Table 111.
Comparison of Oxygen Results on Prepared YF3-Oxygen Standords
Oxygen Added as YlOl
YOF
Theor. Oxygen, % 0.100 0.130 0.169 0.210 0.100 0.161 0.217
Inert gas fusion 0.101 0.130 0.170 0.219 0,099 0.159 0.218
oxygen-containing salts were blended in a Pica blender mill for a total of 2 hours. The samples showed a high degree of homogeneity. A comparison of the results by the various techniques on these samples is shown in Table 111. Again, the KBrF4 method yields consistently low results. When plotted, the results on the yttrium oxide standards by the KBrFl method gave a straight-line relationship having the same slope as the line plotted from the vacuum and inert gas fusion results (Figure 2). This indicates that a constant portion of the oxygen in the samples is recovered by the KBrF4 method When the results
Av. Oxygen, % Vacuum distn. 0.102 0.139 0.165 0,100 0.099 0.169 0.207
KBrFi 0,058 0.085
0.130 0.161 0.015 0.027 0.064
on the yttrium oxyfluoride samples were plotted, however, the KBrF4 technique showed great insensitivity toward any oxygen in the sample (Figure 3). This evidence has led the authors to conclude that, in general, the low results reported by the KBrF4 technique are probably due to the insensitivity of the method toward oxyfluoride, and are a function of the history of the sample. ACKNOWLEDGMENT
The authors express their gratitude to V. A. Fassel and coworkers for
furnishing the vacuum fusion and emission spectroscopic data for this paper. LITERATURE CITED
(1) Abresch, K., Lemm, H., Arch. Eisenhiittenw. 30, 1 (1958). ( 2 ) Banks, C. V., et al., U. S. Atomic Energy Comm., Rept. No. TID45OO (July 1059). (3) Elbling, P.,Gouard, G. W., AKAL. CHEM.32, 1610 (1960). ( 4 ) Gibbs, 1).S., Svec, H. J., Harrington, R. E., Ind. Eng. Chem. 48,289 (1956). (5) Hanson, W. R., Mallett, M . W., Trzeciak, M. J., ANAL.CHEM.31, 1237 (1959):
(6j7Horrigan, V. XI., Fassel, V. A,, Goetzinger, J. W., Ibid., 32, 787 (1960). (7) Peterson, J. I., Xlclnick, F. A , , Steers, J. E., Jr., Ibid., 30, 1086 (1958). (8) Shanahan, C. E. A,, Cooke, F., J . I r o n Steel Znst. (London)188, 138 (1958). (9) Singer, L., IND.ENG.CHEM.,ANAL. ED.12, 127 (1940). (10) Smiley, W. G., ANAL.CHEM.27, 1098 (1955). RECEIVEDfor review March 17, 1960. Accepted June 13, 1960. Pittsburgh Conference on Analytical Chemistry and Applied Spectroscopy, Pittsburgh, Pa., February 29 to March 4, 1960. Contribution No. 877, Ames Laboratory of the U. S. Atomic Energy Commission.
Determination of Oxygen in Beryllium Metal by the Inert Gas Fusion Method SILVE KALLMANN and FRED COLLIER Research Division, ledoux and Co., Teaneck,
b In an inert gas fusion procedure, beryllium oxide i s made to react with carbon in a graphite crucible yielding carbon monoxide which, after oxidation to carbon dioxide, i s measured conductometrically. The work was carried out with a Leco inert gas fusion apparatus which i s slightly modified to accommodate larger sample weights. To provide an intimate contact of the molten sample with the graphite and to minimize the evaporation of beryllium, with the danger of subsequent gettering, nickel i s used as a flux. The method i s rapid and simple enough to permit the use of untrained personnel.
M
ETHODS for the determination of oxygen in beryllium have recently been reviewed and evaluated by Bradshaw (4).
