expected, an essentially linear relationship existed between the degraded and the RON. The line drawn pyrolytic wt through the points in Figure 3 was determined by a leastsquares fit from the empirical data. Thus, vapor-phase F'GC might be a convenient laboratory technique to allow a rapid octane quality survey of pure hydrocarbon compounds and blends of these compounds. degraded With this interesting relationship between wt and RON established, we attempted to duplicate the combustion chamber environment of the motored engine in the pyrolysis chamber by using a 20 vol 02/80 vol Nf mixture as the carrier gas for the pyrolysis of 2,2,4-TMP. The 680 OC pyrolysis temperature was near the peak-cycle temperature required for maximum degradation of 2,2,4-TMP in the motored engine (14). The chromatogram obtained from pyrolytic oxidation of 2,2,4-TMP under these conditions is shown in Figure 4. This chromatogram was obtained with only the 62.0-m capillary column to allow better correlation with the earlier work. The chromatogram in Figure 4 has a peak by peak correlation with a chromatogram published in 1967 for the preflame reaction products trapped from the exhaust of the motored engine (Ref. 14, Figure 3). The peak identities in Figure 4 are given by number in Table X, as well as the mol %composition of the pyrolysis products.
tion of the specific pyrolysis products in the C& range. Th,ese specific hydrocarbon identifications have enabled us to apply the Kossiakoff and Rice theory of thermal degradation to determine to what extent the vapor-phase pyrolyses follow predicted thermal degradation routes. The pyrolyses were carried out at 680 "C to keep the amount of parent compound degraded to less than 10 wt %. Comparison of empirical data with the products predicted by theory was good. A linear relationship between the amount of the parent molecule degraded in the pyrolysis chamber (wt and the Research Octane Number of the TMP isomers has been established. In addition, pyrolysis products of 2,2,4-TMP in a 20 vol % 02/80 vol % Nz atmosphere have been found to be identical with those formed in a motored engine, which suggests that the octane-related preflame degradation of fuels and fuel blends might be rapidly studied in the laboratory.
CONCLUSIONS
RECEIVED for review April 19, 1971. Accepted June 18,1971. This research was conducted in part under the McDonnell Douglas Independent Research and Development Program.
x
High-resolution capillary gas-liquid chromatography used in conjunction with vapor-phase pyrolysis allows identifica-
x)
ACKNOWLEDGMENT
The authors acknowledge the assistance of W. W. Haskell of Shell Oil Company and J. D. Kelley and C. J. Wolf of McDonnell Douglas Research Laboratories for helpful suggestions in preparing this manuscript.
Application of the Carbon Rod Atomizer to Atomic Absorption Spectrometry of Petroleum Products K. G . Brodie and J. P. Matoukek Varian Techfron Pty. Ltd., North Springvale, Victoria, Australia 31 71
Direct determination of AI, cup Cr, Mg, Ni, and Pb in lubricating Oils at concentrations UP to 15 Pg/ml was performed by atomic absorption using non-flame atomization. oil samplesof 0.5-1.0 P l were injected into a PraDhite mini-furnace made bv drillina a transverse hole through a graphite rod. The analy'Sis was carried out by a preset drying, ashing, and atomizing sequence which took about 30 seconds to complete. Samples were applied during the drying cycle and atomic absorption was measured during the atomizing cycle. Working curves showed good agreement between oil and aqueous standards. Higher viscosity Oils were diluted with xylene prior to analysis. Relative standard deviation was better than 4%.
ATOMIC ABSORPTION (AA) analysis of metals in oils, greases, and fuel is being widely and effectively used throughout the industry (1-3). Emission spectrographic methods (4, 5 ) providing direct readout analysis for a number of elements in (1) J. A. Burrows, J. C. Heerdt, and J. B. Willis, ANAL.CHEM., 37, 579 (1965). (2) E. J. Moore, 0. I. Milner, and J. R. Glass, Microchem. J., 10, 148 (1966). ( 3 ) R. A. Mostyn and A. F. Cunningham, J. Inst. Petrol., 53, 101 (1967). ( 4 ) J. J. Lowe and R. J. Martin, Appl. Spectrosc., 23, 587 (1969). ( 5 ) C . W. Key and G. D. Hoggan, ANAL.CHEM., 25,1673 (1953).
