possible that this is due to some side reaction dependent upon the reportedly (IO) different molecular species which exist in 99% Hi304 as opposed to more dilute Ha04 solutions. Within the accuracy of the temperature control these data are in agreement with those reported by Hellin and Jungers (3). They report that rate data for the reaction can be correlated by
+
(CJ%)pSOt
+ HISO, * 2CtHrHSOd
(4)
assuming a second-order reaction equation:
The reaction rate was shown by them to be strongly dependent upon the acid concentration and in fact they made no measurements on solutions with a n initial (C2H6)$04 concentration greater than 50% because of the extremely slow
reaction rates. That the reaction rates are very slow a t low acid concentration is borne out by our work. Only the 24.9 initial mole % (C?H&S04 sample approached equilibrium during the time allowed for reaction; the calculated concentration equilibrium constant for this sample compared favorably with that obtained from the data of Hellin and Jungers. Data for the 65.4 initial mole % (C?H,)$O( sample were taken over an extended period of time and reaction velocity constants were calculated from these data using the above secondorder equation; the data were well correlated by this model and compared favorably with Hellin and Jungers’ data as extrapolated. ACKNOWLEDGMENT
Acknowledgment is made to the donors of the Petroleum Research Fund, administered by the American Chemical Society, for support of this research.
LITERATURE CITED
(1) Breslow, D. S., Hough, R. S., Fairclough, J. T., J. Am. Chem. SOC.76,
5361 (1954).
(2) Ellis, C., “Chemistry of Petroleum
Derivatives,” pp. 321-33, Chemical Catalog, New York 1934. (3) Hellin, M.,Jungers, J. C., Bull. soc.
chim. 1956,386. (4) Plant, S. G. P., Sidgn-ick, N. V., J . SOC. Chem. Ind.40, 148T (1921). ( 5 ) Schroeder, W. C., IND.ENQ.CHEM.;
ANAL.ED. 5,403 (1933). (6) Sheen, R. T., Kahler, H. I,., Zbid., 8, 127 (1936). ( 7 ) Sherrick, P. H., et al., “Manual of Chemical Oscillometry,” E. H. Sargent & Co., Chicago, 1954. (8) Suter, C. M.,“Organic Chemietry of Sulfur,” pp. 8-28, Wiley, New York, 1944. (9) U. S. Atomic Energy Commission, “Master Analytical Manual,” TID7015 (Section 1) (1958). (10) Wyatt, P. A. H., Trans. Faraday SOC.56,490(1960).
RECEIVEDfor review April 4, 1961. Accepted August 14, 1961.
Determination of Oxygen by the Inert Gas Diffusion Method Using Graphite Capsules EDGAR J. BECK and FORREST E. CLARK Parma Research Center, Union Carbide Corp., Parma 30, Ohio
b An inert gas diffusion method i s described for the determination of total oxygen in samples where volatile reactions may be encountered. A graphite encapsulation technique i s used to contain the sample to achieve two important objectives: to prolong the period of time that the sample i s exposed to carbon during decomposition or reaction at high temperatures, and to furnish a miniature reaction chamber in which thin films of metal or carbides are deposited, thus maintaining the walls of the reaction crucible in an active and clean state. The samplecontaining graphite capsule is prepared from a preformed graphite spectroscopic electrode cup and sealed with a graphite plug. Some oxides, inorganic compounds, and metals were analyzed with satisfactory precision and accuracy. Samples of materials which would decompose violently and rapidly when contained in tin capsules react much more smoothly when dropped in this graphite capsule. The gaseous products liberated from the sample b y the high temperature diffuse through the graphite walls of the capsule, the oxygen from the sample being converted to carbon monoxide in this process. The carbon
monoxide i s carried b y a sweep gas of purified argon through a puriflcotion train and then through an oxidizing reagent where it is converted to carbon dioxide. This gas is measured conductometrically. Samples containing oxygen in the range of 0.05 to 40.0% were successfully analyzed b y this inert gas diffusion method.
