Anal. Chem. 2003, 75, 2618-2625
Automated Flow Injection Analyzer with On-Line Solid-Phase Extraction and Chemiluminescence Detection for the Determination of Dodecylamine in Diesel Fuels Philip J. Fletcher,†,§ Kevin N. Andrew,*,† Stuart Forbes,‡ and Paul J. Worsfold†
Department of Environmental Sciences, Plymouth Environmental Research Centre, University of Plymouth, Drake Circus, Plymouth PL4 8AA, U.K., and Analytical Technology Business Group, Shell Global Solutions, Cheshire Innovation Park, P.O. Box 1, Chester CH1 3SH, U.K.
This paper describes the development of a portable, automated flow injection-chemiluminescence (FI-CL) analyzer incorporating on-line solid-phase extraction (SPE) for the determination of dodecylamine (detergent) in diesel fuels. The method is based on the peroxyoxalate/ sulforhodamine 101 chemiluminescence reaction, with SPE required to remove indigenous compounds within the diesel fuel matrix that interfere with the CL response. The automated analyzer achieved a detection limit of 2.9 mg L-1 and a linear range of 2.9-50 mg L-1, which was suitable for determinations of dodecylamine at levels typically present in fully formulated diesel fuels (40 mg L-1). Analyses of base fuels from five different sources demonstrated that an automated FI-CL-SPE system could provide a portable instrument for monitoring the presence/absence of dodecylamine in diesel fuels. Performance-enhancing compounds are added to fuels by manufacturers to achieve specific functions, for example, corrosion inhibitors,1,2 lubricity additives,2 detergents, friction reducers, flow improvers, dehazers, antifoams, and antioxidants. Additives take the form of distinct chemical entities (e.g., dodecylamine, which is used as a detergent additive in certain grades of diesel) or more complex formulations, including mixtures of structurally related compounds having a range of molecular masses. To ensure good fuel performance, it is important that the additives are present at the correct concentrations in the final product and to check that these additives, many of which are surface-active agents, are not being lost throughout the distribution chain. Because the chemical composition of additives used in fuels often differs from one manufacturer to another, their characterization can sometimes be used to differentiate fuels from different sources, and this feature can be employed to combat the problem of fuel counterfeiting which occurs in some countries. †
University of Plymouth. Shell Global Solutions. § Current address: Department of Chemistry, University of Pretoria, Pretoria 0002, Gauteng, South Africa. (1) Hayes, G. E. J. Chromatogr. 1990, 508, 259-64. (2) Black, B. H.; Wechter, M. A.; Hardy, D. R. J. Chromatogr. 1988, 437, 203-10. ‡
2618 Analytical Chemistry, Vol. 75, No. 11, June 1, 2003
At present, determination of fuel additives is usually performed in the laboratory, typically by chromatographic or spectroscopic methods. This necessitates the collection of samples from different points in the distribution chain (e.g., refineries, fuel terminals, or garage forecourts) that are then transferred to a central laboratory facility to await analysis. The process may entail a lengthy and potentially costly delay between sampling and analysis, and an appropriate portable analytical technique that could be used at the point of distribution would, therefore, offer clear advantages. This paper describes the development of an automated, portable flow injection-chemiluminescence (FI-CL) analyzer for the determination of the diesel fuel additive dodecylamine. An FI method based on the peroxyoxalate/sulforhodamine 101 chemiluminescence reaction3,4 is modified for the determination of dodecylamine in acetonitrile/water matrixes, following solid-phase extraction (SPE) of the additive from diesel samples. Univariate and simplex optimization of the method is described, with a detection limit (0.19 mg L-1) significantly lower than 40 mg L-1 (the typical dosing rate of dodecylamine in commercial fuels) achieved. Interferences from indigenous compounds present in diesel fuels are evaluated, and the effects of different solvents on the sensitivity of detection are investigated. The development and optimization of an SPE procedure to extract dodecylamine from diesel matrixes into a solvent that is compatible with the FI-CL system is also described. Gas chromatographic reference methods (using nitrogen-phosphorus or mass spectrometric detection) are used to validate the SPE procedure. Extracts of a number of diesel samples are analyzed by FI-CL, and the results are compared with those obtained using GC analysis. Following the optimization of both the FI-CL and SPE procedures, a fully automated analyzer incorporating on-line SPE is developed in order to produce a prototype field system suitable for deployment at fuel distribution points. Performing on-line SPE offers several potential advantages. The primary reason for incorporating SPE on-line is the ability to fully automate the analytical process. Automated off-line systems exist, but automation is generally achieved by the use of robots5 that are very (3) Katayama, M.; Takeuchi, H.; Taniguchi, H. Anal. Chim. Acta 1993, 281, 111-18. 10.1021/ac034058i CCC: $25.00
© 2003 American Chemical Society Published on Web 05/30/2003
expensive and not readily portable. Other advantages of on-line SPE are that the entire extract is analyzed, whereas with off-line SPE, a subsample of the extract is taken for analysis, and small volumes can be measured out very accurately. An SPE procedure can, therefore, be scaled down by using a smaller sorbent bed and less solvent, thus reducing the cost of extraction and the volume of waste that is produced. Flow injection with on-line SPE has previously been applied only to relatively simple aqueous matrixes, including determinations of pseudoephedrine in aqueous samples using capillary electrophoresis,6 lead in water samples using flame atomic absorption spectrometry,7,8 phenols in waters and wastewaters using spectrophotometic detection,9 caffeine in soft drinks using Fourier transform infrared spectroscopy,10 and bromofenoxim in water samples using electrochemical detection.11 The on-line SPE method described in this paper has been applied to diesel fuels, which are, in contrast to the above examples, complex and variable nonaqueous matrixes. Results for the determination of dodecylamine from a variety of diesel fuels using the automated system are presented and compared with those obtained using manual, off-line SPE with FI-CL detection. EXPERIMENTAL SECTION Reagents and Samples. Dodecylamine (98%) and sulforhodamine 101 hydrate (∼95%) were obtained from Aldrich (Gillingham, Dorset, U.K.). Bis(2,4-dinitrophenyl) oxalate (DNPO) (>97%), aniline (puriss grade), carbazole (purum grade), p-cresol (puriss grade), hexanoic acid (purum grade), indole (puriss grade), and phenol (puriss grade) were obtained from Fluka (Gillingham, Dorset, U.K.). Methanol, toluene (HiPerSolv grade), heptane, propan-2-ol, dichloromethane (AnalaR grade) and 30% (w/v) stabilized hydrogen peroxide were obtained from Merck (Poole, Dorset, U.K.). Acetonitrile, which must be free of residual amines to prevent reaction with DNPO/sulforhodamine, was purchased from Rathburns (HPLC grade; Walkerburn, U.K.). Analytical reagent grade (18.2 mΩ cm-1) deionized water was used throughout, obtained from an Elgastat UHQ II system (Elga Ltd., High Wycombe, Bucks., U.K.). Solutions of 1 × 10-3 M DNPO were prepared by dissolving 0.1056 g in 250 mL of acetonitrile (for the optimization, a 1 × 10-2 M stock solution was prepared). Stock solutions of 1 × 10-5 M sulforhodamine 101 were prepared by dissolving 0.0015 g in 250 mL of acetonitrile. Stock solutions of 0.1 M hydrogen peroxide were prepared by dissolving 2.58 mL in 250 mL of water (for the optimization, a 1 M stock solution was prepared). Stock solutions of dodecylamine (400 mg L-1) were prepared by dissolving 40 mg in 100 mL of 90% acetonitrile/10% water. All standards and reagents were prepared by serial dilutions of the above stock solutions. Solutions of DNPO, hydrogen peroxide, and sulforhoda(4) Katayama, M.; Takeuchi, H.; Taniguchi, H. Anal. Chim. Acta 1994, 287, 83-88. (5) Hoffman, K. L.; Andress, L. D.; Parker, T. D.; Guttendorf, R. J.; Rossi, D. T. Lab. Robot. Autom. 1996, 8, 237-42. (6) Chen, H. W.; Fang, Z. L. Anal. Chim. Acta 1997, 355, 135-43. (7) Naghmush, A. M.; Pyrzynska, K.; Trojanowicz, M. Talanta 1995, 42, 85160. (8) Sooksamiti, P.; Geckeis, H.; Grudpan, K. Analyst 1996, 121, 1413-17. (9) Song, W. L.; Zhi, Z. L.; Wang, L. S. Talanta 1997, 44, 1423-33. (10) Daghbouche, Y.; Garrigues, S.; Vidal, M. T.; de la Guardia, M. Anal. Chem. 1997, 69, 1086-91. (11) Svegl, I. G.; Ogorevc, B.; Novic, M.; Benfenati, E. Analyst 1996, 121, 183943.
