Karl Fischer Determination of Water in Ammonium PerchIorate with Automatic Tit ruti o n Ap p a ratus Evaluation of Reaction Rate Parameters and Statistical Evaluation EUGENE A.
BURNS
and R. F. MURACA
Propulsion Sciences and Space Sciences Divisions, Stanford Research Instifute, Menlo Park, Calif.
b The experimental parameters of the Karl Fischer titration and application of a Beckman Model KF-2 Aquameter for the accurate determination of low water content levels pf ammonium perchlorate are examined and defined. Parameters which have been investigated are: feed rate of titrant, rate of stirring, position of electrodes, volume of solution, effect of current passage in detection system, amount of water present in the sample, selection of end point, and catalysis b y salts in various solvents. The rate of reaction of the Karl Fischer reagent with water i s increased by use as solvent for the sample of a 3 : 1 pyridinemethanol mixture containing appreciable amounts of dissolved salts. An inexpensive attachment suitable for visual observation or recording of dead-stop end points in conjunction with the impedance detection system of the aquameter i s described. A procedure for the determination of the total water content of ammonium perchlorate i s presented with procedures for internal and surface water contents. The results of this method obtained by the nine major consumers and principal vendor of ammonium perchlorate have been subjected to a statistical analysis. With the refined data, repeatabilities of 0.0008 and 0.001 4%, and reproducibilities of 0.0018 and 0.0022E7, were obtained for samples having a water level of 0.01 and 0.05%, respectively.
T
HE accurate determination of the very small water content of amrnonium perchlorate is essential for proper formulation of solid propellants. The determination of water in ammonium perchlorate by Karl Fischer titration is simple and straightforward, as well as accurate (6). However, the use of automatic titration equipment in this determination has resulted in wide divergence of results (8) which has in turn prompted this investigation. This paper examines and defines the chemical parameters of the Karl Fischer titration and the instrumental factors
848
0
ANALYTICAL CHEMISTRY
attending the application of a Beckman Model KF-2 Aquameter for this determination. The major instrumental factor is the lack of anticipation of the end point which is influenced by the following parameters: feed rate of titrant, rate of stirring, position of electrodes, volume of solution, polarization of the electrodes, amount of water to be titrated, and selection of the end point. The major chemical factors are the finite rate of reaction between the Karl Fischer reagent and water, and the very small amount of water present. These factors require careful selection of proper solvents and catalysts, and control of the amount of sampIe. INSTRUMENTAL FACTORS
Feed-rate. The accurate titration of small water contents of ammonium perchlorate requires t h a t t h e commercially available, stabilized Karl Fischer reagent be diluted so the amount of reagent consumed can be accurately measured with the 10-ml. buret generally employed with the Model KF-2 Aquameter. I n direct titrations of the water content of ammonium perchlorate, the rate of reaction of the Karl Fischer reagent is low. Consequently, the Aquameter detects an apparent end point as soon as delivery of the Karl Fischer reagent is begun, and the detection of a n apparent end point actuates the mechanism which stops the flow of the reagent. As soon as tbe excess reagent is consumed by reaction with water, delivery of the reagent is again permitted, only to be stopped nearly as soon as started because of the renewed detection of an apparent end point. Thus, in a diiect titration, the Aquameter periodically delivers increments of the Karl Fischer reagent and, i t is obvious that the end point can be overstepped only by the amount of titrant delivered during the time interval between successive actuations of the reagent-delivery mechanism. Accordingly, the amount of titrant delivered per addition (feed-rate) must be small in comparison with the total volume of reagent required for the titra-
tion to minimize the error resulting from a n overstepping of the end point. I n a back-titration (Le., an excess of Karl Fischer reagent is titrated with a solution of water in methanol), an excess of iodine is always present until the end point is reached, hence the titrant is delivered continuously. Thus, an excess of titrant is delivered during the time which elapses between the detection of the end point and the actuation of the mechanism which stops the flow of titrant, and the end point is overstepped. The results obtained with direct and back-titrations are dependent on the feed-rate of the titrant; the effect of feed-rate on the equivalence of Karl Fischer reagent to the water-in-methanol solution is indicated in Table I for direct and back-titration. For fast feed rates, the apparent end point is early for direct titrations and late for backtitrations; both lead to a low ratio. The direct titration is less sensitive to changes in feed rate and is more reproducible; also the direct titration is not as dependent on the rate of stirring because the reaction of the Karl Fischer addition with water can be completed in the interval between additions. Thus, a direct titration of water with Karl Fischer reagent potentially has a lower error associated with it; therefore, this will be the mode of titration utilized in this work. Figure 1 presents a photograph of the KF-2 Aquameter. The valve diaphragm is sealed against leakage by an O-ring seal ( 7 ) . The burets are filled by dry air pressure, and the solutions are protected against atmospheric water pickup with Drierite. The feed-rate can be varied by adjustment of the stopcock while the valve diaphragm of the Aquameter is held open with the hand; because the feed-rate varies with the height of the solution in the buret, i t is necessary to compare feed-rates over the same volume intervals. A feedrate of 0.035 to 0.05 ml. per addition is desirable and this can be obtained easily by adjusting the stopcock so that the initial 1.5 to 2.5 ml. of titrant are delivered within 30 seconds when the valve