S
the introduction of the inert gas fusion method by Singer (8) and subsequent modifications by Smiley (9), various procedures for the determination of oxygen have been developed (1, 4, 6). By making a few modifications, the authors have been able to adapt a wide variety of samples to the commercially available equipment developed by Bennet ( 2 ) and coworkers. Many samples have shown uncontrollable volatilization properties when heated to the high temperatures encountered in this inert gas diffusion method. Some materials, when dropped in the conventional tin capsule, have been partially lost during this reaction, while others, because of their vaporization, have formed thin films over the inner wall of the graphite reaction crucible. This thin film either reduces the INCE
vapor pressure of the carbon in the reaction crucible or effectively reduces the surface area exposed to the oxygen so that i t cannot react to completion to form carbon monoxide. Successive “drops” of aliquots of a material that tends to form a film show large decreases in the apparent oxygen content based on the conductometric measurement. This phenomenon is demonstrated by analyzing several aliquots of a material that tends to form this thin film on the inner surface of the reaction crucible. The apparent oxygen content of each aliquot, wrapped in the conventional tin container, is appreciably lower than the value attained for the preceding aliquot. We have found that the containment of this volatile reaction is best controlled in a graphite capsule. The capsule, being permeable, allows a controlled gas diffusion through its walls. Mechanical loss is prevented. Graphite’s low thermal expansion, in combination with its high thermal conductivity and strength a t high temperatures, makes it highly resistant to thermal shock. Various vaporizing metals can form carbides inside the capsule rather than forming a film inside the graljhite crucible. I n addition to holdVOL. 33, NO. 12, NOVEMBER 1961
* 1767
from a previously prepared standard curve. EXPERIMENTAL
Apparatus
and
Reagents.
The
LECO oxygen analyzer No. 534-300,
Figure 1. Graphite capsule and:plug with die and tap
ing the sample in the high temperature zone for a slightly longer period of time than the tin capsule, the graphite capsule serves as a miniature reaction chamber in which the solid products of decomposition or reaction are maintained. This permits samples to be analyzed until the reaction crucible is full of spent graphite capsules. These spent capsules are removed from the crucible with tweezers, the unit is reassembled and degassed to a n acceptable blank, and the analyses of samples can be continued. In this inert gas diffusion method, argon, a t a flow rate of 0.4 liter per minute, passes through a purifying train and then through heated titanium sponge or similar material which frees the gas of oxygen prior to entering the reaction chamber. The sample under test, contained within a graphite capsule (Figure l ) , is placed in a loading stopcock, from which it is dropped through a funnel into an inductively heated graphite crucible. This crucible is insulated by a surrounding carbon black packing which is held within a flared quartz thimble (Figure 2). The sample is decomposed by the high temperature within the reaction crucible. The evolved oxygen diffuses through the hot carbon capsule and is converted to CO. Argon then carries this CO from the reaction tube into Ascarite, where acid gases (if any) released by the sample are trapped. The CO next passes through 1 2 0 s , where the CO is converted to COz by the reaction: 120b
+ 5 co 160'
.--f
The resultiiig thiosulfate: 2 NaZS201
+L
5 COa
+ In
(1)
Iz is absorbed in sodium +
2 NaI
+ Na&Oe
(2)
The COz is swept by the argon carrier gas into a Ba(0H)g solution in the conductometric analyzer where a change in conductivity takes place: Ba(OH)2
+ COZ
+
BaCOa 4
+ HzO
(3)
This change in conductivity is measured in ohms, which are correlated against actual oxygen content in the sample 1768
ANALYTICAL CHEMISTRY
which consists of a n induction furnace and a conductometric analyzer, was adapted for use with this method of analysis (6). -4 150-ohm variable oneturn resistor was added to the conductometric unit to compensate for the difference in resistance when different concentrations of barium hydroxide were used. This potentiometer was placed between lead wires connected t o the plug-in on the analyzer and the platinized indicating electrodes. The platinized indicating electrvde tips need never be respaced to accommodate these changing resistances, thereby eliminating a possible mechanical breakdown of platinized surfaces. Temperatures were determined with a Leeds and Northrup optical pyrometer. Figures 1 and 2 show the graphite capsule and redesigned quartz ware. The graphite capsule used to hold the sample was prepared from a Kational spectroscopic preformed graphite electrode (Catalog No. L-4024). The electrode cup was threaded on the inside with a n ID-32-NF GH3 tap. A portion of the section on the end opposite the cup was removed. as indicated in Figure 1. -4plug for the cup was prepared from a National spectroscopic graphite 3//16 X 12 inch solid rod (Catalog KO. L-3086). One end of the rod was threaded with a 10-32NF button die. After the sample is placed in the cup of the capsule, the threaded end of the rod is screwed into the cup, and the portion of the rod protruding from the capsule cup cut off with a small saw. Oxygen blanks of these sample capsules are only slightly higher than tin capsules, are reproducible if handled properly, and yield less than 10 pg. of oxygen per capsule. The reaction crucible used wvas a National 7/* x 31/* inch graphite crucible, which is longer than those originally supplied by the manufacturer but of the same grade. A flared lip quartz thimble was prepared as shown in Figure 2. The electrical discharge over the edge of nonflared thimbles pitted the edges. By utilization of the longer quartz thimble with the 45" flared lip made to accommodate the new graphite reaction crucible, the electrical discharge over the edge was minimized. Undesirable high blanks, which were noted under conditions when electrical discharge occurred, were substantially reduced. Approximately 7 grams of reagent grade barium hydroxide were dissolved per liter of distilled Con-free water according to the procedure in the instruction manual (6). This solution was used where the oxygen content of the saiiiple was greater than 3%. (Samples as high as 60% oxygen have been determined using this solution.) A solution containing about 1 gram of barium hydroxide per liter of COZ-
