average relative precision of the method is within h 13% a t the 95% confidence level. T o investigate the accuracy of the method, the samples were run in a random fashion. The hydrogen content of a lot of uranium tetrafluoride n as established by averaging the results of 10 determinations (sample 4, Table 11). d n other saniplc of the same lot nas spiked with 33.0 p.p.m. hydrogen in the form of barium chloridc dihydrate, and 10 more determinations were run (sample 5 , Table 11). The difference between the averages of the amount of hydrogen found and the residual is 35.4 p.p.m. At the 957, confidence level this value is not significantly different from 33.0, the tlicoretical value of the spike. Consequcntly, the method is considered accurate. The method was used to determine water in refractory magnesium fluoride (Table 111). I n the range studied the average d a t i v e precision is =t 16% at the 95% confidence level. Because a 10-gram sample was the largest that
Table 111. Hydrogen Present as Water and/or Hydrogen Fluoride
P.P.M. 41,49
56,52 40; 49
P.P.31. 53,60 13, 13
pyrohydrolyze
ACKNOWLEDGMENT
The author expresses appreciation to
11,13
C. H. LIcBride, x h o offered valuable suggestions, and S. H. Huston, who
26,22
helped with statistical evaluation of the data.
21,23
8G, 87 29,33
pecially those which easily.
Sample size, 10.0 grams. Samples run in duplicate.
LITERATURE CITED
could be placed in the apparatus, precision was poorer than in the uranium tetrafluoride precision study. CONCLUSIONS
This procedure is suitable for determining hydrogen present in fluoride salts as water and /or hydrogen fluoride. When hydrogen concentration is about 35 p.p.m., good precision and accuracy can be obtained. The method should be applicable to the determination of small amounts of water in other salts, es-
(1) Feibig, J. G., TT-arf, J. C., AXAL. C m x . 2 6 , 927-8 (1954); 1\Isnhattan Proiect Reut. CC-2939 (June 29. 1945). ( 2 ) Hibbits, ‘J. O., Zucker, D.,’ U. S. Aitomic Eneygy Commission, Rept. Y-959 (1953). (3) AIcBride, C. H , llallinckrodt Chemical Works, unpublished TT ork. (4) TT’illard. H. H.. ef al.. “Instrumental Methods of .Snalysie,” 2nd ed., p. 213, Sew York, Van Kostrand, 1951. RECEIVEDfor review August 28, 1958. Accepted December 12, 1958. Second Conference on Analytical Chemistry in Kuclear Reactor Technology, Gatlinburg, Tenn., October 1, 1058.
Automatic Titration of Peroxides in Petroleum Products JOSEPH
S. MATTHEWS and JOAN F. PATCHAN
Gulf Research & Developmenf Co., Pittsburgh, Pa. .An improved method for determining peroxides in petroleum products involves oxidation of iodide ion and the continuous reduction of the liberated iodine with standard thiosulfate solution using an automatic potentiometric titrator. It i s applicable to low or high concentrations of diacyl-type peroxides and hydroperoxides. Olefinic and sulfur compounds do not interfere. The accuracy i s within 3% on peroxide numbers above 1 and within 0.1 peroxide number for lower values.