oil have the advailtage that often no pre-treatment of sample is required. Both techniques are extensively used for detecting traces of metals in lubricating oils, in oil distillates, and metal additives in oils and greases. I n conventional flame AA, organometallic standards in an organic solvent of viscosity close to that of the sample must be used to construct the analytical working curve because of the differing aspiration rates (and hence atomic absorption observed) of oils or organic compounds of varying composition. The atomic absorption observed in organic solvents is considerably greater than that for the Same element in aqueous solution, and furthermore the sequential aspiration of petroleum and aqueous samples is excluded because of the nature of the spray chamber and differencesin flame stoichiometry. The recent development of the carbon rod atomizer (6-8) suggests possibilities for rapid analysis of metal in petroleum compounds without prior ashing or other treatment except perhaps dilution. This paper describes the analysis by carbon rod atomizer of a number of important elements in various lubricants and compares the results with those for aqueous standards. (6) M. D. Amos, P. A. Bennett, K. G. Brodie, P. W. Y . Lung, and J. P. MatouSek, ibid., 43, 211 (1971). (7) J. P. MatouSek, Anzer. Lab., June, 45 (1971). (8) J. P. MatouSek and B. J. Stevens, Clin. Chem., 17, 363 (1971).
ANALYTICAL CHEMISTRY, VOL. 43, NO. 12, OCTOBER 1971
1557
-
0.05 ABSORBANCE
SAMPLE ADDED
Figure 1. Recorder trace for 0.5 pl of oil containing 11.2 pg/ml of aluminum -+ Atomic absorption peak Table I. Preparation of Oil Samples Concentration Element Compound range, pg/ml Base Ag Silver cyclohexane0.2- 5 Jet oil butyrate A1 Aluminum cyclo0.4- 10 Jet oil hexanebutyrate Cu Copper acetylace0.8- 10 Jet oil tonate Cr Chromium acetyl0.6- 10 Jet oil acetonate Mg Magnesium cyclo0.6-2 Jet oil hexanebutyrate Ni Nickel naphthenate 0 . 2 - 15 Jet oil Pb Lead naphthenate 0.6- 10 Jet oil V ... 2-20 Fuel oilQ Zn Zinc dialkyldithio200- 800 Jet oil phosphate ... 500- lo00 Motor lubricating oila Analyzed industrial oil samples. EXPERIMENTAL
Apparatus. All measurements were made with a Varian Techtron AA-5 spectrophotometer used in conjunction with Varian Techtron Model 61 carbon rod atomizer complete with power supply and gas control unit, details of which are given elsewhere (8). The carbon rod was Ringsdorff high purity graphite grade RW 1. Nitrogen was used as the inert gas but the effects of using helium and argon were also studied. A slit height adjustment prevented continuum radiation from the heated carbon rod entering the monochromator, while readout was accomplished by recording peak heights with a G2000 Varian 10-mV chart recorder. In some instances an electronic integrator served to record peak area. Varian Techtron hollow cathode lamps were operated at or slightly above the recommended current depending upon the required intensity. A 5-pl syringe with a gold-plated stainless steel plunger was used to inject samples into the carbon rod. Monochromator entrance and exit slits were kept between 25 and 50 p (spectral band width 0.08 to 0.17 nm) while the standard photomultiplier was replaced with an HTV R106 tube. Reagents. All inorganic reagents were analytical grade (nitrates and sulfates) and were prepared in water once distilled in a borosilicate glass vessel. The lubricating oil was a synthetic grade Mobil Jet Oil 11(Mobil Oil Corp., New York) and the element concentrations in this oil were adjusted to the desired levels by adding organometallic compounds of the individual elements. The compounds used for this purpose appear in Table I, together with the concentration range in the oil which was analyzed. Stock solutions of the elements in the appropriate solvents (toluene, xylene, or chloroform) 1558
were diluted with the oil to give a 5-10Z v/v solvent-to-oil mix to achieve the appropriate metal concentration. Solutions of the oil with the same concentrations of solvent were also used as blanks. The freshly prepared stock solutions of the organometallic standard were analyzed by flame atomic absorption after dry ashing. Samples of industrial fuel oil containing vanadium and motor lubricating oils were also analyzed independently by conventional flame AA using standards prepared in the appropriate organic solvents. The motor lubricating oil was Mobil grade SAE 30. Solvents used for the dilutions were AR grade. Operation. A 0 . 5 ~ 1sample of freshly prepared jet lubricating oil was syringed directly; the fuel oil sample containing vanadium was used undiluted and also diluted 5 0 x v/v with xylene. The lubricating oil containing zinc was diluted similarly. Two methods may be used to apply aqueous samples. In the first method, the carbon rod was impregnated with 3 p1 of xylene before a 0.5-pl sample was added. This prevented the aqueous solution from soaking into the rod and yielded more reproducible results than by injecting the sample onto a cold rod (8). It was evident from the results obtained with and without xylene impregnation for aqueous samples, that the extent to which the solution soaks into the porous graphite considerably alters the peak signal obtained. After permitting an aqueous sample of aluminum (0.5 p1 of 10 pg/ml) to soak into the graphite to varying degrees (by applying the sample to a rod with temperature varying between 20 and about 100 "C), both the integrated and the peak absorbance signal were recorded during the atomization step. This investigation revealed that the total atom population decreased when the solution soaked into the graphitei.e., the ratio of integrated/peak signal remained practically unchanged. As an alternative the aqueous solution was syringed into the sample cavity while the drying step was operating, thus slowly evaporating the solvent upon contact with the carbon surface. Both yielded very similar results with the latter being more convenient. Soaking of the oil into the rod was not minimized (and hence peak signal was not improved) by first impregnating the rod with water. The oil samples were injected onto the rod directly during the drying step. RESULTS AND DISCUSSION