ESTABLISHED PROCEDURES
Vacuum Fusion. Gregory and Mapper (10) recommend the vacuum 1616
o
ANALYTICAL CHEMISTRY
N. 1.
fusion approach using 50 mg. or less of sample, a platinum bath, and alviays maintaining a licryllium-platinum ratio of 1 t o 50. Booth, Bryant, and Parker ( S ) , Iioivever, were unable to duplicatc the findings of Gregory and Mapper except in a limited number of instances. They obtained higher and more consistent results by adding platinum and poirdercd graphite before each sample. Runner (16) rcports quantitative recoveries of oxygen by using a platinum flux technique, a comparativrly low tcmperature, and a beryllium-platinum ratio of 1 to 10. Volatilization. As beryllium oxide resists the attack of chlorine or hydrogen cliloride, volatilization of beryllium chloride and trcatnient of the residual beryllium oxide have been madc the basis of methods reconimended by R,ynasicn.icz ( l 7 ) , White and Burke (B), anti Osinond and Smales (14). Thcse methods are essentially adaptations of methods successfully used for the dctcmiination of oxygen in
zirconium or titanium involving either chlorine (5) or hydrogen chloride (16, 20). Because of the lower boiling point of zirconium tetrachloride (300" C.) compared to beryllium chloride (520' C.), the halogenation procedures appear less attract'ive for beryllium than for zirconium. Differential Solubility. Several mcthods have been suggcsted ivhich arc based on the solubility of beryllium metal and the insolubility of beryllium oxide in various reagents, such as dilute hydrochloric acid (I?'), copper sulfate ( I d ) , and bromine-methanol (6). Again, thew methods are essentially adaptations of earlier procedures designed for the determination of. oxygen in aluminum (8, E?)and have becn evaluated by Eberle and Lerner (6). Other procedures which have becn considered for lhc determination of oxygen in beryliium are based on teehniqurs involving isotopic dilution (11), radio act.ivatior! ( 1 4 ) , and hydrogen evolution (2).
Inert Gas Fusion. T o t h e best of the authors' knowledge, this technique which in principle goes back t o t h e work by Singer (18),and was markedly improved by Smiley (19), has not yet been applied t o beryllium metal, although Uradshaw (4)lists it among the techniques which warrant further investigation. In this method, oxygen is extracted from the metal or alloy by a fusion resulting from induction heating in a graphite crucible in a stream of argon. The oxygen, through the interaction with the graphite, forms carbon monoxide, which is oxidized to carbon dioxide by iodine pentoxide. The more complicated freeze out manometric technique previously used by Smiley (19) has been replaced by a conductometric procedure capable of accurately measuring carbon dioxide. This has been made possible through the introduction of the Leeo oxygen analyzer, a combination of a 4.5-kw. induction furnace operated at 4-Mc. frequency, closely resembling conventional vacuum fusion equipment, and the previously widely used conductometric apparatus. Thus, the carbon dioxide resulting from the oxidation of carbon monoxide by iodine pentoxide is absorbed by a dilute barium hydroxide solution. The amount of oxygen in the sample is determined by measuring the change in conductance of the absorbing solution and comparing this change against a standard curve. The basic reactions involved are as follon.s : hI,O, 5
+ C +CO + M (or M,C,)
co + 1 2 0 6 --+- 5 co, + I,
+ CO, --+- BaCOs + H,O
Ba(OI-I)*
This method has been successfully applied in this laboratory during the last few years to the determination of oxygen in a number of metals (among them, titanium, zirconium, hafnium, niobiuni, tantalum, tungsten, molybdenum, chromium, nickel, cobalt, iron, vanadium, uranium, boron, copper) and has been the subject of studies of various technical groups (1, 7 , 9, IS, 21) with platinum ranking foremost as flux material. DEVELOPMENT OF INERT GAS FUSION PROCEDURE APPLICABLE TO BERYLLIUM METAL
The only data found in the literature of interest for the development of an inert gas fusion procedure for oxygen in beryllium are based on experiments involving vacuum fusion. It was shown (IO) that, following the reaction BeOd,d f Caoild
cog..
