N o . 4, 54-66 (1950). RECEIVED for review February 11, 1955. .\ccepted April 1.5. 195.5. Presented a t the Division of .Inalytical Chemistry a t the 1‘26th Meeting of the A M E R I C A ~C E E M I C ASOCIETY, L N e w York, N. Y . , 1954.