The CRA Power Supply automatically performs the sequential drying (for evaporation of solvent), ashing (for removing organics in oil samples), and final atomizing steps (for actual analysis) according to the preset conditions. The correct parameters of voltage and time for the atomization step were selected for each element using aqueous solutions, while the same conditions of drying and ashing applied to all samples once established. A chart recording for aluminum in oil (Figure 1) shows how the complete analysis of the sample including ashing was achieved in about 30 seconds. The ashing products were responsible for the large non-atomic peak. Working curves were constructed for aqueous solutions of Ag, Al, Cu, Cr, Ni, and Pb in concentrations up to 15 pg/ml. Typical calibrations for aluminum and nickel are shown in Figures 2 and 3. However, the sensitivity for silver was so great that the working curve was useful only up to 5 pg/ml unless the sample was diluted oi a reduced volume used. The sensitivity for magnesium was also very high and in order to obtain a useful working curve without dilution of the sample, the carbon rod was lowered and the optical path viewed above the rod. This reduced the sensitivity by about 200-fold but bending of the working curve is increased-the extent of curvature depending on the height above the rod at which the measurement was made (7).
ANALYTICAL CHEMISTRY, VOL. 43, NO. 12, OCTOBER 1971
O'*I 0.7
0
1
2
3
4
5
6
1
8
8
*
)
ALUM,NUM (w/mt)
Figure 2. Analytical working curve for alumhum at 309.28 nm using a constant sample volume of 0.5 p1 of aqueous standards When an element concentration in a sample is too high for the analytical sensitivity of the most suitable resonance line, analysis may still be carried out by the above method, but it is preferable to carry out a dilution step as this gives a more linear calibration. The working curves constructed for aqueous solutions were used for the evaluation of the analyses of the same volume of oil samples and some of the results are shown in Table 11. Zinc is present as an additive in lubricating oils in the low percentage range (up to 0.1 %) and has been analyzed using its least sensitive line (307.59 nm) in samples diluted 50% v/v with xylene. The viscosity of the oil (grade SAE 30) was such that dilution with xylene was desirable to facilitate syringe operation. It was difficult to analyze vanadium in fuel oil (2-20 pg/ml level) partly because of incomplete vaporization of the metal from the rod. In addition carbon particles formed at the high temperature required to atomize the sample caused nonatomic absorption and also reduced the number of samples which could be analyzed on one rod. A detection limit of about 1 pg/ml from a 0.5-pl sample was obtained. When the unevaporated vanadium was removed from the rod by heating twice prior to the next sample application, good agreement with flame atomic absorption results was obtained at higher vanadium levels (Table 11). At these higher vanadium concentrations, the atomic absorption peak is not affected by the non-atomic absorption peak which immediately follows it. It was observed that while the rod was new, the non-atomic absorption peak was relatively insignificant; as the rod aged however, the peak increased in height to such an extent that replacement of the rod became necessary. The detection limits obtainable using a 0.5-pl sample at the most sensitive resonance lines are as follows: Ag O.OOO4, A1 0.06, Cu 0.02, Cr 0.01, Mg O.OOO1, Ni 0.02, Pb 0.01 pg/ml. As shown in Table 11, the most sensitive resonance lines have not necessarily been used since they were too sensitive for some determinations. The useful range of the method may be extended by choosing a less sensitive resonance line; alternatively the absorption can be viewed above the rod thus decreasing the sensitivity by approximately two orders of magnitude . When the concentration of an element appears too low for the analytical sensitivity of the most sensitive resonance
0
2
4
B
e io 1 2 NICKEL (pg/rnt)
1 4 1 6
Figure 3. Analytical working curve for nickel at 232.00 nm using a constant sample volume of 0.5 pl of aqueous standa rd s
Table 11. Results of Oil Analysis by the Carbon Rod Atomizer
Ag
Wavelength, nm 338.28
A1
309.28
cu
327.40
Cr
427.48
Mg
285.21
Ni
232.00
Pb
283.30
V O
318.54
Zn
307.59
Element
Concentration determined (using aqueous Concentration standards), added, pg/rnl rglml 2.0 2.0 5.0 5 .O 2.2 2.1 10.0 10.3 2.0 2.0 10.2 10.0 1.7 1.6 8.1 8.0 0.6 0.6 2.0 2.1 1.5 1.5 7.6 7.5 1.7 1.7 7.3 7.5 5.8 5 .O 13.1 13.3 18.1 18.5 403 400 800 795
Analyzed industrial fuel oil.
