ratus especially recommends this system for us(' as a process control device. The scintillation detector used in the present system could be replaced by a proportional counter with little loss in sensitivity and resolution and the x-ray unit could be replaced by a beta-excited lowenergy x-ray source (6). I n this way a rugged, inexpensive, process control unit could be constructed. It is possible to make the system selfmonitoring by increasing only slightly the complexity of the electronic circuitry. This can be done by using a two-channel pulse height analyzer in place of the more simple lower-gate discriminator circuit now used. One channel, set to include a part of the spectrum which is insensitive to addition of high 2 material (such as the leading
edge of the Compton smear), can then be used as a control channel while a second signal channel (essentially a lower-gate discriminator) covers all energies greater than 80 k.e.v. Counts from the control channel, after multiplication by a suitable constant, can automatically and almost instantaneously be subtracted from the signal channel count to give an output count dependent almost exclusively on the concentration of high Z atoms. This system automatically compensates for fluctuations in sample density, changes in x-ray flux, or variations of detector sensitivity. Calculations made from existing data have shown the feasibility of this sclf-monitoring scheme and the electronic circuitry required to incorporate this fea-
ture into the system is of conventiona! design. LITERATURE CITED
(1) Bartlett, T. W., ANAL.CHEM.23, 705
(1951).
('2) Birks, L. S., Brooks, E. J., Friedman, H. Roe, R. M., Zbid., 22, 1258 (1950).
(3) dalingaert, G., Lamb, F. W., Miller, H. L., Noakes, G. E., Ibid., 22, 1238
(inm). \ - - - - I
(4) Hughes, H. K., Hochgesang, F. P.,
Ibid., 22, 1248 (1950). (5) Leroux, J., Maffett, P. A , , Monkman, J. L., Zbid., 29, 1089 (1957). ( 6 ) ReifTel, L., Kucleonics 13, No. 3, 22 (1955).
RECEIVEDfor review March 27, 1959. Accepted July 20, 1959. Work supported in part by the U. S. Army Engineer Research and Development Laboratories, Contract No. DA-44-009-ENG-3536.
Flame Photometric Determination of Lead and Manganese in Gasoline GARLAND
W. SMITH and ALTON K. PALMBY
Research laboratories, Ethyl Corp., Detroit, Mich.
b Rapid and accurate flame photometric methods have been developed for determining lead and manganese in gasolines. In each method, the sample is burned in an oxyhydrogen flame. In the lead method, base stock effects are eliminated by using an incremental addition of lead for calibration. In the manganese method, base stock effects are reduced sufficiently by diluting the sample at least twentyfold with iso-octane. Lead naphthenate and tetraethyllead are equally satisfactory reference materials for determining tetraethyllead; manganese naphthenate is satisfactory for determining organomanganese additives. Burner fouling, a maior source of error, is eliminated by periodically aspirating acetone through the burner. The determinations are accurate to about 2% of the amount present for lead, and to about 3% for manganese. Approximately 40 lead or 80 manganese determinations can b e made per man-day.
T
time requirement of chemical methods for determining tetraethyllead in gasoline has prompted considerable research for rapid instrumental methods. Methods employing x-ray absorption, x-ray fluorescence, polarographic, and flame photometric techniques have been reported. X-ray absorption methods are complicated by HE
1798
ANALYTICAL CHEMISTRY
sulfur interfercnce (4). X-ray fluorescencc methods are fast and accurate but require expensive equipment (10). Polarographic mcthods, while more rapid than conventional chemical procedures, are still relatively slow (2, 5, ?, 12, IS). Flame photometric methods previously reported require either the unlradrd bas[! btork of the sample or thv handling of concentrated solutions of tetraethyllead ( 3 , 6 , 8 , 9 , I I ) . This papcr describes a rapid and accurate flame photometric method for detc,rmining tetraethyllead based on the lead content of the gasoline which eliminatm the need for the unleaded basr stock and concentrated tetraethyllead solutions. I n addition, a technique is incorporated into the method to prevent fouling of t,he burner. The method is applicable to tctraethyllead concentrations from 0.1 t o 6 ml. per gallon, and is accurate to about 2% of the amount of tetraethyllead present, bascd on 95% confidence limits. Approximatcxly 40 samples can be analyzed per man-day. The recent d e d o p m e n t of a new antiknock compound, (methylcyclopentadieny1)manganese tricarbonyl, necessitated a rapid method for determining manganese in gasolinc. The success of the flame photometer in determining tetraethyllead suggested its use for manganesc. For this determination, the effects of differences in gasoline base stock composition are minimized by a twentyfold dilution of the sample with iso-
octane. The method has been used to determine manganese concentrations from 0.1 to 4 grams per gallon. Based on 95y0 confidence limits, the method is accurate to about 3% of the amount of manganese present. About 80 samples can be analyzed per man-day. APPARATUS