mine 101 were stored in the dark when they were not being used. Sulforhodamine stock solutions were stable for several months when kept in the dark. DNPO and peroxide solutions were stable for 1 week. Stock solutions (1000 mg L-1) of potentially interfering compounds (aniline, p-cresol, carbazole, hexanoic acid, indole, and phenol) were prepared by weighing 100 mg into 100 mL volumetric flasks and diluting to volume with 90% acetonitrile/10% water. Serial dilutions were made as required. Diesel fuels (U.K. (Shell Haven), Sweden, U.K. (Stanlow), Germany (Hamburg), Holland (SNV), and Brazil) and a diesel fuel additive package were obtained from Shell Global Solutions (Cheshire Innovation Park, Chester, U.K.). Spiked diesel stock solutions (400 mg L-1) were prepared daily by dissolving 40 mg of dodecylamine in 100 mL of diesel fuel or heptane. Samples (40 mg L-1) were prepared by serial dilution. Fully formulated fuel samples (containing the additive package) were prepared by dissolving 85 mg of the additive package in 100 mL of diesel fuel to give the required 40 mg L-1 dodecylamine concentration in the fuel (dodecylamine is present at a concentration of 4.7% (m/v) in the additive package). All glassware was soaked overnight in 5% Decon 90 solution (Fisher Scientific) then rinsed with copious amounts of water and left to air-dry. Instrumentation. Manual FI-CL Manifold. A schematic diagram of the manual FI-CL manifold is shown in Figure 1. Chemiluminescence emission was measured using a CamSpec CL-2 detector (Cambridge Instruments, Cambridge, U.K.) containing a low-power, 12-V Hamamatsu photomultiplier tube with a quartz flow cell (path length 12 cm, volume 120 µL).12 The 0-10 V analogue output from the detector was acquired using a multifunctional DAQ-700 data acquisition/digital I/O card (National Instruments, Berkshire, U.K.) attached to a laptop computer. Software written in LabVIEW (National Instruments) was used to record and process the signal. A variable speed peristaltic pump (Gilson Minipuls 3, Villiers-le-Bel, France) was used to propel reagents, and a fixed speed peristaltic pump (Ismatec, Westonsuper-Mare, U.K.) was used for the sample stream. Ismaprene (Ismatec) pump tubing was used in all cases (1.02-mm i.d. for DNPO, sulforhodamine, and sample and 0.5-mm i.d. for H2O2 and carrier). A 6-port, 2-position rotary injection valve (Rheodyne 5020, Bensheim, Germany) was used to inject the sample into the carrier stream. All manifold tubing was 0.75-mm i.d. poly(tetrafluoroethylene) (PTFE) tubing (Fisher Scientific, Loughborough, U.K.). PTFE T-pieces (Omnifit Limited, Cambridge, U.K.) were used to merge reagent streams. All other connections were made using 0.25-in.-28 flanged polypropylene FI fittings (Omnifit). Solid-Phase Extraction. The SPE columns used in this work were 200-mg solid-phase, 3-mL reservoir Aminopropyl Isolute SPE cartridges (International Sorbent Technology; Mid Glamorgan, U.K.). A 12-port vacuum SPE manifold (Whatman, Maidstone, U.K.) was used for all off-line SPE procedures, with the pressure maintained below 15 cm Hg. Automated SPE-FI-CL Manifold. A schematic diagram of the automated SPE-FI-CL manifold is shown in Figure 2. Gilson Minipuls 3 peristaltic pumps were used to propel the reagent and SPE solvent streams, while an Ismatec 12 V peristaltic pump (MS/ (12) Sanders, M. G.; Andrew, K. N.; Worsfold, P. J. Anal. Commun. 1997, 34, 13H-4H.