diaphragm is held open. Timing of this operation is accomplished conveniently by having the solution in the titration vessel past an end point, setting the end point duration for 30 seconds, actuating the titration button simultaneously with the manual opening of the titration valve, and observing the volume dispensed when the end point indication is heard. Selection of E n d Point. Two end point adjustments can be made: the potentiometer in the a r m of the bridge circuit which permits the concentration of iodine necessary to actuate the valve relay t o be varied, and the maximum interval between additions. V i t h these adjustments, a compromise between accuracy and speed can be obtained. The titration is more rapid and less accurate when the potentiometer is adjusted SO that the reagent addition is made when a relatively high concentration of iodine is detected by the electrodes, or =hen a short interval between additions is selected. The advantage of approaching the end point antomatically is that yon 'can reproducibly fall short of the stoichiometric end point; hence, analysis time
Figure 1. meter
Beckmon Model KF-2 A quo-
Table I. Effect of Feed Rate on Karl Fircher Water-in-Methanol Ratio Av. Ratio No. of MI. KF/Ml. Mode of Titration Feed Rate Detns. WM Std. Dev. Direct 0 035" 10 0.465 0.002 0.O6Oc 4 0.463 0 003 0 123" 6 n (1114 0, 195n
4
Feed rate expressed as milliliters per addition, averaged after 4 ml. s flow rate, milliliters per minute, for initial 8 ml.
tolerable errors in the water determination will resuit. However, it has been pointed out (2) in a similar case, if the titrant is standardized using exactly the same volume required for the determination, the error which arises from stopping a t a point on the titration curve other than the stoichiometric end point vanishes. We have found that adjustment of the potentiometer in accordancepiith the manufacturer's instructions (1) together with an end point period of 30 seconds to be optimum end point conditions. This is comparable to a "deadstop" end point of 4 to 7 @., between 1.0 sq. em. smooth platinum electrodes, 0.5 em. apart when the relay permits reagent addition. Rate of Stining. As noted above, the results of a back-titration are dependent on the rate of stirring because the velocity with which the added reagent is brought into intimate contact with the bulk of the solution will determine whether the electrodes detect a n excess of reagent; on the other hand, accurate results can be obtained in direct titrations when the rate of stirring is sufficient t o mix the solution satisfactorily because the rate of reaction appears to be slower. There is a wide range of speeds that are usable
tor u r e c t titrations Gnat are unsatisfactory for back-titrations. I n general, the maximum rate of stirring which does not splash or beat air into the solution is recommended ( 1 ) . Position of Electrodes. The time required for titration is determined by the relative position of the indicating electrodes and the point of entry of the titrant. T h e buret valve will be maintained in a n open position until iodine is detected by the indicating electrodes and the relay is actuated t o close the valve. T h e valve will again open when the excess of iodine has reacted with solution in the titration vessel. Location of electrodes in the solution away from the point of entry of the titrant results in a large feedrate and a relatively short time for titration. The merits of the low feedrate outweigh the short time required for titration; thus, it is necessary to have the electrodes as close as possible to the point of entry of the titrant. Volume of Solution. It was ohserved early in this work t h a t the length of time required for t h e titrations determining the equivalence of the Karl Fischer reagent and the water-in-methanol solution was dependent on the volume of solution in the titration vessel; Figure 2 shows
Figure 2. Volume of Karl Fischer reagent delivered os function of time of titration and initio1 volume
I
I I IO mi
Figure tector
RECORDER
3.
Deadstop end point de-
the results of a series of determinations obtained in t h e same cell and indicates that the rate of reaction is dependent on the concentration of reactants. Actuation of the buret valve occur3 when a certain low concentration of iodine is present in the vessel. When there is a smaller volume in the vessel, a given volume delivered from the buret will give a higher concentration of iodine than when a larger volume is in the vessel; hence, the response time of the instrument will be smaller, with smaller volumes the rate of reaction of the Karl Fischer reagent will be larger, and the titration will be effected within a shorter period of time. Unfortunately, the time required for titration of different concentration levels is not related directly as predicted by rate theory because of catalysis of the reaction by "spent" Karl Fischer reagent salts.