T
HE determination of peroxides in petroleum distillatrs has long been studied because of their effects, such as the formation of sediment, gum, color, or odor. Most techniques are based on reduction of the peroxide b y agents such ns iodide, arsenite, or ferrous ions. The ferrous method (10) is shonm b y Kolthoff and hfedalia (5) to be subject t o error and not reliably accurate. The arsenite method of Kalker and Connay (9) is applicable only to hydroperoxides. The negative hydroperoxide numbers are attributed to’the presence of reducing agents in the sample (6, 9). Generally, iodometric procedures have been considered to give the more nearly
theoretical results. However, there has been criticism of their application to petroleum products because the liberated iodine accumulates until the oxidation reaction is complete, and thus. side reactions n ith iodine as well as volatilization can give rise to unreliable results. Iodometric methods (2-5, 8 ) , however, have been shon-n to yield accurate results n ith knon-n concentrations of peroxides in the absence of interfering compounds of the type encountered in petroleum-derived material. -411 these methods begin with complete liberation of iodine followed by titration with staiidard thiosulfate solution. Csually starch indicator is used to determine the end point in aqueous titrations or the disappearance of the iodine color in nonaqueous solutions. K h e n solutions are colored, such end points are inapplicable. Abrahamson and Linschitz (1) describe a method for electronietric determination of the end point. I n the proposed method, the sample containing the peroxides is allowed to react with potassium iodide in a seniiaqueous medium and the liberated iodine is immediately and continuously reduced back to iodide as it forms, b y automatie addition of standard thiosulfate solution. The thiosulfate addition is controlled b y an automatic
titrator which is set a t a potential 10 to 20 mv. more negative than the equivalence potential of the titration of iodine by thiosulfate in the given medium. The iodine concentration is rarely allowed t o exceed a very small value and then only momentarily, to minimize side reactions. This method is applicable to the determination of any peroxides which liberate iodine from potassium iodide if the sample is soluble in the medium. APPARATUS
The titration cell consists of a 400-ml. Berzelius beaker cut to a height of 9 em. fitted with a five-hole cork stopper. A saturated calomel reference electrode, a platinum indicator electrode, buret tip, stirrer, themometer, and a coarse porosity fritted filter disk are inserted through the stopper. The cork also has a stoppered hole for introduction of the sample. For increased response, the electrodes and buret tip are positioned about the stirrer to direct the titrant onto the indicator electrode. The fritted disk is located below the stirrer to facilitate removal of dissolved oxygen. The assembled titration cell is heated by an asbestos-insulated, resistance wire heater controlled by a variable transformer. A Beckman Model K automatic titrator fitted with t n o delivery units was VOL. 31, NO. 6, JUNE 1959
1003
used, One unit is fitted with a 10ml. buret graduated in 0.05-ml. units for use with samples having peroxide numbers greater than 10. The other has a 5-ml. buret graduated in 0.01ml. units for peroxide numbers below 10. PROCEDURE
Dissolve 15 grams of potassium iodide in 10 ml. of m t e r , and add 100 ml. of isopropyl alcohol and 1 ml. of glacial acetic acid in the assembled titration cell. Fill the buret with standard thiosulfate solution. Turn on the stirrer and start oxygen-free nitrogen bubbling through the solution. Prepare the oxygen-free nitrogen by scrubbing prepurified nitrogen through a solution of sodium anthraquinone-@-sulfonate ( 7 ) . Turn on the heater and adjust the heat input by a variable transformer until the temperature of the medium is maintained a t 55" =t 2" C. By the time this temperature is reached, the medium will be deaerated. While the reaction medium is reaching temperature, deaerate the sample by bubbling oxygen-free nitrogen through it for about 5 minutes with the aid of a fritted-glass disk. If the sample is volatile, cool i t while it is being deaerated. K h e n the medium reaches operating temperature, set the titrator a t -50 mv. and the anticipator a t 10, and then the acid-base switch to acid to reduce any free iodine nhich may have formed. Zero the buret and turn the anticipator control to a setting between 1and 5 . Add colorless samples a t a rate to keep the formation of iodine color to a minimum. When the sample is colored, and titrant is adding too rapidly, add the sample a t the rate of %bout 1 ml. per 5 seconds. Cse 25nil. samples for peroxide numbers below 1, and 5-ml. samples for values above 1. The titration required for 25- and 5-mI. sample volumes should not greatly exceed 5 and 10 ml., respectively. If these limits are exceeded, repeat n-ith a smaller sample. Year the end of titration, turn the anticipator control back to 10 to prevent overtitration. 9 good criterion for the end point is when no further addition occurs for three times the time interval between the last two additions of titrant. If the time intervals are less than 1 minute, wait about 5 minutes before considering the titration complete. K i t h an automatic reset timing device, the end point can be automatizedn h e n no more titrant adds for a given time, the titrator is automatically turned Off.