-k Bei,,,,d
the pressure of carbon monoxide in equilibrium with beryllium oxide and gmphitcat 1560' C. is 3.2 X 10-3mni. of mcxury and that the rate of reduction of h~r!~lliunio d e by carbon alone is too
slow for the vacuum fusion method. The use of a molten platinum bath speeds up the reaction by acting as a medium or solvent for bringing the beryllium oxide and graphite into intimate contact. However, a t the temperature necessary for the reduction of beryllium oxide, the high vapor pressure of the metal (about 10 mm.) leads to the evaporation of the latter with the inherent danger of chemisorption or gettering. As a starting point, the following physical and chemical properties derived from previous vacuum fusion experiments were therefore taken into consideration. The melting point of beryllium is 1350' 20' C., its boiling point is 2970' C., and the vapor pressure a t 1950' C. is almost atmosphere. At this temperature, the graphite reduction of refractory beryllium oxide is very slow. A platinum bath does not sufficiently increase the rate of reaction between the beryllium oxide and the graphite to allow the operations to be carried out a t a temperature where the vapor pressure of the beryllium metal could be disregarded. Preliminary tests carried out in this laboratory indicated that platinum was even less suited for the inert gas fusion procedure than for the vacuum fusion technique. To achieve an optimum reaction rate between beryllium oxide and graphite, temperatures of about 2600" C. are required which lead t o excessive evaporation of beryllium and prohibitive gettering. [Platinum metal is the flux chosen for determining oxygen in titanium, niobium, and zirccnium (1, 7, 9, bl).] The next metal investigated was copper. Its potential use was indicated by the availability of a commercial copper-beryllium master alloy containing up to 3.5% of beryllium. When portions of this alloy with a known oxygen content were introduced into the hot graphite crucible, reaction of the oxide with graphite was immediate and without any apparent gettering. On the other hand, when beryllium was introduced into a bath of molten copper, difficulties were encountered, inasmuch as the first determination usually yielded the full oxygen content while subsequent runs indicated an increasing loss of oxygen. Apparently, beryllium metal vaporizes before it has a chance to dissolve in the molten copper and before the oxide reacts with the graphite. When the copper was introduced into the crucible simultaneously with the beryllium, thus acting as a flux, the precision of the results improved markedly. It was frequently possible to obtain as many as six determinations before the oxygen recovery decreased significantly. Copper, however, is unable to conibinc with more than 3.5 weight % of berylliuni, thus requiring
*
about 30 parts of copper for each part of beryllium. This makes copper virtually useless for beryllium metal with an oxygen content below 0.05%, as a minimum of 0.25 gram of beryllium and 7.5 grams of copper would be required to provide sufficient sensitivity for the conductometric measurement. The search for a better flux material finally led to nickel. Although other metals are superior to nickel in regard to their alloying affinity toward beryllium (beryllium and zirconium, for instance, form well-defined alloys with a ratio of about 1 to 1) they could not be obtained with a sufficiently low oxygen content to be useful as a flux material. Some of the potential alloying constituents even require platinum to release their own oxygen. Nickel looks attractive as a flux for several reasons. It can be obtained with a very low oxygen content ( < l o p.p.m.), its melting point of 1455' C. is fairly close to that of beryllium (1350' C,), and its boiling point of 2900' C. is virtually the same as that of beryllium (2970" C.). As a result of the fusion, a nickel-beryllium solution containing up to 15% of beryllium is formed, which effectively reduces the volatilization of the beryllium. It was found advantageous to continue using copper for w r a p ping the beryllium because many samples are submitted for analysis in powder or flake form which is unsuited for direct introduction into the graphite crucible. Copper foil u-ith a very low oxygen content ( < l o p.p.m.) is used and the resulting Monel-like nickel alloy can dissolve up to 17% of beryllium. EQUIPMENT
The only commercially available apparatus is that manufactured by the Laboratory Equipment Corp. The instruction manual (12) should be consulted for a detailed description of the equipment and for a suggested procedure suited for the determination of oxygen in titanium, niobium, and zirconium based on a platinum flux technique. The main features of the equipment are outlined in Figure 1. The following additional information is of interest: The power source for this unit is a Leeo 4.5-kw., 3.6-Me. radiofrequency generator necessitating the use of carbon black as insulating material. A report from another installation using homemade equipment ( 7 ) indicates that a different power source, such as the Lepel 2.5kw., 450-kc. radiofrequency generator, would allow the use of graphite insulation in the crucible assembly which is preferable to carbon black. The latter is more difficult to blank out and has a tendency to be blown into the graphite crucible whcn the flow of argon is increased. The sample-loading stopcock supplied with the original equipment is not well suited for the determination of oxygen in beryllium. It is apparently VOL. 32, N O . 12, NOVEMBER 1960
1617
I
I
FURNACE
Figure 1. 1.
a.