Catalysis of the Reaction. The time required for a titration is considerably reduced in the presence of dissolved salts and with the employ of a 3: 1 pyridine-methanol solution as a titration medium. In experiments, the titration of sodium tartrate dihydrate in a 3 : 1 pyridine-methanol solvent required only 20 minutes whereas the same titration performed in methanol required 31 minutes (a total volume of about 9 ml. of reagent was required). When the pyridinium salts from pretitration of excess water present in the pyridine-methanol mixture were allowed to remain in the titration vessel, the time required for titration of sodium tartrate dihydrate was reduced to 14 minutes. It has also been demonstrated
Table 11.
54
Water Taken, Mg.
12.92 25.83 12.92 25.83
50 83 79 2,897 mg. HeO/ml.
850
1.0
130 ELAPSED
Figure 4. curve
CHEMICAL FACTORS
Volume a t Start of Titration, Ml.
that anhydrous sodium bromide will also reduce the time of titration; however, the limited solubility of sodium bromide or other inorganic salts curtails their use. On the other hand, the 3 : l pyridinemethanol solvent does dissolve a relatively large amount of ammonium perchlorate (about 0.22 gram per ml.); this salt also shortens the titration time. This is fortunate because, as mentioned above, it is necessary to dissolve about 15 grams of ammonium perchlorate to titrate its lorn water content with good accuracy, and the salt itself will thus serve to reduce the titration time. I n general, titrations of the water content of ammonium perchlorate require 2 to 3 minutes with an automatic apparatus. The reduction of the time required for titration is caused by catalysis of the reaction between water and the Karl Fischer reagent. Changes in the resistance of the solution do not lead to reductions of the time required through changes in the response time of the Aquameter. The Beckman KF-2 Aqua-
140 TIME
-
270
28 0
min
Typical current vs. time
meter does not employ exclusively the principle of dead-stop indication ; the electrodes are one arm of an impedance bridge and the instrument is designed to stop the flow of titrant when the electrical current in the bridge is rendered large by a lowering of the impedance across the electrodes caused by a change in the state of polarization of the electrodes. The change in the specific conductance of the solution from 0.085 mho. per em. n hen a pretitrated 3: 1 pyridine methanol solvent is used to 0.151 mho. per mi. xhen 15 grams of ammonium perchlorate are added to 100 ml. of this solution is not sufficient to cause activation of the Aquameter circuits.
Amount of Water Titrated. The duration of the titration also depends on the amount of water initially in the reaction vessel. Because the Aquameter cannot anticipate the end point, the general trend is t h a t the time required to titrate a given amount of water decreases as the amount of water titrated increases; this trend is indicated by the results given in Table 11. Because the duration of titration also depends on the many factors already discussed above, it is not worthlyhile to establish the exact relat'ionship; in particular, it would be difficult to use such a relationship in practice because the actual amount of dissolved pyridinium salts is usually unknoxn. ANTICIPATION OF END POINT
It is possible to reduce the titration time still more by a preliminary manual addition of the titrant to within a few milliliters of the end point, and then permitting the end point to be achieved by an automatic titration. This procedure is particularly advantageous for titrations of sodium tartrate dihydrate and pretitration of the solvent. The manual manipulation of the delivery valve can best be done with experience; however, it is also possible to add titrant to a point -1 ml. before the end point predictably by observing the frequency of addition of the reagent. To aid this end a simple deadstop device n-as built (Figure 3), and the current that flon-s between two l-sq. em. smooth platinum electrodes (0.5 em. apart) n a s fed into a 10-mv. recorder. typical current us. time curve is given in Figuie 4. The average frequency of additions as a function of volume from the end point is plotted in Figure 5 . Employing this figure it is possible to observe the frequency of four additions and then manually add the predicted volume less 1 ml. to the solution, and then pcrmit the instrument to approach the end point automatically, and hence reproducibly. TYith experience, it is possible to predict the volume 30
l
i
!
,
,
,
l
3 4 5 6 VOLUME F R O M END P O ' N T
7
Effect of Water Taken on Time of Titration
Volume5 KF Solution, M1. 4.45 8.92
ANALYTICAL CHEMISTRY
4.50 8.86
Titration Time, Min. 11.3 16.6 14.0
18.2
\F7ater Found,
Titration Time, Min./Mg.
Mg.