.I Beckman
Model 2400 DU spectrophotometer, equipped with a Beckman Model 9200 flame photometry attachment and a Beckman Model 4300 photomultiplier attachment, was used. A &volt storage battery was used as the power source to ensure independent operation of the dark current and sensitivity controls of the instrument; Lhe Beckman DU power supply, Model 23700, is equally sntisfactory. Beckman aspirator-type oxyhydrogen burners, Model 4050, with small-bore capilh i e s were used. Rotameters wcre installed in the oxygen and fuel gas lines to aid in maintaining constant flame conditions. The plated tubing in the gas regulator unit was replaced with stainless steel tubing because particles of metal plate sometimes flaked off and lodged in the burner. A neoprene disk was used to cover the sample cup to prevent ignition of gasoline vapors. The disk, cut from a '/le-inch neoprene sheet, was positioned by slipping the burner capillary through a pinhole punched in the disk. PROCEDURE
Lead Determination.
A reference
standard containing lead equivalent t o 3.00 ml. of tetraethyllead per gallon (3.171 grams of lead per gallon) is prepared by dissolving chemically analyzed lead naphthenate in isooctane. The instrument is warmed up until galvanometcr readings bezome constant. The oxyhydrogen burner is lighted and 5 minutes allowed for the flame to come to equilibrium. The selector switch is set a t 0.1 and the photomultiplier switch is set to give adcquatc amplification. The slit width is set a t approximately 0.03 mm. A portion of the lead reference standard is aspirated into the fiame, and the wave length setting is adjusted to 405.8 mp by noting the point of maximum lead emission. The concave mirror in the burner housing is adjusted until maximum lead emission is obtained. The unburned portion of the reference standard is removed from the burner, and acetone is aspirated into the flame to clean the capillary of the burner. A 5ml. portion of the sample is diluted to 10 ml. with the lead reference standard. A second 5ml. portion of sample is diluted to 10 ml. with clear isooctane. A portion of the g m l i n e sample which was diluted with the lead reference standard is aspirated into the h e . The wave length scale is set a t 400 mp, and the transmittance scale at zero. With the shutter open, the darkcurrent control is adjusted until the galvanometer needle coincides with the last fiducial mark to the left of zero on the galvanometer scale. This fiducial mark was chosen arbitrarily. The closed-shutter, dark current galvanometer reading is noted. This reading must not change during the subsequent steps. With the sample still being aspirated, the shutter is opened, the transmittance scale is set at loo%, and the wave length scale is adjusted until maximum emission is obtained for the lead line a t 405.8 mp. The sensitivity control is adjusted until the galvanometer needle coincides with the fiducial mark. If the closed-shutter, dark current galvanometer reading changes after the first adjustment of the sensitivity control, a second approximation of the dark-current and sensitivity control adjustments must be made. A portion of the gasoline sample which was diluted with lead-free isooctane is aspirated into the flame and the transmittance scale adjusted until the gslvanometer needle returns to the fiducial mark. The reading on the transmittance scale is noted. The unburned sample is removed from the burner housing, and acetone is aspirated into the flame. The tetraethyllead content of the gasoline sample is calculated as follows: Tetraethyllead (milliliters per gallon) = AB/(100 - 8) where A = tetraethyllead concentration in ml. per gallon equivalent h the lead content of the reference standard B = transmittance scale reading
Manganese Determination. A reference standard containing 0.10 gram of manganese per gallon is prepared by dissolving the required amount of chemically analyzed manganese naphthenate in iso-octane. The instrument is placed in operation, and the oxyhydrogen flame adjusted for optimum maiiganese emission. The slit width is set a t approximately 0.02 mm. A portion of the manganese reference standard is aspirated into the flame and the wave length setting is adjusted to 403.3 mp by noting the point of maximum manganese emission. The concave mirror in the burner housing is adjusted until maximum manganese emission is obtained. The mirror position IS usually the same for Irad and rnanganese. The unburned portion of the reference standard is removed from the burner, and acetone is aspirated into the flame. A 5.0-ml. portion of the sample is diluted to 100 ml. with iso-octane. Because the final solution must contain less than 0.10 gram of manganese per gallon, samples containing more than 2.0 grams of manganese per gallon required a greater dilution. A portion of clear iso-octane is aspirated into the flame. With the shutter open, the transmittance scale is set a t zero and the dark-current control is adjusted until the galvanometer needle Coincides u ith the last fiducial mark to the left of zero on