Peroxide number as used here is defined as milliequivalents of active oxygen per liter. For peroxide numbers above 20, tn-o modifications may be used. Either the sample is diluted with a suitable solvent or a stronger standard thiosulfate solution is used. EXPERIMENTAL
The accuracy of this method was evaluated on synthetic blends of hydro1004
ANALYTICAL CHEMISTRY
Table 1.
Analysis of Mixtures
Peroxide Numbers Found Reference Arsenite Method (4) Titrator (9) Acyl-type peroxide mixtures 14.5 4.54 0.45 (by dilut'ion)
14.9 4.45 0.52
Hydroperoxide mixtures 14.1 3.74 0.36 (by dilution) Table II.
14.1 3.81 0.35
13.5 3.22 0.38
Determination of Peroxides in Petroleum Products
per- Peroxide Number osidea Found Petroleum Number Titra- ilmenite Product Added tor (9) Catalytically cracked gasoline None 0 43 0 21 0 60
Fuel oil S o . 2 None
1.19
None
0 83
None
Cs alcohol
0.60 14 4
None
0 55 0 63 14 4 14.1 6 56 Not deter-
JP-4fuel
None
None
0.60
0.54
mined Sone 0.78
Platformate None 0.19 Segative Premium gasoline Sone Sone Sone 0.35 0.33 0.22 15.2 15.5 13 4 Sone 3 92 Sotdeter-
Propylene po1y mer mined a Five grams of a mixture of tert-butyl and cumene hydroperoxides in 450 ml. of 1: 1 benzene-heptane solvent was analyzed and appropriate dilution made.
peroxides and diacyl-type peroxides. I n each case, the peroxide m s dissolved in 1 to 1 n-heptane-benzene solvent, A11 peroxides were commercially available and were used without purification. They were, however, assayed for active oxygen content by the Kokatnur and Jelling method ( 4 ) and all results are expressed in terms of active oxygen. The diacyl-type peroxide mixture consisted of benzoyl and lauroyl peroxides in the benzene-heptane solvent. Table I lists the results of these analyses by the reference method (4) and the titrator method. The synthetic blend containing a peroxide number less than 1 could not be analyzed by the reference method (4). The value shown was obtained by dilution of a higher concentration which had been so analyzed. An equivolume mixture of tert-butyl hydroperoxide and cumene hydroperoxide was dissolved in the benzeneheptane solvent and assayed for active oxygen (4). Solutions having peroxide numbers between 0 and 1, 1 and 10, and 10 and 20 were analyzed by the an arsenite reference method (4,
method ( 9 ) , and the titrator method, The results are shown in Table I. I n determining the peroxide content by the Kokatnur and Jelling method (4)it was observed that for the less reactive peroxides, all the active oxygen had not completely reacted after heating 5 minutes on a steam bath and a definite end point could not be reached because iodine kept reforming after titrating rrith thiosulfate. Because on heating, a blank did not liberate iodine, all determinations by the Kokatnur and Jelling method were done by reheating and retitrating until no more iodine was liberated. Thus, results reproducible to within 0.1 ml. of titrant were obtained consistently. The titrator method was tested on various petroleum products m-ith and 11ithout added hydroperoxides. The same samples were analyzed by the arsenite method ( 9 ) for comparison because the reference iodometric method (4) could not be used because of the color of the samples. Table I1 reports the results of duplicate determinations. RESULTS AND DISCUSSION