H20
12.89 25.82 13.OA 25.65
0,875 0.643 1,084 0,704
I
0;
;
e
- ml Figure 5. Period of addition as function of volume from end point AVERAGE
0.5 to 1 ml. short of the end point without the aid of the external circuit. Using this latter technique it is possible to pretitrate the solvent and titrate a 15gram ammonium perchlorate sample in a period of only 6 to 8 minutes. REAGENTS AND PROCEDURE
Reagents. Stabilized Karl Fischer reagent, 1.4 to 2.3 mg. of H20 per ml. Dilute 750 ml. of commercially available stabilized Karl Fischer reagent (with water equivalence 5 to 7 mg. per ml.) to 2000 ml. with absolute methanol (O.Olyowater, max., commercially available) or stabilized Karl Fischer diluent; mix well and allow to stand overnight before use. Water-in-methanol solution, 1.5 mg. of H20 per ml. Methanol-pyridine, 1: 3 v./v. Prepare this mixture by either of the following methods. Method 1. Mix 5 pints of anhydrous pyridine with 850 ml. of methanol (max. 0.017, water, commercially available). Anhydrous pyridine can be prepared by any of the following: Treat with calcium hydride until hydrogen is no longer evolved; separate the liquid from the residue and distill it, or use only the clear supernatant liquid. Treat overnight with barium cxide; distill the supernatant liquid. Determine the water content of the pyridine; add 10% excess of benzene over that required to form the 8.9% water-benzene azeotrope (b.p. 69.3" C.). Distill the mixture until the temperature rises to 115.6" C.; the residue is anhydrous pyridine. Method 2. In preparation of large quantities of 3: 1 pyridine-methanol mixture, this method may prove to be more convenient. Mix a 5-pint bottle of reagent grade pyridine (max. 0.1%
Table 111. Sample
Day 1
A
2
1
0.009 0,008 0,009 0,008 0,009 0.009 0.008
3 AV.
Std. dev. 1
B
2 3 Av.
Std. dev.
0,009 0,009 0.0087
0,0005
0.052 0.052 0.047 0.049 0.051 0.047 0.048 0.048 0,047 0.0490 0.0021
water) and 850 ml. of absolute methanol (max. 0.01% water, commercially available), Add concentrated, stabilized Karl Fischer reagent from a graduated cylinder until the mixture is almost dry. (The solution becomes colored.) Often as much as 400 to 600 ml. of Karl Fischer reagent is required. If too much Karl Fischer reagent is added, add about 100 ml. of 1 : 3 methanolpyridine. A 4- to 5-liter amber-glass vessel will be required for storage of this mixture. Sodium tartrate dihydrate. If the water content of the sodium tartrate dihydrate is in question, it may be determined by volatilization a t 150" C. to a constant weight (minimum of 4 hours for 2- to 3-gram sample). If the per cent water is not 15.66, then the determined value should be used for the water equivalence of the Karl Fischer reagent. Recommended Procedure. Follow the operating instructions furnished by the manufacturer as much as possible. Set the apparatus so t h a t the determination can be accomplished by direct titration and set the timing mechanism for a n end point duration of 30 seconds. Adjust the titrant delivery rate so t h a t the first milliliter of Karl Fischer reagent is dispensed in 20 to 30 seconds when the automatic valve is held open. The accuracy of the determination of water is severely affected by the flow rate of the titrant. Because the flow rate varies with the height of Karl Fischer reagent in the buret, use the same volume interval, 0.00 to 1.00 ml., when adjusting the delivery valve. I n general, this adjustment ensures that 0.03 to 0.05 ml. is delivered per addition of reagent near the end point. Arrange or modify the titration vessel of the automatic equipment so that a slow stream of dry air or nitrogen can be passed continually over the solution in the vessel. The gas can be dried
Water Content (%)of Replicate 2 3 4 0.009 0.013 0.008 0.007 0.010 0.011 0.007 0.010 0.011 0.006 0.011 0.010 0,010 0.006 0.011 0.013 0.009 0.006 0,009 0.007 0,011 0,010 0.007 0.011 0,010 0.010 0.007 0.0103 0.0068 0.0107 0.0007 0.0011 0.0012 0,039 0,051 0.046 0,040 0.052 0.045 0.048 0.045 0.044 0.046 0.048 0.043 0.044 0.051 0.044 0.044 0.050 0.047 0.047. 0.050 0.044 0.044 0.047 0.045 0.047 0.049 0.044 0.0439 0.0493 0.0448 0.0026 0.0014 0.0012