Analytical Chemistry, Vol. 75, No. 11, June 1, 2003
2619
Figure 1. Flow injection-chemiluminescence (FI-CL) manifold for the determination of dodecylamine.
Figure 2. Automated on-line SPE-FI-CL manifold, showing column 1 loading with sample while column 2 is eluting (DCM ) dichloromethane; ACN ) acetonitrile).
CA4-E/8-100C) was used for the sample stream. The autosampler was a 5-port, 4-position electric rotary valve (11526) purchased from Omnifit Limited (Cambridge, U.K.). The sample injection valve (6-port, 2-position [C22-3716EH]), SPE solvent selector valve (11-port, 10-position [C25-3710EMH]), and column selector valve (10-port, 2-position [C22-3716EH]) were Cheminert microactuated valves (Valco Instruments, Switzerland). On-line SPE columns were formed from 5-mm-i.d. chromatography columns (Omnifit Limited, Cambridge, U.K.) packed with 50 mg of Isolute aminopropyl SPE sorbent (International Sorbent Technology, MidGlamorgan, U.K.). Microporous tubing (Global FIA, Washington) was used to remove air bubbles from the dead volume of the SPE column. All pumps and valves were switched using TTL logic controlled with the DAQ-700 card. LabVIEW software (National Instruments) was used to provide a graphical user interface for the data acquisition and the TTL input/output to control the switching valves and pumps. 2620
Analytical Chemistry, Vol. 75, No. 11, June 1, 2003
Procedures. Optimization of the FI-CL Manifold. The manifold shown in Figure 1 was univariately optimized with respect to the concentrations of each reagent, the percentage of water in the acetonitrile carrier stream, and the flow rates of each reagent/ carrier stream. A 10 mg L-1 solution of dodecylamine in 90/10 (v/v) acetonitrile/water was injected five times into the carrier stream, and the mean peak height was calculated. The baseline was recorded for 2 min, and the peak-to-peak noise was used to determine the baseline noise level. A further simplex optimization of the parameters DNPO/sulforhodamine 101 flow rate, hydrogen peroxide/carrier flow rate, and mixing coil length was then performed in the same way. Optimization continued until there was no increase in the signal/noise (S/N) for five consecutive experiments. Solvent Effects. Different ratios of organic solvent/water from 50 to 100% acetonitrile, methanol and propan-2-ol were used as carrier solvents using the FI-CL manifold shown in Figure 1.