the galvanometer scale. The closed-shutter, dark current reading must not change during the subsequent steps. The unburned portion of iso-octane is removed from the burner and a portion of the manganese reference standard is aspirated into the flame. With the shutter open, the transmittance scale is set at 100% and the sepsitivity control is adjusted until the galvanometer needle coincides with the fiducial mark. A portion of the gasoline sample which was diluted with iso-octane is aspirated into the flame and the transmittance scale adjusted until the galvanometer needle returns to the fiducial mark. The reading on the transmittance scale is noted. The instrument need not be recalibrated before proceeding to the next sample. After five or six determinations, acetone should be aspirated into the flame to clean the capillary. The instrument should then be recalibrated. The manganese content of the gasoline sample is calculated as follows: CDE Manganese (grams per gallon) = 100
where C
=
D E
= =
manganese concentration in grams per gallon in the reference standard transmittance scale reading dilution factor EXPERIMENTAL
Lead. Gilbert (6) investigated the use of flame photometric techniques for determining tetraethyllead in gasolines. His studies showed t h a t lead emission in a flame source varied with the type of gasoline being analyzed,
but that the flame photometer n:is rapid and highly accurate when unleaded base stocks were available for the preparation of calibration standards. A method based on these considerations was described by Jordan (9). Gilbert suggested the use of flame background emission as an internal standard whrn unleaded base stocks \+-erenot available, but did not explore the feasibility of this approach. Burress and Grant (3) proposed a flame photometric method which did not require the unleaded base stocks for caSbration standards. These workers based their method on three observations: For all gasolines tested, flame background a t 402 mp was identical to that a t the wave length of lead emission a t 405.8 mp; lead emission a t 405.8 mp increased linearly with tetraethyllead concentration; flame background a.t 402 nip was not affcctod by tctr:wthyllcad additions. Thus, an incremcmtal addition of tetraethyllead to a portion of :m unknown gasoline provided a means of calibrating thr flame photometer for the analysis of the sample. In this method, a 400-ml. sample was specified and tvtraethyllead was added as the dilute fluid used in engine laboratories. A similar method was drscribcd by Linnd and Wulfken (11). Jentoft (6) developed a method which eliminated the need for unlcadrd bas? stocks and the. use of concentrated solutions of trtraethyllead. In this method a portion of the gasoline sample was diluted with kerosine containing a known amount of tetraethyllead. The added lead was then used to calibrate the flame photometer as described by Burress and Grant. Because the calibrating solution and sample were both kerosine dilutions of the gasoline, no base stock effect existed. Flame photometric methods for determining tetraethyllead have been studird rxtensively in these laboratories with particular attention to those proposed by Gilbert (@, Burress and Grant (S), and Jentoft (8). The nirthod proposed by (filbert (6) has the advantages of simplicity and s p e d The precision is inherently greatrr than that of procedures based on incrrmcmtal additions of tetraethyllead. In these latter mrthods, the length of the transmittance scale is in eflect reduced by the calibration procedure. Any error in transmittance readings is compounded in the calculations. Tmts of the Gilbert method substantiatrd his claims RS to the accuracy of the procedure whrn the unleaded base stock was available for the preparation of calibration standards. These trsts also vcbrified the author’s conclusion that standard srtmplrs prepared from one base stock could not be used t o dctermine tetraethylload in another base stock. VOL 31, NO. 11, NOVEMBER 1959
0
1799
In testing the Gilbert method, results obtained using oxyacetylene and oxyhydrogen flames, and burners with large-bore (0.012-inch diameter) and small-bore (0.009-inch diameter) capillaries were compared. Table I shows that the best precision was obtained using an oxyhydrogen flame and a burner with a small-bore capillary. A technique for correcting for base stock variations was studied. This technique is similar to that described by Bauserman and Cerney (1) for determining sodium and potassium in solutions containing sugar and alcohol. I n this method, a known amount of a metal other than the one being sought is added to the sample. The apparent concentration of the added metal is determined. The ratio of actual to apparent concentration of the added metal, called the concentration correction factor, is used to correct for base stock effects. I n this technique, it is assumed that the base stock has the same effect on the metal being sought as on the added metal. This technique was studied by adding a known amount of tetracthyltin to base stocks under test. Analytical curves were established for tin and tetraethyl-
Table 1.