by passing it through a glass trap cooled in dry ice or liquid nitrogen; a long tube (about 3 feet by 1 inch in diameter) containing a desiccant is also satisfactory. Alternatively, the titration vessel can be flushed out with air from a rubber hand bulb connected to a long drying tube. Spent solution in the titration vessel can be removed conveniently by a suction tube. This expedient permits running titrations without removal of the titration vessel. It is recommended that Karl Fischer titrant be added manually to within 0.5 ml. of the end point and that the titration then be completed automatically. By this means, the time required for each titration will be significantly reduced. Add 25 ml. of the methanol-pyridine mixture; start the automatic titration equipment and obtain an end point. Condition the apparatus by determining the relative strengths of the Karl Fischer reagent and the water-in-methanol solution by titration of about 10 ml. of the water-in-methanol solution with the Karl Fischer reagent; use the maximum rate of stirring which does not splash or beat air into the solution. Repeat until the ratios obtained from three successive titrations agree within ten parts per thousand. The apparatus needs to be conditioned only a t the start of the day and after interrupted use during a day. Transfer approximately 15-gram portions of the sample into dry weighing bottles; protect samples from loss or gain of water a t all times. Weigh each of the bottles (closed). Transfer 100 ml. of methanol-pyridine mixture to the titration vessel, pass dry air into the vessel (about 75 strokes if a rubber hand bulb is used) and obtain an end point. Open a weighing bottle and quickly empty as much of its contents as possible into the reaction vessel using a powder funnel and suitable small
Samples Found by Laboratories 5 6 7 0.008 0.013 0,009 0.008 0.012 0,010 0.010 0.008 0.013 0.009 0.008 0.012 0,009 0.008 0.011 0.013 0,009 0.008 0.009 0.008 0.011 0.009 0.008 0.011 0.008 0,009 0.012 0.0092 0.0080 0.0120 0.0004 0,0000 0.0009 0.047 0.048 0.048 0,047 0.049 0.045 0.048 0.048 0.046 0,044 0,047 0,044 0,046 0.048 0.049 0.047 0.049 0.047 0.044 0.047 0.047 0.045 0.049 0.046 0.044 0.046 0.047 0.0458 0.0479 0.0466 0.0016 0.0011 0.0015
8 0.016 0,014 0.011 0.009 0.008 0,012 0,010 0,009 0,013 0.0113 0.0026 0.055 0.053
0.052 0.053 0.051
0,053 0.056 0.052 0.057 0.0535 0.0020
9 0.001 0,001 0.005 0,001 0,001
0,001 0.003 0,002 0.003 0.0020 0.0014 0.044 0.038 0.049 0.045 0.042 0.043 0.049 0.054 0.052 0.0462 0.0030
VOL. 34, NO. 7, JUNE 1962
851
brush. Slant the funnel towards the center of the vessel. Do not try to transfer all the contents of the bottle because it is desired to minimize exposure of the sample t u the atmosphere. Allow the solution to stir about 30 seconds and titrate to an end point. A11 of the ammonium perchlorate will dissolve by the time the titration is completed. Reweigh the emptied weighing bottle and, from the difference in weights, obtain the weight of sample taken for titration. Then calculate the per cent water. If desired, determine the water equivalence of the Karl Fischer reagent immediately following the titration of sample. I n case of doubt or dispute it is required that this determination be made in the same medium as that of the titration of the sample. Dissolve 5 grams of ammonium perchlorate in 100 ml. of the methanolpyridine mixture and obtain an end
Table IV. Mean, Repeatability, and Range of Acceptable Results
Sam- Mean, ple % A
Repeatability,
%
Range,
70
0 0088 0.00102 0 0057-0 0119 0 0096" 0,00076' 0 0073-0.0119"
0.0474 0.00194 0.0416-0.0532 Calculated after rejecting data of laboratory 9 and correlating individual wild results. (See text.) B 0
Table V. Inspection of Data Via Dixon's Q-Test to Find Laboratories with Widely Deviating Averages Q Valuea
Sample A B
On On highest result lowest result 0,237 0.237 0.075 0,026
a Value must exceed 0.683 for rejection a t 99% level.
Table VI.
V , X lo6
1
2.28 1.06 1.50 0.889 0.556 0.611 1.83 3.67b 3 . 67b
3
4 5 6 7 8
9
STATISTICAL EVALUATION
The evaluation of the recommended procedure was accomplished by a round robin, participated in by members of the Joint Army-Navy-Air Force Panel on the Analytical Chemistry of Solid Propellants and supervised by R . P. Rice of the American Potash and Chemical Corp. Samples of two water content levels (approximately 0.01 and 0.05%) were selected at random and distributed to the major users of ammonium perchlorate for iocketry. The results transmitted by participants to the American Potash and Chemical Corp. are recorded in Table 111. The replicate data for the water content of both samples are listed as a function of the day of the series and the participating laboratory (coded). Also listed are the laboratory average and pooled standard deviation (calculated when day-to-day variances are ignored). The following nine laboratories participated in this program: Aerojet General Corp., Sacramento, Calif.; American Potash and Chemical Corp., Henderson, Nev.; Grand Central Rocket Co., Redlands, Calif.; Stanford Research Institute, Propulsion Laboratory, Menlo Park, Calif.; Thiokol Chemical Corp., Brigham City, Utah; Thiokol Chemical Corp., Elkton, Md.; Thiokol Chemical Corp., Huntsville, Ala.; Thiokol Chemical Corp., Marshall, Texas, and U. S. Naval Propellant Plant, Indian Head, Md .