Standards prepared in the same solvents (0-10 mg L-1) were analyzed to determine which solvent ratio gave the best S/N. Interferences. Samples containing 40 mg L-1 dodecylamine in 90/10 (v/v) acetonitrile/ water were spiked with varying concentrations (1-1000 mg L-1) of aniline, carbazole, p-cresol, hexanoic acid, indole, and phenol. The FI-CL manifold shown in Figure 1 was used with 90/10 (v/v) acetonitrile/water as the carrier solvent. Direct FI-CL Analysis of Diesel Fuels without SPE. Heptane and diesel samples (unspiked, spiked with 40 mg L-1 dodecylamine, and spiked with the additive package to give a final concentration of 40 mg L-1 dodecylamine) were diluted to 25% (v/v) in propan2-ol. These samples were analyzed using the manifold shown in Figure 1 with a carrier stream consisting of 25/75 (v/v) heptane/ propan-2-ol. Solid-Phase Extraction. The SPE sorbent was first conditioned by sequentially passing small volumes of methanol, toluene, and heptane (4-5 mL each) through the column. Following loading with an aliquot of 40 mg L-1 dodecylamine in heptane, the sorbent was sequentially washed with 2-3 mL of heptane, toluene, and dichloromethane. Finally, dodecylamine retained by the sorbent was eluted using 90/10 and 80/20 (v/v) mixtures of acetonitrile/ water. Extractions were performed in triplicate, and eluates were analyzed using GC-NPD. Automated SPE-FI-CL Procedure. The on-line SPE-FI-CL manifold permitted complete automation of extraction and analysis in a four-step process. A schematic diagram of the automated manifold is shown in Figure 2. • Step 1 (Sample Loop Filling/Column 1 Conditioning/Column 2 Eluting). Initially, the autosampler was switched to the position of the sample to be analyzed, and the sample injection valve was set to the load position. The sample pump was then activated to fill the sample loop. When the loop was completely filled with sample, the pump was switched off. Simultaneous to this, each of the conditioning solvents was pumped though the solvent selector valve for a predefined length of time in order to condition the SPE sorbent. Column 2 was eluted in parallel with column 1 conditioning in order to reduce the time required to perform a series of analyses. • Step 2 (Column 1 Loading/Column 2 Eluting). The solvent selector valve was switched to the sample position, and the SPE solvent pump was activated, forcing the sample in the sample loop through column 1. Column 2 continued to elute during this step. • Step 3 (Column 1 Washing/Column 2 Eluting). Each of the wash solvents was pumped through column 1 for a predefined length of time using the solvent pump. While this was happening, the sample loop was washed with heptane. Column 2 continued to elute during this step. • Step 4 (Column 1 Eluting/Column 2 Conditioning). The column switching valve was then switched, resulting in column 1’s undergoing elution and column 2’s commencing the conditioning, loading, and washing cycle. RESULTS AND DISCUSSION Determination of Dodecylamine Using the FI-CL Manifold. The manifold used by Katayama3,4 was evaluated for the detection of dodecylamine in 90/10 (v/v) acetonitrile/water. When DNPO and sulforhodamine 101 were premixed, they decomposed within 24 h (a complete color change from light pink to yellow
Table 1. Optimized Conditions for SPE and FI-CL Detection of Dodecylamine in Diesel Fuels parameter
optimum condition
SPE
step
vol/mL
solvent
conditioning
1 2 3 4 5 6 7 8
5 5 5 10 2 2 2 1.5
methanol toluene heptane 25/75 (v/v) sample/heptane heptane toluene dichloromethane 90/10 (v/v) acetonitrile/water
loading washing eluting
FI-CL DNPO concn sulforhodamine 101 concn hydrogen peroxide concn water content in acetonitrile stream carrier and peroxide flow rates sulforhodamine and DNPO flow rates
1.0 × 10-3 mol L-1 8 × 10-7 mol L-1 0.1 mol L-1 10% (v/v) 0.75 mL min-1 3.0 mL min-1