The titrator results listed in Table I are the average of triplicate determinations lyhich \?-ere precise to within 0.2 peroxide number for values above 1, and within 0.04 peroxide number for values below 1. The accuracy in all cases, compared TT ith the reference iodois n-ithin 37, for metric method (4, peroxide numbers greater than unity and within 0.1 peroxide number for values below 1. The advantage over the arsenite method is that this method is not limited to hydroperoxides. Because peroxides differ in reactivity, methods 11-hich choose arbitrary tinies for reaction may produce low results. The titrator method, however, has no arbitrary time limit. The only limit is the time for all the active oxygen to react. For example, certain peroxides may require a fen- minutes for complete reaction, TT hereas less reactive peroxides niay require a half hour or longer. V h e n a sample is first introduced into the reaction medium, the rate at lyhich iodine is liberated may vary depending on the concentration and reactivity of the peroxide present. During the titration, the addition of thiosulfate solution slows as the concentration as n ell as reactivity of active oxygen decreases. ,4t the beginning of the titration, additions of titrant niay occur a t intervals on the order of every few seconds. Xear the end point, the titrator adds thiosulfate a t intervals of 1 minute or more, if less reactive peroxides are present. Even so, titrations by this method are several times faster than b y the reference iodometric (4). The point when all the active
oxygen has reacted is taken as the time when no more thiosulfate solution adds from the titrator. Once the addition of titrant stops, no more is added, even if heating and stirring are continued for a half hour or more. The equivalence potential u-as determined b y a manual titration of knoivn amounts of iodine using a Beckman Model G p H nieter and the same titrant and reaction conditions as described. The potentials found were between - 30 and - 40 mv. HoneTTer, n hen the automatic titrator is employed in this range, the iodine concentration may build u p and never completely disappear except a t the end of the titration. By setting the instrument 10 t o 20 nip. more negative, a t the beginning and end of the titration there was about 0.0005 nieq. of thiosulfate present. Thus, free iodine could not increase to visible concentrations except for short periods. The reaction medium is not stable to air oxidation. K h e n heated and stirred in a n open beaker, iodine is liberated after 20 or 30 minutes as noted by the addition of titrant. Even trace amounts of oxygen in prepurifid nitrogen caused interference in the presence of tetraethyllead which n-as shown b y the continuous formation of iodine, much beyond the amount of peroxides present. Tetraeth>-llead was shown to be a n interference in two ways. TVhen the titrator method n as applied to a marine gasoline which contained the same additires as the leaded gasolines but no tetraethyllead, no interference was found. Also, nhen 0.3 nil. of pure tetraethyllead nas added to the reaction medium in the absence of peroxides but in the presence of air, the liberation of iodine was evident. Thus, titrations must be performed under an atmosphere of oxygen-free nitrogen. -4possible side reaction in published iodonietric methods n as the reaction of iodine with sulfur compounds. To test this, n mixture consisting of 0.35 gram of n-octyl mercaptan, 1.09 of isoamyl disulfide (isopentyl disulfide), 1.27 of diphenyl sulfide, and 1.22 of thiophene was dissolved in 1 liter of the benzene-heptane solvent containing four hydroperoxides to bring the peroxide number to about 13 (some peroxide is consumed through oxidation of the sulfur compounds). The unreacted perouides n ere then analyzed by the arsenite ( 9 ) and titrator methods. The peroxide content by the arsenite method nas 11.9 and that by the
Table Ill.
Effect of Titrant and Sample Volumes on Titration
ilpproximate
Volume of Thiosulfate, 1\11. 4.3 8.6
13.9 17.2 1.1
‘2.2 3.3 4.4 5.5 4.3
Sample
Volume,
Titration Error, %
5
0 -0.1 +0.8
1\11.