line, the sensitivity of the method may be improved by simply increasing the sample volume. The signal increase however was not linear as shown in Figure 4. For both aqueous and oil solutions the sample was added while the drying step was operating and up to 5 pl has been applied in this manner. When constructing working curves, it is preferable to use a constant volume of solution of variable concentrations, rather than differing volumes of a constant concentration solution. In the latter case, increased curvature will result. In addition, precision of the method is better when a constant volume is used. For the carbon rod atomizer, a sample volume of 0.5-1.0 pl would be the optimum as it provides sufficient sensitivity. Further, the relative standard deviation calculated from ten replicate measurements at 2 pg/ml level in oils using 0.5 pl
ANALYTICAL CHEMISTRY, VOL. 43, NO. 1 2 , OCTOBER 1971
1559
"'I
Figure 4. Analytical working curves for lead at 283.30 nm and zinc a t 307.59 nm using a constant concentration of 2.5 pg/ ml for lead and 200 pg/ml for zinc Aqueousdnc 0 Lead inoil was found to be: for Ag 1.9%, A1 2.6%, Cu 1.7%, Cr 2.4%, Mg 3.1 %, Ni 1.9%, and Pb 2.1 %. For Zn, the relative standard deviation was found to be 2.3% at the 400 pg/ml level and for V 4.3 at the 5 pg/ml level. The effect on the peak absorbance of using helium, nitrogen, and argon as the protecting inert gas was examined for aluminum and lead. For lead the absorbance was practically the same in nitrogen, argon, and air (i,e., no gas flowing), while in a helium atmosphere the absorbance was about 6 0 z lower. With helium flowing at less than about 2 l./min, the absorbance increased with decreasing helium flow. The absorbance value reached that obtained in nitrogen when the helium flow finally reached zero. In nitrogen and argon the absorbance was practically independent of the flow rate. With aluminum, the ratio of the peak absorbance in helium, nitrogen, and argon was 1 :1.5 :3, respectively, when the flow rate of each gas was about 1.5 l./min. As with lead, the peak absorbance eventually reached the value obtained in air (or nitrogen) as the helium or argon flow was decreased to zero. This increased absorbance for aluminum in argon compared with nitrogen has been observed with a graphite tube through
z
1560
which the inert gas was continuously flushed (9). The authors considered that possible aluminum nitride formation reduced the free atom population. Our results would indicate that the main cause was different rates of diffusion of the metal atoms from the sample cavity. The rate of diffusion depends on the atomic weight of the metal and the nature of the gas (IO). Chemical interferences experienced with the carbon rod are widely different from those observed with conventional flame AA (6, 7). If compounds of high volatility are formed (e.g., PbClz), they will be partly removed from the rod before dissociation can occur with resultant loss in peak signal. Further, the presence of anions which form a less volatile metal compound will result in an increased peak area which will be evident when integrated peaks are measured. Measurement of the integrated absorbance signal as well as peak absorbance signal will ascertain whether the vaporization rate of the analyte had altered in the presence of the oil matrix or whether the atom population had changed because of matrix interference. The comparison of peak and integrated measurements for aqueous solutions with those for oil samples containing the same concentration of the analyte showed very close agreement. This indicates that neither chemical interference nor a change in vaporization rate was present when these oil samples were analyzed. The method could provide direct analysis of trace metals in petroleum products using aqueous standards. No chemical pretreatment of the oil is required and dilution would be necessary only in a few cases. The method is also very rapid as a complete single analysis of the oil sample can be achieved in about 30 seconds.
ACKNOWLEDGMENT We thank Mr. J. B. Sanders for the analysis of many oil samples by conventional flame methods. RECEIVED for review March 29, 1971. Accepted June 21, 1971,
(9) D. C . Manning and F. Fernandez, At. Absorption Newsleft., 9, 65 (1970). (10) B. V. L'vov, Specfrochim. Acta, 17,761 (1961).
ANALYTICAL CHEMISTRY, VOL. 43, NO. 12, OCTOBER 1971