was observed). However, without premixing these two reagents, peak heights remained constant, and the individual solutions were stable for at least 1 week. A 2-m mixing coil was therefore inserted to ensure the two reagents (pumped separately) were thoroughly mixed before reacting with hydrogen peroxide. The univariate optimization graphs shown in Figure 3 indicate that all parameters gave clear optimum values. Maximum S/N was observed using 90/10 (v/v) acetonitrile/water, because higher concentrations of water result in a proportional increase in the hydrolysis of DNPO,13 therefore reducing its availability to participate in the CL reaction. Hydrogen peroxide concentration has been shown to be directly proportional to reaction rate for peroxyoxalate chemiluminescence.14 The decrease in CL response observed for peroxide concentrations above the optimum value of 0.1 M is likely, therefore, to be due to the maximum CL intensity occurring before the sample arrives at the flow cell. Optimal settings of flow rates and mixing coil length are also necessary to ensure the CL window (the zone of maximum CL emission) coincides with the time that the sample passes through the flow cell. A simplex optimization was carried out on these physical manifold parameters using the optimum conditions from the univariate optimization as a starting point. The S/N was improved from 112 (arbitrary units) after the univariate optimization to 148 after the simplex optimization (a 32% increase). The optimum conditions were DNPO/sulforhodamine flow rate of 3.0 mL min-1, hydrogen peroxide/carrier flow rate of 0.75 mL min-1, and a mixing coil length of 30 cm. Table 1 summarizes the optimum FI-CL parameters. Solvent Effects. The effect of changing the carrier/sample solvent is summarized in Table 2, with the greatest sensitivity achieved using acetonitrile/water mixtures with g90% acetonitrile. The much lower sensitivity in other solvents, particularly with respect to S/N, was due to increased hydrolysis of DNPO by nucleophilic attack. It has, for example, been reported that 1% (v/v) methanol can result in the chemiluminescence intensity being halved.14 A further reason for the decrease in sensitivity (13) Orosz, G.; Dudar, E. Anal. Chim. Acta 1991, 247, 141-47. (14) Hanaoka, N.; Givens, R. S.; Schowen, R. L.; Kuwana, T. Anal. Chem. 1988, 60, 2193-97.
Analytical Chemistry, Vol. 75, No. 11, June 1, 2003
2621
Figure 3. Univariate optimization of the FI-CL manifold with the following FI parameters: (a) DNPO concentration, (b) sulforhodamine 101 concentration, (c) peroxide concentration, (d) percentage of water in the acetonitrile carrier stream, (e) carrier and peroxide flow rates, and (f) sulforhodamine and DNPO flow rates (error bars are 3s, n ) 5). Table 2. Effect of Different Carrier/Sample Solvents on CL Sensitivity
solvent 50/50 (v/v) acetonitrile/water 60/40 (v/v) acetonitrile/water 70/30 (v/v) acetonitrile/water 80/20 (v/v) acetonitrile/water 90/10 (v/v) acetonitrile/water acetonitrile methanol propan-2-ol
RSD LOD S/N (S/N ) 3), (at 10 (at 10 mg L-1), % (n ) 5) mg L-1 mg L-1) 4.3 2.4 1.5 0.4 0.2 0.6 1.3 0.2
9 31 86 270 1100 1000 26 180
3.3 1.3 1.9 0.5 4.1 4.5 3.5 4.4
seen with methanol and propan-2-ol is that the kinetics of the peroxyoxalate CL reaction are considerably faster in these solvents than in acetonitrile.15 A system optimized for the relatively slow acetonitrile kinetics would, therefore, not give the greatest sensitivity for solvents with different reaction rates. Interferences. The effects of indigenous compounds in diesel fuel on DNPO chemiluminescence are shown in Table 3. Aniline and indole interfered at relatively low concentrations (1000 mg L-1) suppressed the CL emission at much higher concentrations. This is consistent with the above theory, because the electron-donating methyl group in cresol (15) Weinberger, R. J. Chromatography 1984, 314, 155-65. (16) Vanzoonen, P.; Kamminga, D. A.; Gooijer, C.; Velthorst, N. H.; Frei, R. W.; Gubitz, G. Anal. Chem. 1986, 58, 1245-48. (17) Vanzoonen, P.; Bock, H.; Gooijer, C.; Velthorst, N. H.; Frei, R. W. Anal. Chim. Acta 1987, 200, 131-41.
2622 Analytical Chemistry, Vol. 75, No. 11, June 1, 2003
Table 3. Interfering Effects of Indigenous Compounds in Diesel Fuel interfering agent
concn causing interferencea
effect
aniline indole p-cresol hexanoic acid carbazole phenol