! 5
$1.3
25 25 25 25 25 50
0
0 +0.6 +0.2 +0.2
+5.1
titrator method \vas 12.2, a difference of about 2.5%. I n the presence of these sulfur compounds the titration proceeded much more slorrly. Besides a sample of cracked gasoline, Tvhich had a n olefinic content of SO%, a synthetic blend \vas tested ivhich contained purified DL-limonene and 1octene. For this test, 0.9851 gram of a n equal neight mixture of these coniponents TI as placed in the reaction niedium Tvith 5 nil. of a previously analyzed hydroperoxide mixture of peroxide nuniber 16.2. The peroxide nuniber found for this solution by the titrator method in the presence of these olefinic conipounds n a s 16.0, a difference of about 1.2%. The cracked gasoline listed in Table I1 n a s a difficult sample because of its instability to\\ ard peroxide formation. K h e n first received, i t titrated reasonably nell. Honever, after standing a fe\v days in the presence of oxygen, there was a rather rapid increase in peroxide content n hich TT as more readily noted by the titrator than by the arsenite method. The longer this gasoline \vas permitted to oxidize, the more difficult it became t o titrate t o a definite end point even while using oxygen-free nitrogen. This gasoline had a conjugated diolefin content of 1.5%. Therefore, this behavior may be clue to the presence of autoxidized conjugated diolefins 11hich Kagner, Smith, and Peters ( 8 ) shoned reacted slon 1y and incompletely n i t h iodide. Because the equivalence potential might be affected by the titrant and sample volumes added to the reaction medium, eaperinients TI ere conducted in TI hich 1-nil. increments of iodine in isopropyl alcohol n ere added to reaction
media containing 0, 5, 25, and 50 ml. of benzene-heptane solvent and titrated with 0.01.1’ sodium thiosulfate solution. The value in the absence of solvent was taken as the reference value. The results are given in Table 111. K i t h a sample volume of 5 ml. the titrant volume should not exceed about 10 ml. The titration error with about 9 ml. of thiosulfate was -O.l%, but at 13 ml. was +0.87,. Therefore, the higher peroxide numbers may be titrated if the titrant volume does not exceed about 10 ml. Sample volumes of 25 ml. did not produce appreciable error when 5 ml. of the titrant was consumed. A 5O-ml. sample volume was unsuitable because appreciable error 1%as produced when only about 4 to 5 ml. of titrant n a s added. Tests have shown t h a t as the solvent is made more aqueous the equivalence potential becomes more positive in these titrations, which explains the overtitration when larger volumes of titrant are used. The peroxide content of petroleum distillates, charge stocks, and fuels as well as petrochemicals such as propylene polymers, oxo alcohols, and aldehydes has been successfully determined b y this method. It is sensitive, reproducible, and accurate for hydroperoxides as IT-ell as diacyl-type peroxides in media containing compounds nhich would interfere in other iodometric procedures. The method should be applicable to the determination of any active oxygen compounds which liberate iodine from potassiuni iodide under the conditions of the test. LITERATURE CITED
(1) ilbrahamson, E. W., Linschitz, Henry,
AN.4L. CHEM. 24, 1355 (1952). (2) Dastur, N. N., Lea, C. H., Analyst 65, 286 (1940). (3) Ibid., 66, 90 (1941). (4)Kokatnur, V. R., Jelling, Murray, J . ilm. Chem. SOC.63,1432 (1941). (5) Kolthoff, I. AI., Medalia, -4.I., ANAL. CHEV.23,595 (1951). (6) Rippie, C. IT., 0 2 1 Gas J . 54, 215 (19~6). (7) Stafford, C., Puckett, J. E., Grimes, &I. D., Heinrich, B. J., ANAL. CHEM. 27, 2012 (1955). (8) Wagner, C. D., Smith, R. H., Peters, E. D., Ibid., 19, 976 (1947). (9) Walker, D. C., Conway, H.S., Ibid., 25, 923 (1953). (10) Yule, J. A. C., Kilson, C. P., Jr., Ind. Eng. Chenz. 23, 1254 (1931).
RECEITED for review June 20, 1958. Accepted January 28, 1959. American Petroleum Institute, Los Angeles, Calif., M a y 1958.
VOL. 31, NO. 6, JUNE 1959
1005