Inspection of Data Via Barlett's x ; - ~Test to Find Laboratories with Widely Deviating Repeatabilities
Laboratory 2
point. Rapidly transfer 90 to 110 mg. of reagent grade sodium tartrate to the titration vessel using a powder funnel and suitable brush. Slant the funnel tonards the center of the breaker to prevent the salt from adhering to the sides of the titrat-ion vessel. Again flush the titration vessel with dry air (about 35 strokes) and titrate to an end point. Repeat the standardization until two successive results agree within ten parts per thousand.
log v, -5.642 -5.975 -5,824 -6.051 -6.255 -6.214 -5.738 -5.436 -5.436
$0.199 -0.134 + O , 017
-0,210 -0,414 -0.373 +D. 107 $0,405 $0,405
(log VO)I = -5.841, X Z - ~ = 15.95 (Significant at 95% confidence level). (log V O )=~ -5.957, xZ-1 = 5.23O (Significant a t 10% confidence level). Critical values of xf-,, 99% confidence level:
xi
(I
+0.315 -0,018
+ O , 133
-0,094 -0,298 -0.257 +0.219
20.09, xi = 16.81 Log VO- (log V0),obtained after rejecting data of laboratories 8 and 9. Only 6 degrees of freedom used in this calculation. xl-I calculated after rejecting data of laboratories 8 and 9.
852
=
ANALYTICAL CHEMISTRY
Statistical Evaluation of RoundRobin Results. The statistical evaluation of the results of this round robin conforms to the method outlined by Burns and lIuraca (3) and McArthur et al. (6). Estimates of the repeatabilities, So, within laboratories were made for the raw data of both samples so that the data could be combed for "vcild" results. I n general, any data which differ from the grand mean of the sample by more than three times the repeatability, 3S,, are considered rejectable. Table IV lists the mean, repeatability, and range of acceptable data for both samples. Wild data are replaced by the mean of the remaining data of the respective day. For sample A , the data of laboratory 9 are obviously out of control as well as value 1, day 1, laboratory 8. Several other results are wild on the basis of the first approximation; however, the significantly low results of laboratory 9 have strongly influenced the first evaluation of mild data. Therefore, computation of the second evaluation of wild data was afforded by rejecting all of the data of laboratory 9 and replacing value 1, day 1, laboratory 8, with the value 0.013. The second evaluation reveals that 20 pieces of data are wild; it is not practical to adjust this amount of data, hence the data were not corrected any further. Similarly, it is seen from Tables I and I1 that for sample B, the mean of laboratory 8 is outside t'he limit of acceptable values, hence the data of laboratory 8 are rejected. Also, t'he following wild values are replaced: Laboratory 2, day 1, values 1 and 2 ; laborat'ory 9, day I , value 2 , and day 3, value 2 with 0.044, 0.044, 0.047, and 0.051, respectively. The averages of samples A and B for the remaining 8 laboratories were compared via Dixon's Q test (4,pp, 145-6) a t the 99% confidence level. The Q ratios are shown in Table T.'. The Q test shows t'hat no laboratory has averages vihich are sufficiently high or low to be rejected. Data of the remaining laboratories Tyere examined to determine whether some show a significantly greater spread in results than others. The repeatability of the data from each laboratory was dctermined (the pooled standard deviation, 12 degrees of freedom) and Bnrtlett's x2,-] value (9, p. 276) was calculated to be 15.95 (Table VI). This is not high enough to prove (with 99% confidence) that the remaining data lack homogeneity. S o further data need be reject'ed. However, this value of is suspiciously high (shows a lack of homogeneit'>- a t 95% confidence level). I t is apparent that laboratories 8 and 9 have a much higher within-sample variation than t'he other laboratories. From a practical standpoint, the data of these laboratories were
deleted and the new value of x2n-1of only 5.23 shows that the balance of data is homogeneous. The repeatability of the two samples was examined to ascertain whether it varied significantly with the magnitude of the water content of the samples. An F test yields a value of 6.08, indicating that there is a 9970 significant variation of rcppatability; hence the data for the two samples must be handled separately. The analysis of variance of all of the remaining corrected data of samples A and B is recorded in Tables VI1 and VIII. Because the day-by-day laboratory interactions for samples d and B are not significant a t the 99% level, i t is permissible to pool the interaction and replicate variance to evaluate the repeatability ( 9 , p. 7 7 ) . The variance between laboratories is significant at the 99% level, hence the components of variance due to this factor must be incorporated in the calculation of repeatability ( 6 ) . The reproducibility of the refined and raw data was calculated by evaluating the components of variance due to sources which were found to be significant, summing the variance, and taking the square root of the sum. The calculated mean, repeatability, and reproducibility of the raw and refined data of samples A and B are shown in Table I X . Conclusions. T h e application of automatic titration apparatus t o the determination of water in ammonium perchlorate by the Karl Fischer titration yields values of repeatabilities and reproducibilities which are influenced by the level of water in t h e sample. On the average, a single water determination made in the laboratory will lie within the following intervals : Water Content 0 01yo 0.05%
Reproducibility 3 ~ 00031% absolute, 68% of the time 1 0 0062% absolute, 95% of the time f O 0036% absolute, 68% of the time f0.007270 absolute, 95% of the time
However, a laboratory will generally reject wild or obviously erroneous results; then a single determination made in any laboratory will lie within the following intervals : Ifrater Content 0.017~ 0.05%
Reproducibility 10.0018% absolute, 68% of the time + O . 0036% absolute, 95% of the time 10.0022% absolute, 68% of the time =!=O.0044% absolute, 95% of the time
Table VII.