Burner Capillary
Diameter, Inch 0.012
0.009 0.009 0.012 0.009
0 009
Table 11.
Base Stock
Effect of Burner Type on Lead Determination
Tetra-
ethyllead Concn. Range, Std. hll. per Dev., % Flame Gallon Oxyacetylene 1 to G 8.0 Oxyacetylene 1 to 6 1.14 Oxyacetylene 1 to 3 1.04 Oxyhydrogen 1 to 6 0.85 Oxyhydrogen 1 to 0 0.61 Oxyhydrogen 1 to 3 0.75
lead in iso-octane. Tin emission was measured at 303.4 mp. Results obtained from other base stocks were referred to these curves t o obtain apparent concentrations of tin and tetraethyllead. Apparent tetraethyllead concentrations were then multiplied by the tin concentration correction factor t o obtain corrected tetraethyllead values. The effect of applying correction factors is illustrated in Table 11. While results were improved by this correction, base stock effects were not entirely e l i i nated. Tests of the Burress and Grant (3) method substantiated the claims of the authors as to its accuracy and its a p plicability to the determination of tetraethyllead in all base stocks. It was felt, however, that the large sample size (400 ml.) and the necessity for handling hazardous concentrations of tctraethyllead in the dilute fluid were serious disadvantages. Results obtained by this method are compared in Table 111 with results obtained by the Gilbert method. Tests of the method proposed by Jentoft (8) showed that the kerosine diluent contributed to burner capillary fouling and gave a samewhat unsteady flame. Burner capillary fouling was less severe when iso-octane was used as the diluent. Burner capillary fouling from hydrocarbon components in gasolines proved to be the major obstacle to the development of a satisfactory method. I n the course of making only a few determinations, the burner became sufficiently fouled to affect precision and accuracy seriously, although the flame might appear normal, and no serious loss of sensitivity might occur. This persistent difficulty was eliminated by aspirating acetone through the burner into the flame a t frequent intervals. The effect
Determination of Tetraethyllead Using Concentration Correction Factors
Tin Added, Grams per
Liter
c
0.5105 0.5105 0.5105 0.5305 0.5105 0 . 5 105 0.5105
c' C
c
.4 .I A
Correction Factor 0.9132 0 9132
0.9132 0,9132 0.8720 0.8720 0.8720
Table 111.
Tetraethyllead, hI1. per Gallon Apparent Corrected Added 3.42 2.43 1.65 0.90 3.54 2.28 1.14
3.12 2.22 1.51 0.82 3.08 1.99 0.99
3.00 2.08 1,39
0.70 3.01 1.82 0.91
Dev., % 4.0 6.7 8.6 17.2 2.3 9.3 8.8
Comparison of Gilbert Method with Burress and Grant Method Gilbert Method Burress and Grant Method Tetraethyllead, Av. dev., ml. Av. dev. ml. JII. per Gallon per gallon Av. dev., yo per gallon Av. dev., 7c 0 01 1.00 0.7 0 01 0 8 1 5 0 01 0.7 0.03 2.02 0 03 1 9 3.00 0.8 0.06 1 4 0 02 0.5 0.06 4.00 4 0 0 02 0 20 0.4 5.00 1 9 0 02 0.6 0.07 Av.