Hzsoc
4. 5. 6.
7.
Catalyst heater 10 grams of Ti, Zr, or U Quick disconnect Pressure relief volve
designed to accomniodate 0.1 to 0.25 gram of titanium and zirconium and about 0.5 gram of uranium. The 9-mm. bore hole of ' t h e stopcock holds less than 0.1 gram of solid beryllium metal, thus making it impossible to analyze samples below 0.05% oxygen. Therefore, the original stopcock was replaced by a larger one (20-mm. bore hole) made of Teflon and requiring no lubrication. Also, the ball joint connecting the stopcock was eliminated, thus ensuring greater rigidity of the loading device. The size of the funnel was increased to allow the passage of larger sample weights and the funnel itself was suspended in the ball joint of a larger furnace tube. Figure 2 shows a combined loading lock and storage tube which is suitable for introdu,cing a number of samples of wrious weights into the graphite crucibles. This, or a similar device ( 7 ) , is necessary when the oxygen content is 0.02yoor below, thus requiring a sample of beryllium metal which cannot be accommodated by any stopcock-type loading device. Graphite crucibles, 2 inches long with a center hoic of S/'*-inch diameter, prepared from cylindrical 'isinch rods were used and found superior to the onrs suppiied by the manufacturer. Thrj, accommodate larger quantities of samplrs 2nd eflectively prevent the carbon black from being b!own into the crucible. 1618
ANALYTICAL CHEMISTRY
I
I/--_
I
I N DUCT1O N
CONDUCTOMETRIC A NALY Z E R
2. Ascarite and Mg(C104h 3. Flowmeter
I
9. 10. 11. 12. 13. 14.
PURIFYING
TRAIN
Inert gas fusion apparatus
Ring seal Pedestal Thimble Carbon black Carbon crucible Reaction tube Funnel
PROCEDURE
Consult the manufacturer's manual for assembly instruct,ions and also for directions for the preparation of the barium hydroxide solution. See also Figures 1 and 2. Outgassing and Blank Rate of Apparatus. Regulate the argon flow on the purifying train to 0.5 liter per minute by use of the needle valve, leaving the loading device in the open position. T o prevent oxidation of the metal in the argon purification train, allow argon to flow for 5 minutes before turning on the catalyst heat'er. Turn on the water supply, furnace filament, and the high-voltage supply to the generator. Advance the t,einperature on the generator, heating the crucible assembly by 100 ma. a t 1-minute intcrvals until the maximum heat is obtained. Maintain this temperature of approximately 2700" C. for 10 minutes, then turn off the high voltage and cool the crucible. Obtain the blank rate of the apparatus by turning on the highvolt,age switch and adjusting temperature to 2600" C. Maintain this temperature setting for 1.5 minutes and follow with R 5.5-minutc flubh of the syst'pm by th:: argon. R,cpcat. t'his procedure iiiitil a blank of almost 1.5 ohms for the 7-minute q ~ l ise obtained. Blank Rate of Flux. Transfrr U.4 gram of copper foil (1.5 inches 2 mil) and 1 grani of nickel ( I j - n d sheet)
15. Metal clip 16. Clamp 17. Loading stopcock 18. Ascarite 19. Catalyst heater 20. Iodine pentoxide 21. Sodium thiosulfate
(both previously cleaned with liydrochloric acid then rinsed in succession with reagent grade 1,1,2-trichloroethylene and acetone) to the saniplcloading device. Repeat the cycle previously described. The flux blank should not exceed 0.5 ohm, particularly when samples of a low oxygen content are analyzed. Preparation of Standard Curve. Prepare copper foil packages (100 mg.) containing either silver oxide or lead oxide (PbO) with an oxygen content of betwcen 100 and 400 mg. Introduce the foil packages individually into the sample-loading stopcock. If the multiple sample-loading device is used, a number of packages can be introduced a t the same time. Adjust the stopcock in such a way that argon will flush out any carbon dioxide and oxygen from the loading device. Adjust t,he variable temperature control knob to bring the plate current to 700 mLi. and allow the graphite crucible to stabilize in temperature (about 2600" C.). Turn the loading stopcock to lock position t>oallow argon to flov through the iodine pentoside, sodium thiosulfatr, and the conductometric cell. Close tlie inlet stopcock on thc conductomrt-ic unit. Fill the conductometric measuring cell with barium hydroxide solution, Allow the lines on the oscilloscope t o become ,$able. Drop the sample from tlie load1.n: stopcock into the crucible