M
PEDESTAL
Figure 2. Redesigned quartz ware and crucible
free water (6) n-as used for samples containing less than 3% oxygen. Degassing and Blanking. The system was freed of adsorbed oxygen, in most cases, by gradually heating the crucible t o 2600" to 2800" C. while maintaining a flow of 0.4 liter per minute of argon, and holding the temperature for 20 t o 30 minutes. A 1/4-inch elevation of the crucible over the rim of the flared thimble was the most favorable position when packing the crucible in the carbon black. This keeps carbon black from falling into the crucible. The temperature during degassing 17 as gradually increased to the 2800" C. level by slo.cvly advancing the variable temperature control. This prevents the possible rapid escape of gases which tend to blow carbon black over the inside surfaces of the quartz reaction chamber. After a brief cooling period, the loading stopcock was closed and the stopcock of the conductometric inlet for "products of reaction" was simultaneously opened. The conductometric measuring cell was filled with barium hydroxide solution and the induction furnace adjusted so that the temperature of the graphite reaction crucible would reniain constant at 2050" C. within the 2minute preheating period. The oscilloscope of the conductometric unit was zeroed. The furnace temperature was maintained for exactly 3 additional minutes and then turned off, allowing 2 minutes for cooling, making a total of exactly 5 minutes for the total reaction time. If the readings from the decade knob on the conductometric unit did not exceed 2.0 ohms for 1 gram of Ba(OH)* per liter of solution or 0.2 ohm for 7 grams of Ba(OH)2 per liter of solution, degassing was considered adequate. Calibration. Various standards were used in setting up a curve which showed the relation of the change in conductance (ohms) us. grams of oxygen. T h e selection of a standard oxygen sample was dependent upon the strength of barium hydroxide solution used in the analysis. Silver oxide ( E ) , with a n oxygen content between 100 and 400 rg., placed in either the manufacturer's recommended tin capsule or the spectroscopic graphite capsule, was used with 1 gram of barium hydroxide per liter of solution. Potassium acid phthalate solution (prepared by dissolving dried primary
standard grade reagent) aliquoted with a microsyringe into tin capsules with subsequent drying can also be used for calibrating this same 1 gram of barium hydroxide per liter of solution. This method ( 7 ) saves time and error due to weighing, and eliminates the need of a microbalance. Reagent grade, dried silica (about 240 mesh, floated powder), placed in graphite capsules, was used in calibrating with the 7 grams of Ba(0H)Z per liter of solution. A reproducible standard curve was obtained for evolved oxygen quantities between 1.0 and 5.0 mg. Following the same procedure as used in blanking, the standard was dropped from the loading stopcock into the reaction chamber crucible after the preliminary 2-minute preheat. The induction furnace was allowed to heat the graphite crucible for 3 additional minutes, with a final %minute cooling flush. Although the reaction due to the induction heat was almost instantaneous when a tin capsule was used, it was not so in a graphite capsule, as heat penetrstion was slowed by the graphite capsule walls. Three minutes !+ere sufficient for this 2050" C. reaction temperature to effect fusion and diffusion of gas through the capsules. Readings from the decade control were taken after this 5-minute run. All standards were determined in the same manner. An occasional blank was run to amire complete recovery. -411 blank readings should be less than 2.0 ohms for 1 gram of Ra(0H)z per liter, less than 0.2 ohm for 7 grams of Ba(OH)2 per liter. Standard curves are plotted for each concentration of barium hydroxide solution. An average blank value (in ohms) is established for each concentration of solution and subtracted from each standard reading; the standard curve is plotted from ohms (corrected for blank) against grams of oxygen in each standard. Procedure. The inside dimensions of the graphite capsules, which are 3 X 5 mm., limit the sample size. Powders are ideally suited for placing in these capsules. Uranium Oxide (UsOs, National Bureau of Standards sample 950, Table I). This oxide serves as a favorable example for use in this procedure. Five capsules were threaded and placed in a drilled graphite sample holder. One of the capsules was transferred to a 1-ml. beaker on the balance and weighed. A Mettler semimicro one-pan Gram-atic balance was found satisfactory for use in these weighings. Stoichiometric U308 with a theoretical oxygen content of 15.2Yc would require a sample size of 12 to 15 mg. This approximate quantity of dried uranium oxide was placed directly in the weighed, empty capsule cup and reweighed. Tweezers were used exclusively for manipulating the capsule when filling with the sample. The threaded graphite rod was screwed into the capsule cup, and the extending portion cut off with a jeweler's saw. A portion of the section on the end opposite the cup
Table I.