1800
ANALYTICAL CHEMISTRY
of burner capillary fouling is illustrated in Figure 1. The time required for each set of six determinations was approximately 5 minutes. Figure 1 shows that the burner capillary fouled sufficiently in less than 3 minutes to affect the results seriously. Some fuels are very prone to foul the burner, but acetone was an effective cleaner for all gasolines encountered. Studies were made of lead emission a t 368.3 and 405.8 mp. Emission at 368.3 mfi was not linear with respect to lead concentration for tetraethyllead concentrations over 2.5 ml. per gallon. Emission at the 405.8 mfi wave length was linear for tetraethyllead concentrations as high as 25 ml. per gallon. Because the determination of tetraethyllead by methods employing incremental additions of lead to the sample depends upon a linear relationship between lead emission and lead concentration, the 405.8mp line was considered more suitable. Under the conditions chosen flame background intensity from 400 to 402 mp was independent of lead concentration and equal to the background intensity a t the wave length of the lead line for gasolines containing only tetraethyllead. However, manganese emission interfered a t 402 mp. For this reason bnckground intensity measurements were made a t 400 mp. The effect of other elements upon lead emission a t 405.8 mp was studied. Manganese interfered, but only when tetraethyllead concentrations of less than 0.35 ml. per gallon wcre being determined. Low concentrations of tetraethyllead can be determined in thr presence of manganese by measuring lead emission a t 368.3 mp and flame, background emission a t 3iO mp. Other elements commonly found in gasolines. such as phosphorus, halogens, anel sulfur, do not interfere. The need for handling concentrated solutions of tetraethyllrad to preparr the lead reference standard was considered a srrious disadvantage for any routine method. Consrquently, othw suitable lead compounds wcrc sought. Two compounds, lcad nal hthenate and lead di-n-butyldithiocdxmatc, werc? satisfactory. Standards containing lcad naphthenate were studird cstcnsively. Thrse studies showcd that lrad naphthrnate and tetracthyllend were equally satisfactory. Manganese. Preliminary studies were made of mangancse emission i n the flame. The manganese lines at 403.1, 403.3, and 403.4 mp appear as a singlo strong line at 403.3 mp making it ~io~sihlcto detcct estrrnicly small amounts. For mnngiincse concc.ntr:ttions lx~low0.25 gram pcr gallon. innngancse conccntr;ition \;-as dircct1)- 1 rol'ortionid to (,mission nt 403.3 1111. G:isolinc, base stock had a markvd oCe.ct upon mangmcw emission, but a t\r.c,ntxfold
2,501 I
1
I
I
I
I
I
1
lev., MI.
2.48
Ethyllead per Gallon0 ing Instrumental 0.0135 0.00706 1 .OO 0.0205 0.00873 2.00 0.0263 0.0130 3.00 4.00 0.0303 0.0175 0.0343 0.0218 5.00 6.00 0.0380 0.0257 a SM. dev. of mean of duplicate determinations made on single dilution.
I I
2.46 -
I
i
I
W 4
2
I 2.44-
d I2.42
-
2.40
-
2.38
I
1
I
2 3 4 5 SUCCESSIVE DETERMINATIONS
I
I
I
1
I
6
Figure 1. Effect of capillary fouling on precision of lead determinations A. 6.
Table IV. Comparison of Sampling and Instrumental Errors in l e a d Determination
Burner capillary cleaned after each determlnation Burner capillary not cleaned
dilution with iso-octane minimized these effects to a point where they could be tolerated. Tests of a large number of different organomanganese compounds showed that all had identical emission characteristics. Because of their stability over long periods of time, isooctane solutions of manganese naphthenate were used as reference standards with which to calibrate the instrument. The effect of other elements upon manganese emission at 403.3 mp was studied. Elements commonly found in gasolines, including lead, do not interfere. An oxyhydrogen flame was used for both lead and manganese determinations because of its cleanliness and low flame background. The flow rates for oxygen and hydrogen were determined experimentally for each burner used and were the same for both lead and manganese emission. SOURCES OF ERROR
Instrumental. Gross errors resulted unless the flame photometer was kept at optimum operating conditions. Best operation 11as obtained when the instrument was kept turned on a t all times. A 4-hour warm-up period was usually necessary to eliminate a drift in readings. Burner capillary fouling resulted in nonreproducible transmittance readings and an eventual loss of sensitivity. Small I articles n hich sometimes became lodged in the oxygen nozzle, causing an abnormal flame appearance and loss of sensitivity could usually be flushed out by running tap nater into the ouygen inlet and through the burner. A varnish coating which sometimes formed around
the rim of the hydrogen nozzle could be removed with a 1% nitric acid solution followed by thorough rinsing with water. If all the dry cells were not a t full strength, poor reproducibility and a loss of sensitivity were observed. With continued use, the sensitivity control potentiometer contacts became dirty, and caused erratic galvanometer fluctuations while sensitivity settings were being made. Alcohol was satisfactory for cleaning the contacts. Sampling. Errors associated with sampling, preparing the sample dilutions, and d a y t o d a y variations in operator skill contributed more t o t h e total error than errors associated with t h e instrument. A comparison of errors from different sources is given in Table I V for t h e lead determination. Similar results were obtained for manganese. Sample dilutions had to be accurate, and the effect of temperatures on sample volume had to be considered. It was advantageous to perform the dctcrminations in an air-conditioned laboratory.