and start a stopwatch. Alternatively, m e a magnetic sample manipulator if the multiple sample-loading device (Figure 2) is used. The oscilloscope iines should not spread more than 0.5 ohm in the first 30 seconds. Oxygen will be detected in 40 to 60 seconds as noted by the movement of the lines on the oscilloscope. A t exactly 1.5 minutes from the time the sample was dropped, shut off the furnace. Flush for exactly 5.5 minutes, balance the bridge, and take the reading of the resistance duodial. Repeat for each of the standards. Correct the resistance reading in ohms of the standards for the apparatus blank and plot the corrected readings us. the known oxygen of the standards (in micrograms) on linear coordinate paper. The graph should be a straight line up to 400 fig. Check the curve daily with a t least two calibration standards. Preparation and Analysis of SamMaintain all ples. Precaution: health physics requirements for beryllium. T h e following additional precaution should always be observed. When opening the exit stopcock t o allow the escape of argon, attach a glass tube in such a way as to permit the gas to pass through water. To analyze solid samples, cut off pieces with a hacksaw. File or sand the surface and degrease with 1,1,2-trichloroethylene and rinse in acetone. Allow to dry, then weigh. No sample preparation is possible for beryllium powder or flake. ITrap 0.25 gram of beryllium into a 1.5-inch sheet of copper foil if the oxygen cuntcnt is between 0.01 and O.lyG.Use 0.1-gram sample for an oxygen content from 0.1 to 0.4%, and 0.05 gram for the 1% oxygen range. To this, add nickel to the extent of ten times the weight of tlie beryllium and introduce it into the sample-loading device. If a 0.25-gram sample is used, the multiple sampleloading device (Figure 2) must be used. Adjust the stopcock in such a way that argon will flush out any carbon dioxide arid oxygen from the loading device. From thi? point on, follow the instructions given under Preparntion of Standard C'1m-c. Calculations: P.p.m. oxyger! = A / B where A
micrograms of oxygen read from curve and corrected for apparatus and flux bhk B = ?ample weight in grams =
E=+-
mples as such were U t e standards were nrcpured by mixing varying proportions c?;iipl~.? I and 11. Saniple I was a !!~~.m powdc,- bi-ith :a oxy5~11conf i .!?:%>, as rktrrmined bv the
Results Obtained by the Proposed Procedure
Oxygen, % Present Found
w Figure 2. Modified sampleloading device
indicated by the facts that a 2-gram sample yielded no weighable residue by the bromine-methanol method and that a 0.25-gram sample treated by the proposed procedure did not produce more carbon dioxide than the combined apparatus and flux blank. DISCUSSION
1.05 1.05 1.05 1.05 0.30 0.30 0.30 0.30 0.12 0.12 0.12 0.12 0.78" 0.78"
0.97 0.99 1.02 1.10 0 33 0 32 0.27 0.30 0.10 0.13 0.09 0.11 0.80 0.76
Oxygen, % Present Found 0.92. 0.97a 0.05 0.05 0.05 0.05 0.025 0.025 0.87 0.87 0.62 0.62 0.24 0.24
0.92 0.97 0.04 0.052 0.038 0.060 0.029 0.018 0.85 0.90 0.65 0.58 0.25 0.25
Beryllium-zirconium alloy. CompariEon with HC1 volatilization method. 0
The results in Table I would indicate that the method is valid within the range tested. It is difficult to extend the method to a much lower oxygen level because the weight requirements both for sample and flux are such that the graphite crucible could not hold more than two runs. The results in Table I do not show how many determinations failed when the furnace was not operating a t ful! efficiency or when, for other often unexplainable reasons, results suddenly dropped off. [It is important that the power supply is 230 volts or the required temperature range (about 2600' C.) cannot be obtained. With lower temperatures, the reaction between oxide and graphite is incomrlete.] Failure is indicated when duplirate determinations disagree widely. When this happrns, it is a good practice to introduce immediately a standard of lrad or silver oxide. If the standarcl also fails, it is time to change the graphite crucible and replace the carbon black, .4s many as ten determinations have been carried out in the same graphite crucible and it has also happened that the crucible had to be changed after three or four determination. All in all, it is felt that with proper safeguards the proposed method is supericr to other procedures in the range under investigation. ACKNOWLEDGMENT
The constructive criticism of Charies
D. Houston of WADC is gratefully RESULTS
Table I.