Analysis of Variance of Refined Data of Sample A
Sum of Squares
Source
+
2 167 17 14 32 202
667 745 678 667 345 757
1.33 27 96 1 47 0 350 0 599
2
6 12 42 54 62
Fa
2 20 46 A8b
Calculated using pooled estimate of variance due to differences of replicate test, 0.599, Significant a t 99% confidence level. Table VIII.
Analysis of Variance of Refined Data of Sample 6
Source
Sum of Squares x 106
Between days Between laboratories Interaction Redicates Pobled replicates interaction Totals
8.603 200 111 18 174 89.334 107,508 316 222
+
5
Variance x 106
of
Freedom
X 105
Betyeen days Betyeen laboratories Interaction Replicates Pooled replicates interaction Totals a
Degrees
Degrees of
Freedoni
Trariance
x
lo6
4.30 33 35 1.51 2.13 1.99
2
6 12 42 54
62
Fa
2.16 16 76b
Calculated using pooled estimate of variance due to differences of replicate test, 1.99. Significant at the 99% confidence level.
These tabulated values represent the over-all reproducibility to be expected from a sample of ammonium perchlorate which has been sealed, shipped, opened, sampled, and analyzed for mater content with automatic titration equipment. These values, then, represent the deviation expected among distant laboratories analyzing the same sample. The results are acceptable in spite of the fact that the between-laboratory reproducibility is significantly greater than the within-laboratory repeatability. This fact is explained by “better-than-shouldbe-expected” results within laboratories, which is accounted for by obtaining operator-bias common to routine operations associated with handling of sample, stirring rate, selection of end point, flow rate, and reading of buret. It can be concluded that control of the experimental variables can be achieved readily enough to recommend the procedure as a referee method for the analysis of water in ammonium perchlorate.
Table X.
Table IX. Comparison of Mean, Repeatability, and Reproducibility of Raw and Refined Data
ReRepeat- producSamhlean, ability, ibility, ple Data % 70 % A Refined 0.0094 0.0008 0.0018 0.0088 0.0012 0.0031 Raw B Refined 0 0469 0 0014 0.0022 Raw 0 0474 0 0024 0 0036 SURFACE AND INTERNAL WATER
I n the manufacture of amnionium perchlorate, inclusions of mother liquor are always obtained within the c-rystal. In addition, n-ater adsorbs on the surface of the crystals. Hence, the total mater content is the sum of the “internal” and “surface” components. A novel determination of these quantities is presented as follows: Surface water is obtained by the loss in weight by volatilization a t 110” C., and also by
Determination of Total, Surface, and Internal Water Contents of Typical Ammonium Perchlorate Samples
Per Cent Water Total Water by Surface water by Sample Karl Fischer Oven drying Karl Fischer A B C
D E
0 0 0 0 0
0385 0295 0186 0141 0366
0 0056 0 0056 0.0137 0 0083 0 0007
0 0 0 0 0
0047 0047 0124 0067 0010
Internal xater byDifference Karl Fischer 0 0 0 0 0
0338 0248 0062 0074 0356
0 0 0 0 0
VOL. 34, NO. 7, JUNE 1962
0341 0256 0079 0082 0361
853
Karl Fischer titration of a sample placed in a water-free saturated ammonium perchlorate-in-methanol solution; internal water is determined after solution of oven-dried material by Karl Fischer in dry pyridine-methanol solvent, by Karl Fischer titration, and also evaluated from the difference between total and surface water contents. Examples of the type of results that are obtained for these determinations are shown in Table X. LITERATURE CITED
( 1 ) .Beckman Instruments, Inc., Instruc-
tion Manual Models KF-2 and KF-3
Aquameters, Fullerton, Calif., Decemb& 1958. ( 2 ) Burns, Eugene A., R . F. Muraca, Anal. Chim.Acta 23. 136 11960). (31, Burns, Eugene A’., R.‘ F. Muraca, Round Robin 13-C of the Joint ArmyKavy-.4ir Force Panel on the .4nalytical Chemistry of Solid Propellants: Evaluation of the Karl Fischer Method for the Determination of Water in Animonium Sitrate.” Progress Report No. 20-354, Jet Propulsion Laboratory, California Institute of Technology, April 1, 1958. (4) Dixon, W. J., Massev, F. J., “Introduction’ t o Statistical “Analysis,” X c Graw-Hill, Xew York, 1951. 1 5 ) JANAF-PACSP Handbook on Recommended Analytical Methods and Specifi-
cation Procedures, Method 300.1 and 701.0, July 1959. (6) McArthur, D. S.,Baldeschwieler, E. L.. White. W. H.. Anderson. J. S..ANAL. CHEM.26. 1012 11954). ’ ( 7 ) Muraca,’ R. F:, Burns, Eugene A., Chemist Analyst 5 0 , 121 (1961). (8) Rice, R. P., Summary of Round Robin conducted by American Potash and Chemical Corp., August 19, 1959. (9) Villars, D. S., “Statistical Design and Analysis of Experiments for Development Research,” FTilliam C. Brown Co., Dubuque, Ion-a, 1950. RECEIVEDfor review July 10, 1961. Accepted April 5, 1962. Division of Analytical Chemistry, 138th Meeting, ACS, New York, N. Y., September 1960.