Three flame photometric determinations were made fgr each set of dilutions. Estimates of the error between sets of dilutions, 'the error associated with the instrument, and the error associated with sampling were calculated for each tetraethyllead concentration level using the following expressions:
/n,nl-
B =
where A is an estimate of the error between sets of dilutions for a particular tetraethyllead concentration level B is an estimate of the instrumental error C is an estimate of the sampling error i refers to the determinations j refers to the dilutions X denotes the result of a single - determination x denotes the average of the indicated determinations n, is the numhrr of sets of dilutions per sample n2is the number of determillations made on each set of dilutions n3 is the number of dctcrminations made on each sample
PRECISION AND ACCURACY
Lead.
Precision and accuracy were estimated from the results of 756 determinations made on 18 standard samples. These samples covered tetraethyllead concentrations from 0.1 t o 6.0 ml. per gallon and included three widely different gasoline base stocks. A total of 378 determinations were made using a reference standard containing lead naphthenatc and an equal number with one containing tetraethyllead. I n each case, seven sets of dilutions were made for each sample.
The precisions obtained for the anal!.sis of different base stocks were compared by the F trst ( I d ) . Gasoline 1)nsv stock had no significant effect on the Iirecision of the method (Table V). The F test also s h o w d that lcad naphthtinate was as satisfactory as tctracthyllcad for preparing the lead rcfercncc stantlard (Table VI). Bccausc the method was not significantly biastd, it n.as concludctl that accuracy was equal to the prccision. Precision and accuracy of the method are summarized in Table VII. The flame photometer method apVOL 31, NO. 1 1 , NOVEMBER 1959
1801
pears to be somewhat leaa precise than x-ray duoresccnce procedures, giving a standard deviation of approximately 0.035 ml. of tetraethyllead per gallon as compared to a standard deviation of 0.028 ml. of tetraethyllead per gallon reported by Lamb, Niebylski, and Kiefer (IO) for x-ray fluorescence determinations. The flame photometer method appears to be comparable t o x-ray absorption procedures in precision. Unlike x-ray absorption methods, which require corrections for the sulfur content of the gasoline, the flame photometer method appears to be free from serious sample matrix effects. The flame photometer approaches x-ray methods in speed, allowing about 40 determinations per man-day. The Gasoline Testing Division of Ethyl Corp. exchanges gasoline samples regularly for comparison of analytical results. Tetraethyllead in these samples is determined by both chemical and x-ray absorption methods. A large number of these exchange samples have also been analyzed by the flame photometer and the r mlts compared with those obtaincd by the other methods (Table VIII). On the avrrage, flame
photometric results agreed with chemical values within 0.04 ml. of tetraethyllead per gallon, and with x-ray absorp tion values within 0.03 ml. of tetraethyllead per gallon. Manganese. The precision of t h e manganese method was estimated from t h e results of 252 determinations made on 12 synthetic standard samFles. These standards were prepared t o contain known amounts of (methylcyclopentadienyl) manganese tricarbonyl in three widely different gasoline base stocks. A manganese concentration range from 0.1 to 1.0 gram per gallon was covered in each base stock. Seven dilutions were made of each sample. Three flame photometric determinations were made for each dilution. Estimates of the error between dilutions, the error associated vi-ith the instrument, and the error associated with sampling were calculated for each manganese concen-
ACKNOWLEDGMENT
Table VII.
Tetraethyllead Concn., M1. per Gallon 0 1 0.5
...
Tetraethyllead Concn., MI. per Gallon
Mn Concn., Grams per Gallon
2.0 3.0 4.0 5.0 6.0
... ...
5.50 4.20 2.80 1.50 0.50 0.20
...
...
... ... ... ...
... ... ...
1.0 0.5 0 3 0.1
...
F Ratioxz/yt xz/zt 1.55 2.28 0.81 1.21 2.78 1.09 0.43 1.03 1.69 1.93 2 03 1.88 3.24 3.55 1.86 1.26 2.40 2.34 0.91 0.75 are 2.22 (5% level)
Critical values for F and 3.14 (1 % level). x* is variance for base stock 2. y’ is variance for bme stock y. z* is variance for b u e stock z.