Precision Studies on Several Inorganic Materials Using Graphite Capsules
Sample A1
Ti02
Sample No. 9020
0.08 0.08 0.08 34.2 34.8 34.2 1.3 1.4
2400
+ Ti
2400
AIN
9360
2400
UaOs
NBS 950 before 900" C.
2400
1.L
U3Of
heat NBS 950 after 900" C. heat, 1 hour
2400
(187) 9492
2400
Uranium oxide
0
14.1 14.2 14.8 15.0 15.1 15.0 14.8 16.3 16.4 16.4 16.4
15.0 15.2 15.0 15.1 14.9
Standard deviation =?=0.12ojO. Coefficient of variation 0.8%. Table II.
Sample Titanium oxides UOn Ag,OsN03 Rare earth monosulfide US08
Si02 B BzOs Mn-Ag oxide
+
Oxygen Studies of Various Inorganic Materials
Specific Sample NO.
Range0
... ...
30-50 -12 -20 -5 -1 5 -54 2-20 -28 -2 -2
9128
... ... ...
9356 9724
% Oxygen Inert gas method, av. of 3 detns. ... ...
By differenceh
...
...
18.1
18.3
7.5 30.8
7.3 31.2
... ... ...
Optimum Tzmp.,
... ... ...
c.
2400 2400 2400 2000 2400 2050 2350 2500 2400 2000
... ... ... ... ... CeS ... Time per analysis. 10-1 5 minutes. Operating conditions. Conductometric operation; 7 g. Ba(0H)Jliter; variable resistor; threaded spectroscopic preformed electrode capsule; long graphite reaction crucible plus, flared lip thimble, temperature as needed for type sample. 4 Variety of samples in this range have been successfully analyzed by this method. * 100yo - (matrix elements, in yo,determined by wet chemistry). AlN
was also cut off. The capsule dimension of 6 X 10 mm. fits well within the area of the stopcock loading cavity. The induction furnace was adjusted so that the temperature of the graphite reaction crucible would remain constant a t 2400" C. after the preliminary 2minute preheat. The sample was placed in the loading stopcock and immediately run, following the procedure as outlined for calibration of the working curve. The average blank (ohms) was deducted from the sample reading (ohms). The grams of oxygen in the sample were determined from the standard oxygen curve using this net ohms value. All samples were analyzed in the same manner. Results. Uranium oxide (UaOs NBS 950), which was dried a t 105' C. for 1 hour, was used for the initial study. As shown by Table I, the precision was found to be acceptable, with an oxygen content of 14.1 and 14.2%.
The accuracy appeared to be biased, as the expected yield of 15.2% was not found by analysis. This material was heated at 900" C. for 1 hour and allowed to cool. The uranium oxide was again analyzed. Satisfactory precision once more was attained. The statistical evaluation of the ten oxygen determinations of this uranium oxide (UaOS, after 900' C. heating) resulted in a precision measurement with a standard deviation of +0.12TC and a relative standard deviation of 0.8%. The arithmetic mean of 15.0Q/o now indicated no significant bias from the expected theoretical yield of this material. X-ray diffraction patterns on reaction products formed in the graphite crucibles from samples of uranium oxide and silicon dioxide indicated the complete conversion (greater than 95%) to metal carbides. This information, VOL. 33, NO. 12, NOVEMBER 1961
1769
with satisfactory reproducibility of the oxygen value, gave assurance that the temperature was sufficiently high to attain a complete reduction reaction of these oxides. Optimum temperatures for the determination of oxygen in silica and uranium osides were 2050” and 2400” C., respectively. Heating above 2150” C. was avoided with silica, as silicon carbide, which forms inside the capsule, sublimes a t 2200” C. If allowed to sublime, this would coat the inside crucible walls, thereby inhibiting the active carbon surface. Table I1 lists a number of oxygen determinations successfully obtained by this inert gas diffusion method. Some of the inorganic samples were chemically analyzed. The oxygen found by difference compared favorably with the oxygen results of the direct inert gas diffusion method.