N e w Method for Determination of Creatinine in Urine by Ion Exchange Separation and Ultraviolet Spectrophotometry WILLIAM S. ADAMS, FRANCES W. DAVIS, and LOUIS E. HANSEN Department of Medicine, School of Medicine, University of California at los Angeles, and Wadsworth Hospital, Veterans Administration, los Angeles, Calif.
A specific method for creatinine determination utilizes physical rather than chemical properties. The procedure is based on the separation of creatinine in urine by an ion exchange resin and subsequent ultraviolet spectrophotometric measurement of the compound a t pH 10.4. Triplicate analyses and standard recovery procedures show good reproducibility. To date, no interfering substances have been encountered. The reagents used in this procedure are stable over long periods of time. The number of variable factors which must b e closely controlled in colorimetric methods are minimized in this procedure.
A
entirely new method for the determination of creatinine in urine uses an ion exchange resin and the maximum ultraviolet absorption wavelength of creatinine at 234.5 mp in an alkaline solution. Because physical properties and not chemical characteristics are utilized, the method is specific, as well as simple, rapid, and accurate over a wide range of concentration. The creatinine value indicates the completeness of a 24hour urine collection in many laboratories. Several methods (2, 7, 8) may be used for the determination but most colorimetric procedures currently employed are not entirely specific or require close con854
N
ANALYTICAL CHEMISTRY
trol of many variables (3, 9). To eliminate the presence of possible chromogenic substances which interfere in the creatinine assay, two major modifications of the Jaffe reaction have been proposed in recent years. Dubos and Miller (4) used a microorganism, Clostridium ureafaciens,to destroy quantitatively the creatinine present, but this method is not without objections (9). Gaebler (5) employed Lloyd’s Reagent to separate creatinine from the chromogenic “pseudocreatinine,” and a modification has been described by Haugen and Blegen (6). The work in the authors’ laboratory involved the separation of various components of 24-hour urine collections by ion exchange resins. To standardize the difference in total output of patients, creatinine values were used as a criterion to determine the aliquot of urine t o be sorbed on to a resin column. Thus a rapid accurate procedure was necessary. The method of choice appeared to be the modification of Haugen and Blegen (6), but unsatisfactory reproducibility of results on some specimens was encountered, even though the HartmanLeddon Co. preparation of Lloyd’s Reagent was used. I n the technique employed for urinary fractionation with a Dowex 2 resin column, creatinine was found by ultraviolet absorption measurements to be in the first portion of the eluate ( I ) .
Further analysis, both by spectro.photometry and paper chromatography revealed the additional presence of creatine and N-methyl-2-pyridone-5carboxamide. Creatine is normally found only in trace quantities in urine. The absorption curve of this compound exhibits no characteristic pattern except for an increased absorption in the region below 230 mp. However, to obtain this high end absorption, there must be a concentration far in excess of that found in urine. Consequently, with the possible exception of extreme cases of creatinuria there would be no distortion of the creatinine curve due t o creatine. I n a series of fractionations of normal and leukemic urine, an average of 34 mg. of N-methyl-2pyridone-5-carboxamide was found to be excreted in normal urine per 24 hours. The highest average value obtained for the different types of leukemia studied was 83 mg. per 24 hours although an isolated case of chronic granulocytic leukemia showed an excretion of 210 mg. However, since the pyridone a t p H 10.4 exhibits a minimum at 233 mp whereas creatinine has a maximum at 234.5 mp (10, 11), interference from the pyridone would be negligible. If creatinine had little affinity for the resin column and other substances eluting at the same time were either present in trace amounts or had an absorption maximum far removed from creatinine, a rapid