-
acknowledged. LITERATURE CITED
il) Am, Soc. Testing Materials, Rept. of Oxygen Subgroup Niobium Task Force, Division hl, Committee E-3 ( J u n e 19m; --_-,
'2) Bergholz, ,T. A , LT. Resetrch Kat!. Bur. Slandards48,201-5(1952,. 13) Booth, E . Bryant, F I., Parkpi, A., dnrlldst 8 2 , 50 (1957). 1'4) Bradshaw, W. G., Lockheed Aircraft S o ~ p .Rept. . LMSU-2312 (March 1958). ( 5 ) Corbett, J. -4 dnalyst 76, 652 (1951). ~
(6) Eberle, A. R., Lerner, M. W., USAEC Rept. NBL-133 (1957). (7) Elbling. P.. Goward. G. W., ANAL. CHEM.32, i6io (1960): (8) Fischer, J., Rechtel, H., Erzb, Z., Met. Ab&. 20. 194 (1952). (9)p w a r d , G: W:, .Jacobs, R. M., Restinghouse Bettis Plant Rept WAPD-M (GLA)-790 (1959). (10) Gregory, d. N.,Mapper, D., Analyst 80,230 (1955). (11) Kirshenbaum, A. D.,Grosse, A. V., Trans. Am. Soc. Metals 45, 758-67 11953). (12)Laboratory Equipment Corp., Instruction Manual for Operation of Leco Oxygen Analyzer No. 534-300 (1958). 113).Odess, S. L., Pratt & Whitney Co., Aircraft Piogress Rept. Task Force on Analysis of Interstitials in Chromium Metal (September 1959). (14) Osmond, R. G., Smales, A. A., Anal. Chim.Acta 10,117(1954). (15) Read, E. B., Zopatti, T. P., MIT Rept. 1038, U. S. Atomic Energy Comm., AECD-2798 (February 1950). (16) Runner, M., Chemical Research Services, Addison, Ill., private communication, 1959. (17) Rynasiewicz, J., U. S. Atomic Energy Comm., Rept. AECD-3720 (1951,declassified 1955). (18) Singer, L., IND. ENG.CHEM.,ANAL. ED. 12,127 (1940). (19) Sniiley, W. G., ANAL. CHEM.27, 1098-102 11955). (20) Tighe, J. J., Gerdes, -4. F., Center, E. J.; Mallett, If.W., Battelle Memorial Institute, Rept. BMI-799 11952). (21) U. S. Army, Ordnance Corps, Metallurgical Advisory Committee on Titanium, Rept. by Task Force on Oxygen in Titanium, 1959. 122) Wcmer, O., 2. anal. Chem. 121, 385 11941). ( 2 3 ) White, D. W., Jr., Burke, J. E., "The Metal Beryllium," American Society for Metals, Cleveland, Ohio, 1955.
RECEiVED far review hkrch 10, 1960. pcepted July 18,1960. Work supported oy the United States Air Force under Contract A.F. 33 i 11:6)-6281 monitored by the Materiala Laboratory, Wright Air Development Center, Wright-Patterson Air F o r x Base, Ohio. VOL. 32, NO. 12, NOVEMBER I900
1619