The reaction crucible together with the flared lip thimble when cooled can be lifted out of the reaction chamber and the used capsules removed with tweezers. By reinserting the thimble and the empty crucible into the reaction chamber and degassing for 15 minutes at 2600” C., the assembly can be reused. Blanks should again be well Rithin limits after this degassing time. This operation of emptying and reusing the crucible can be carried out as many times as desired, as long as a film of metal or carbide does not form on the inside of the reaction chamber. This method offers a variety of possibilities on the future application for the determination of oxygen in volatile samples. Precision and accuracy combined with speed and ease of operation make this a highly reliable and economical method for the direct determination of oxygen in a wide variety of materials.
DISCUSSION
ACKNOWLEDGMENT
Tables I and I1 indicate various inorganic samples that have been successfully analyzed by this technique. Four to five determinations can be carried out in any one reaction crucible.
The authors acknowledge the assistance of William E. Chambers, and the x-ray diffraction interpretation of Donald E. Mentzer. The met chemical analyses were performed by Delores
Leonard and Martha F. Sweeney to confirm total add-up of compounds. The assistance of the National Carbon Co. Research Laboratory in carrying out this investigation is deeply appreciated. REFERENCES
(1) Banks, C. V., O’Laughlin, J. W., Kamin, G. J., ANAL.CHEM.32, 1613-
16 (1960). (2) Bennet, E. L., Laboratory Equipment Corp., St. Joseph, Mich., personal communication. (3) Crumpler, T. B., Yoe, J. H., “Chemical Computations and Errors,” pp. 131-4, M7iley, New York, 1940. (4) Elbling, P., Goward, G. W., ANAL. CHEM.32,1610-13 (1960). (5) Kallman, S., Collier, F., Ibid., 32, 1616-19 (1960). (6) Laboratory Equipment Corp., Instruction Manual for Operation of LECO Oxygen Analyzer N o . 534300, 1958. (7) McKinley, T. D., private communication, “Procedure for Preparation of Absolute Oxygen Standard.” (8) Singer, L., IND.ENQ. CHEM.,ANAL. ED. 12. 127 (1940). (9) Smiliy, W. G.; ANAL. CHEM. 27, 1098-102 (1955). RECEIVEDfor review June 12, 1961. Accepted September 1, 1961. Division of Analytical Chemistry, 140th Meeting, ACS, Chicago, Ill., September 1961.
Behavior of Substituted Aromatic Acids in Selected Nonaqueous Solvents ROY R. MlRON California Research Corp., Richmond, Calif.
DAVID M. HERCULES Departmenf of Chemistry, Juniata College, Huntington, Pa.
b The effect of nonaqueous solvents on the acidic behavior of substituted benzoic acids and phenols has been correlated with structural properties of the acids in a variety of nonaqueous solvents, Specifically, a linear relationship between AHNP and pK, has been demonstrated. A correlation has also been established between Hammett’s u value and AHNP in the same solvent series. Four of the acids studied deviated from “normal” behavior and the causes for these deviations are discussed.
S
1936 there has been considerable effort directed toward improving acid-base titrations in nonaqueous solvents, and extensive reviews have been written (11, 12, IS). Streuli and Miron (17) have correlated the behavior of substituted acids in pyridine with their behavior in water INCE
1770
ANALYTICAL CHEMISTRY
by plotting the difference in halfneutralization potential from a standard acid (AHNP) against pK, in water. Recently this work has been extended by Streuli to titration of bases in nitromethane (16). Hall (3) has conducted a n extensive investigation of the titration behavior of a large number of acids in a variety of solvents. The present investigation was undertaken to determine if the relationships established by Streuli and Miron in pyridine held for the titration of substituted benzoic acids in solvents of widely differing dielectric strength and basicity. This was accomplished by studying the relationship between AHNP and pK, for the substituted benzoic acids in a variety of solvents. Also, it was felt desirable to correlate AHNP with some property of the acid, more readily available to investigators than pK, values. Because tables of Hammett’s u values for substituent
groups on the benzoic acid nucleus are readily available, and are related to pK,, we felt that a correlation between AHNP and u would be desirable. The investigation has demonstrated that linear correlations between AHNP and pK. do exist in a variety of solvents, and that linear relationships between AHNP and Hammett’s u exist in the same solvents. Several of the submstituted benzoic acids-namely, aminobenzoic, p-aminobenzoic, pmethylbenzoic, and p-nitrobenzoic acids -show consistent deviations in one or the other of these plots. In most cases, these deviations can be explained in terms of equilibria or resonance considerations. EXPERIMENTAL
Reagents and Solutions. All of the acids studied were either Eastman Kodak White Label grade or were obtained from university stock.