Table VI. Lead Naphthenate vs. Tetraethyllead as Reference Standard
Tetraethyllead Concn., MI. per Gallon
F Ratio u11/02=
5.50 4.20 2.80 1.50 0.50 0.20
1.18 1.15 1.37 1.55 1.23 0.64 Critical value for F is 1.58 (5% level). u,*. u2*.
Variance with lead naphthenate. Variance with tetraethyllead.
1802
1 .o
ANALYTICAL CHEMISTRY
... ... ... ...
Precision
Manganese Estimated Errop 70 of Concn., Gram Metal Present, Per (95% ConfiGallon dence Limits)
Effect of Base Stock on Precision
Table V.
tration level using Equations 1,2, and 3. The precision of the method when different base stocks were analyaed was cornpared using the F test (14). Gamline base stock appeared to have a slight effect on precision in only one case (Table V). Precision of the method is summarized in Table VII. In the case of one gasoline base stock, a significant bias was noted when concentrations of manganese from 0.25 to 0.50 gram per gallon were determined. This bias was in the positive direction and amounted to about 2% of the amount of manganese present. Accuracy, within 95% confidence limits, for the manganese concentration range from 0.25 to 0.50 gram per gallon was within 5% of the amount of manganese present when the bias was included. Therefore, the bias was not considered to be serious. In all other cases, accuracy was approximately equal to the precision.
... ..
...
... ...
0.10 0.25 0.50 1 .o
19.2 4.88 3 17 2.32 2.12 1.93 1.82 1 75 5.94 3.14 2.20 2.20
Estimated error of mean of duplicate determinations made on single dilution. y
Table VIII. Comparison of Flame Photometer with Chemical and X-Ray Absorption Methods
(Results, ml. of tetraethyllead per gallon) X-Ray Flame Chemical0 Absorption“ Photometerb 1.15 1.75 1.97 2.03 2.59 2.60 2.86 2.90 2.92 3.14
1.10 1.76 1.93 2.00 2.54 2.60 2.81 2.91 2.91 3.04
1.13 1 69 1.94 2.02 2.54 2.57 2.81 2.9% 2.92 2 . 9gd
a Average of results reported by five laboratories. b Average of duplicate determinations from one laboratory. e Greatest disagreement between flame photometric and x-ray absorption results observed in 40 samples. d Greatest disagreement between flame phqtometric and chemical results observed in 40 samples.
The authors express their appreciation to W. C. Healy and G . W. Thomson of the Ethyl Research Labomtones for their aid in thc stat.istica1 analysis of the data. LITERATURE CITED
(1) Bauserman, H. M., Cerney, R. R., Jr.. ANAL.CHEM.25, 1821 (1953). (2) Borup, R., Levin, H., Am. Soc. Testrng Materials, Proc. 47, 1010 (1947). (3) Burress, G. T., Grant, J. A., Generai
Papers, Division of Petroleum Chemistry, 122nd Meeting, ACS, Atlantic City, N. J.: September 1952. (4) CalingaeTf, G., Lamb, F. W., Miller, H. L., Noakes, G. E., ANAL.CHEM. 22, 1238 (1950). (5) Frediani, H. A., Bass, L. A., Oil Gas J . 39, 51 (1940). (6) Gilbert, P. T., Jr., Am. Soc. Testing Materials, Spec. Tech. Publ. No. 116, 77 (1952). (7) Hansen, K. A., Parks, T. D., Lykken, Louis, ANAL.CHEM.22, 1232 (1950). (8) Jentoft, R. E., “Determination of
Tetraethyllead in Gasoline by the Flame Photometer,” California Research Corp. Laboratory commimica- Rept., - . private tion. (9) Jordan, J. H., Jr., Petrol. Refiner 32,
139 (1953). (10) Lamb, F. W., Niebylski, L. hl., Kiefer. E. W.. ANAL. CHEM.27. 120 ( 1955)‘. (11) Linn6, V. W., Wulfken, H. D., E‘rdol u. Kohle 10, 757 (1957). (12) Offut, E. B., Sorg, L. V., ANAL. CHEM.22, 451 (1947). (13) Swanson, B. W., Daniels, P. H., J . Znst. Petrol. 39, 487 (1953). (14) Youden, W. J., “Statistical Methode for Chemists,” pp. 29-32, Wiley, New York, 1951.
RECEIVED for review February 13, 1959. Accepted July 16, 1959. Pittsburgh Conference on Analytical Chemistry and Applied Spectroscopy, Pittsburgh, Pa., March 1959.